Category: Semiconductor Stocks

Nvidia CEO Predicts AI Spending Will Increase 300%+ in 3 Years

Posted on by io-fund

Nvidia has traversed choppy waters so far in 2025 as concerns have mounted about how the company plans to sustain its historic levels of demand. It began with DeepSeek in late January, was furthered by suppliers providing mixed signals on the timing of its premiere Blackwell NVL systems, then saw rumors of data center cancellations from a major customer in February.

What better place to address these issues than the GPU Technology Conference (GTC) in San Jose, now dubbed the Super Bowl of AI. In the keynote held on Tuesday, Jensen Huang threw cold water on many of Wall Street’s assumptions, helping to alleviate concerns that demand for Nvidia GPUs will slow. In addition, I appeared on Fox News during the keynote to discuss why valuation is the great equalizer for this stock – along with my prediction for which quarter this year Nvidia will likely explode higher.

Nvidia Explains Why Cheaper Models Will Not Result in Less Compute

CEO Jensen Huang kicked the conference off with a wild remark about the current pace of progress in AI and the need for compute: “the scaling law of AI is more resilient, and in fact, hyper-accelerated, and the amount of computation we need at this point, as a result of agentic AI and reasoning, is easily 100x more than we thought we’d need at this time last year.”

The proof of this is easily seen as Blackwell chip sales have significantly outperformed Hopper year-over-year, with 3.6 million GPUs ordered so far in 2025 by the top 4 CSPs, versus a peak of 1.3 million Hopper GPUs in 2024. And this is just sales to the 4 largest CSPs, not including CoreWeave, Meta, xAI, Tesla, Nebius and many others that will be acquiring the chips. Huang added that “demand is much greater than that, obviously” — with the readthrough being this is what they’re able to ship, with demand that exceeds current capacity.

screen shot of Nvidia CEO Jensen Huang at GTC

Nvidia’s Blackwell chip sales so far in 2025 have far exceeded Hopper’s peak. Source: NvidiaNvidia

Huang further illustrated that due to AI being able to reason beyond pretrained data, it now generates more tokens at 10X for a complex model, yet compute has to be 10X faster, resulting in 100X more computation.

“Well, it could generate 100x more tokens and you can see that happening, as I explained previously, or the model is more complex, it generates 10x more tokens. And in order for us to keep the model responsive, interactive so that we don't lose our patience waiting for it to think, we now have to compute 10x faster. And so 10x tokens, 10x faster, the amount of computation we have to do is 100x more easily. “

Models will need to generate more tokens, more quickly; meaning, AI remains a hardware problem that Nvidia is uniquely positioned to solve. The amount of computation required for inference is significantly higher than previously estimated – and it’s this demand that Nvidia’s future generations of GPUs will aim to meet.

Huang Forecasts Capex to Grow more than 300% in 3 Years

Nvidia has been a massive beneficiary of big tech capex budgets. Our firm has been tracking Big Tech capex as a proxy for AI spending since 2022, when I publicly stated in my newsletter: “However, it has been our stance for some time that Big Tech capex is the true leading indicator for AI semiconductor companies. Despite an enormous increase in Big Tech capex primarily driven by data centers, this line item does not get the attention it deserves in terms of follow-through to the semiconductor industry.”

We’ve continuously reminded our readers that data center capex provides visible read-throughs for Nvidia as it captures a lion’s share of that spend, and GTC provided another clear signal that not only is capex not slowingnot slowing as analysts fear, but is accelerating ahead of expectations.

At GTC, Huang pulled forward his view for $1 trillion in data center buildouts, saying he now sees the $1 trillion mark being reached as soon as 2028, ahead of prior expectations for 2030, representing an expansion of Nvidia’s addressable market.

Huang explained that he was confident that the industry would reach that figure “very soon” due to two dynamics – the majority of this growth accelerating as the world undergoes a platform shift to AI (the inflection point for accelerated computing), and an increase in awareness from the world’s largest companies that software’s future requires capital investments.

Nvidia stock CEO Jensen Huang at GTC explains that data center capex is accelerating and could reach $1 trillion as soon as 2028, ahead of prior views for 2030.

Nvidia CEO Jensen Huang predicts data center capex may reach $1 trillion as soon as 2028 as AI drives an inflection in computing. Source: NvidiaNvidia

Not only did Big Tech hit the $250 billion threshold in 2024, but these companies are on track to significantly exceed that in 2025, with Microsoft, Meta, Alphabet and Amazon likely to spend close to $330 billion on capex this year. This is easily more than double what was spent in 2023, and as whole, that represents 33% YoY growth for the four purchasing Blackwell en masse.

Based on Huang’s prediction that data center expenditures could reach $1 trillion by 2028, that’s 3x growth in 3 years, and Big Tech alone (not even including Oracle and others) is already at one-third of that this year.

Graph showing Big Tech capex surging 33% YoY in 2025, on track to reach to $330 billion.

Big Tech’s capex is on track to approach $330 billion in 2025, up 33% YoY and more than double what was spent in 2023. Should Huang’s prediction prove true, it will represent 300% growth in the AI DC infrastructure market in three brief years.

China’s tech firms are also quickly raising capex to remain competitive in the global AI war, with Alibaba signaling capex of $52 billion over the next three years, more than what it has spent over the past decade, while Tencent outlined faster capex growth as it purchases more AI chips. I have said previously on Fox Business News that AI spending goes up in times of war – and neither China nor the US will want to lose to the other when it comes to AI dominance.

The I/O Fund specializes in covering lesser-known AI stocks on our research site with trade alerts and weekly webinars. Learn more here.The I/O Fund specializes in covering lesser-known AI stocks on our research site with trade alerts and weekly webinars. Learn more here.here.

Huang Explains Why Nvidia’s GPUs will Remain in High Demand

The breakthroughs we’ve seen in recent months and the rapid progression to complex problem solving and reasoning are increasing token usage by 100x and resulting in 10x faster computing power required to power the next stages of AI.

Tokens are the core factor going into the economics of an AI model – tokens for training represent the core part of the model costs, while tokens for inference generate revenue and thus profit. In a demo at GTC, Nvidia showed that for a complex problem with multiple constraints, a reasoning model like DeepSeek’s R1 would reason through the possibilities and answer with 20x more tokens using 150x more compute than a traditional model like Meta’s Llama 3.3-70B.

Translating this to the data center shows why Blackwell is in such high demand, to the tune that it has sold more than 2.5x as many GPUs already in 2025 versus Hopper’s peak. With Blackwell, which delivers up to 30x faster performance on inference versus the HGX H100, at 116 tokens per second per GPU versus 3.5 tokens per second, with 25x better energy efficiency. For a reasoning model, Huang explained that with Nvidia’s new Dynamo inference serving library, Blackwell can deliver up to 40x performance for reasoning models.

Here's why this is important. We explained last week in a brief writeup Unlocking the Future of AI Data Centers: Which Fuel Source Reigns Supreme in Efficiency? that power was the core chokepoint and the key enabler for AI’s future, as AI cannot exist without new sources of electricity to power its applications. Huang highlighted this at GTC, explaining that data centers are power limited, meaning revenues are power limited, hence why customers are looking for the most energy efficient chips they can get.

A 100MW data center (which is becoming more commonplace for hyperscalers) could house 1,400 H100 NVL8 racks and produce a maximum of 300 million tokens per second. With Blackwell, the same data center could house 600 racks but produce a maximum of 12 billion tokens per second, in theory a 40x increase. Increased inference performance leading to higher token outputs both lowers costs and increases revenue potential – Nvidia pointed out that DeepSeek-R1 based software optimizations improved token output and revenue generation by 25x and lowered inference costs by 20x.

While these maximums are theoretical in nature, the underlying notion that a data center can serve substantially more tokens at a lower cost supports Blackwell’s high demand, from a superior TCO profile and increased revenue generating ability.

Larger (and more) data centers expand the opportunity ahead for Nvidia – in the follow-up analyst call at GTC, Huang explained that “every gigawatt [of data center] is about $40 billion, $50 billion to Nvidia.”

According to CBRE, approximately 9.5 GW of data centers have gone under construction since the start of 2023. given an average construction timeline of 18 to 36 months (depending on constraints such as power supply), Huang’s comments imply a $380 billion to $475 billion revenue opportunity over the next 1 to 3 years just from that existing footprint under construction since 2023. We’ve already seen large data center announcements in 2025, with construction on the first $100 billion data center for Stargate commencing and Crusoe securing 4.5GW in natural gas for future data centers.

Upcoming GPU Roadmap Positions Nvidia to Capture $1T Data Center Spend

Nvidia is continuing to move at a break-neck pace when it comes to upgrading its GPU lineup, and maintaining this rapid release cycle is allowing it to continually pry away Big Tech’s capex year after year due to the performance, energy and TCO advantages each generation offers over the last.

At GTC, Huang unveiled Blackwell Ultra, the GB300 lineup, Vera Rubin and Vera Rubin Ultra, Blackwell’s successors, and an initial view at Feynman, Rubin’s successor.

GB300 NVL72 Delivers 1.5x Performance Upgrade

Notably, Nvidia provided little mention of the GB200 NVL72 during the keynote and offered no concrete evidence of shipping timelines for the superchip, opting to discuss Blackwell Ultra instead.

Blackwell Ultra, the GB300 NVL72, is due in the second half of 2025, with Huang expecting a smooth transition to the upgraded platform. The GB300 NVL72 provides up to a 1.5x performance boost versus the GB200 and delivers 50% more FP4 dense compute with a 50% boost to memory capacity, both of which will increase inference throughput.

Rubin Offers 3.3x Boost to GB300

Nvidia’s Vera Rubin NVL144 is scheduled for release in the second half of 2026, a year after the GB300 NVL72. Rubin is expected to be “drop-in compatible” to existing Blackwell infrastructure and offers up to a 3.3x boost to FP4 inference performance versus the GB300, with 3.6 exaFLOPs compared to 1.1 exaFLOPs.

Per chip, Rubin offers 50 petaFLOPs of FP4, up 2.5x from 20 petaFLOPs for Blackwell. Rubin also marks a shift to HBM4 memory, while remaining at 288 GB capacity.

Rubin Ultra Sees up to a 14X Increase in Inference Performance

Perhaps the largest boost in performance comes with Rubin Ultra NVL576, set to be released in the second half of 2027. Nvidia says the upcoming platform will offer up to 15 exaFLOPs of FP4 inference performance, a more than 4x increase from Rubin and nearly 14x increase from the GB300 in just two years.

While this leaves much for the supply chain to address in a short period of time (as we know Nvidia likes to break the limits of what’s possible), Nvidia is proving that it remains committed to the two things that matter most as AI continues to scale past generative AI to agentic AI and physical AI – it will continue to significantly boost inference performance via hardware improvements and software optimizations and reduce costs and thus TCO for its customers.

Put simply, data centers can handle more inference requests, process more tokens, and make more in revenue with each upgrade with the same power requirements.

Nvidia’s Valuation is the Equalizer

The major takeaway from GTC is that we’re only on the very brink of what AI can ultimately achieve. The need for compute will continue to rise as the industry progresses from generative AI to advanced reasoning models, to comprehensive AI agents, to autonomous vehicles and robotics where real-time inference is an absolute necessity for split-second decision making.

I spoke with Charles Payne on Fox Business News live during GTC to explain why I believe that the event’s major takeaway is that GPU demand is secular, not cyclical. I explained that Huang is “answering for investors why Nvidia’s GPUs will remain in demand. It does not matter if cheaper models are run on a single GPU, because ultimately, for these advancements to continue, we need to see that 10x in [faster] computing power, and we all know which company will serve that demand.”

Huang put it quite simply: “every single company only has so much power. And within that power, you have to maximize your revenues, not just your cost.”

While a lack of clarity and little mention of the GB200 NVL72’s timing during the keynote was likely a factor behind the muted stock price reaction, I would argue that Nvidia’s stock is absurdly cheap ahead of Q3 and Q4’s volume ramp.

Graph of Nvidia stock's forward P/E ratio showing stock is trading at the same valuation level as prior to Hopper's breakout May 2023 quarter. Source: YCharts.

Nvidia is trading at 26.5x forward earnings with growth of over 51% expected this year. Source: YChartsYCharts

Nvidia is currently trading at 26x this fiscal year’s earnings with earnings growth forecast to be 51.5% to $4.53, and at 20x next year's with 27% growth to $5.76. That 26x multiple is nearly a 25% discount to Nvidia’s average forward PE ratio over the past two years, and the same multiple it commanded before May 2023’s Hopper-driven breakout quarter.

Conclusion

Although there are many details from Nvidia’s GTC conference keynote worthy of discussion — Big Tech capex is the single most important point for investors as the sheer amount of capital pointed at data center infrastructure from a handful of companies is truly unparalleled in the history of the markets.

We’ve continuously reminded our readers that data center capex provides visible read-throughs for Nvidia as it captures a lion’s share of that spend, and GTC provided another clear signal that not only is capex not slowingnot slowing as analysts feared but is accelerating ahead of expectations.

In the more immediate term, we have mixed signals from suppliers on the exact timing of Blackwell’s GB200 NVL72s. The premiere SKU was originally expected to ship in volume in Q1 and that did not happen. Going into the February earnings report, I stated my spidey senses were up in the article “Nvidia Suppliers Send Mixed Signals for Delays on GB200 Systems – What It Means for NVDA Stock and cautioned the earnings report was unlikely to offer the blowout that investors have become accustomed to. This was despite Wall Street growing exuberant into the print and aggressively raising price targets.

Later, I/O Fund Portfolio Manager Knox Ridley stated that if Nvidia breaks $123-$119, the stock would likely find support between $102 and $83. This scenario remains a possibility given the weakness we have seen in the broad market. With that said, we see any dips on Nvidia as a buying opportunity as the stars are aligning for Q3-Q4 in terms of volume shipments on the Blackwell and Blackwell Ultra GPUs.

The I/O Fund has a strong track record on this stock, discussing every twist and turn publicly for our free stock newsletter readers with documented gains of up to 4,100% as far back as 2018 based on a very-early AI thesis. The I/O Fund sends real-time trade alerts for every entry and exit, and our research members will be notified via text when we deem the risk/reward favorable and resume buying Nvidia. Learn more here.

Disclaimer: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund own shares in NVDA at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

Nova Limited: Riding the AI/HPC Wave with Advanced Nodes and Packaging

Posted on by io-fund

Key Takeaways:

  • Nova's metrology solutions are poised for long-term growth as AI/HPC drives the demand for advanced nodes and advanced packaging solutions.
  • Gate all around (GAA) presents a catalyst for NVMI as Taiwan Semiconductor moves toward a new advanced packaging architecture.
  • Over 100 new fabrication facilities will be built globally by 2030, according to Jeffries, buoyed by NVIDIA and AMD shifting to annual GPU releases.
  • Nova experienced a revenue and earnings growth spurt starting in Q3 2023, potentially hitting a peak in its Q1 2025 guide, as analyst estimates indicate a flattish plateauing year with rangebound revenue and EPS.
  • Nova collected 39% of its total revenue from China in 2024, but that will shrink due to the growth in its advanced nodes business, which China lags due to U.S. trade restrictions.
  • Nova’s headquarters are based in Rehovot, Israel, located 20 km from Tel Aviv, making it susceptible to geopolitical risks including the ongoing Israel-Hamas war and tensions with Hezbollah in Lebanon.

Nova Limited (NASDAQ: NVMI) is a leading provider of metrology tools for advanced process control in the semiconductor manufacturing industry. The company primarily focuses on dimensional and materials metrology and inspection solutions. Its tools are a necessity for chip manufacturers, and their business correlates to the supply and demand trends of the semiconductor industry.

The AI revolution is providing an extended runway as the need for metrology grows with the evolution of more advanced chips that are required for artificial intelligence (AI) and high-performance compute (HPC) applications. Nova’s metrology solutions are applied to advanced logic (AI-enabled), memory and advanced packaging.

Advanced Nodes Will Drive Growth

As AI drives the need for more powerful and efficient chips, manufacturers are scaling up their designs by making them more complex in terms of size, materials, and packaging. This complexity means that chips are becoming harder to produce, and even small deviations during manufacturing can affect the yield. Therefore, precise process control becomes critical, which in turn increases demand for metrology equipment.

I/O Fund pointed out that the evolution to advanced nodes is what will drive the AI boom and demand for metrology in our article, “Nova and Onto Innovation: Growth in Metrology and Semiconductor Process Control.”

“Advanced nodes require more process steps as node sizes shrink, so for a chipmaker or foundry like TSMC or Intel to move from primarily producing on 5nm nodes to 2nm or below over the next couple of years, there will be a greater need for metrology equipment. Currently, the 5nm and 4nm nodes are primarily being used for AI chip production, such as that for Nvidia’s Hopper and Blackwell chips, while 3nm production is ramping at TSMC, with volume production at the 2nm node expected in 2025, primarily for smartphone applications. This is because the manufacturing tolerances shrink as nodes shrink in size – the chipmaking process now becomes increasingly more sensitive to minute deviations in the process. Moving to more advanced nodes warrants much greater precision throughout the entire manufacturing process, as the smallest of deviations could greatly affect the process yield.“

Moreover, the frequency of accelerated chip development timelines is an added boon for the urgency and demand for Nova’s metrology tools — “Nvidia and AMD are both shifting to annual release cadences, aiming to bring next-generation GPUs to market once per year, compared to prior cadences of every two years. This is a major technological feat – as we had said previously for Nvidia, it’s a ‘move-fast-break-things’ problem, where Nvidia is pushing the boundaries of what had previously been seen as impossible in the chip industry. By moving to these quicker release cycles, there’s a much greater emphasis on metrology and process control to ensure that the manufacturing process remains sharp while also ensuring a faster ramp and high yields to meet mass production thresholds and demand.”

The increasing complexity of chips is driving higher metrology intensity. Gate all-around (GAA) field-effect transistors (FETs) require 30% more metrology steps. High bandwidth memory 3 extended (HBM3E) used for AI, ML, graphics processing, and scientific computing consumes 3X more wafer supply as double data rate 5 (DDR5) SDRAM used for mainstream computing applications on desktops and laptops. There are over 100 fab projects planned globally, supported by over $300 billion in funding and incentives by 2030. The U.S. alone has 28 fabs costing around $52 billion.

Advanced Packaging Revenues Doubled in 2024, Driven By AI/HPC Demand

AI and HPC workloads drive the need for advanced nodes, which deliver higher performance and energy efficiency. However, as chips become denser and generate more heat, advanced packaging techniques become essential to manage thermal challenges and improve overall chip performance. Nova’s 2024 advanced packaging revenues more than doubled YoY. It now contributes 15% of product revenue, and its integrated metrology solutions have been adopted by four of the top five advanced packaging manufacturers.

CEO Gabriel Waisman addressed the areas that boosted their advanced packaging segment in 2024 during the Q4 conference call. He stated this.

“So first, the advanced packaging had contribution from both our chemical metrology division as well as the dimensional metrology division. We have our integrated metrology in all of the top five advanced packaging manufacturers, and we have a significant adoption of our PRISM standalone OCD platform. So it's both divisions that contributed to this growth. And we do expect this year to expect to continue and grow by double-digit growth.”

Gate All Around (GAA) Presents Catalyst for Nova

In June, we covered how TSM is moving from FINFET transistors to gate all around (GAA), stating “with FinFET, the gate is wrapped on three sides, whereas with gate-all-around (GAA), as the name implies, the gate is wrapped around on all sides. FinFET is used in 14nm, 10nm and 7nm nodes. TSMC uses FinFETs in the 5nm, yet will phase out FinFET after the 3nm. As TSMC moves toward GAA for the 2nm, having the gate wrap “all-around” will create a greater surface area for better electrostatic control and to also reduce leakage.”

The write-up also pointed out: “The 2nm will be the first node to use gate-all-around field-effect transistors (GAAFETs), which will increase chip density. The GAA nanosheet transistors have channels surrounded by gates on all sides to reduce leakage, yet will also uniquely widen the channels to provide a performance boost. There will be another option to narrow the channels to optimize power cost. The goal is to increase the performance-per-watt to enable higher levels of output and efficiency. The N2 node is expected to be faster while requiring less power with an increase of performance by 10%-15% and lower power consumption of 25%-30%.

For TSMC, the 2nm will feature NanoFlex technology, which is similar to FinFlex to where designers can use cells from different libraries. However, due to the new gate-all-around (GAA) nanosheet transistors, there are additional benefits, such as customizing the width and height of cells.

Intel’s 20A will be the first to feature backside power delivery for faster switching and to alleviate routing congestion. With this release, Intel is introducing the “angstrom” era” which translates to future process generations where the process nodes are not smaller necessarily, rather the transistors they’re built with will be improved upon. For Intel, instead of the GAAFET, the company is introducing RibbonFET transistors where multiple flat nanosheets are stacked to enable better current flow.”

CEO Waisman committed to $500 million of GAA revenue from 2024 to 2026. If our math is correct, this can be achieved in one of two scenarios:

  • A two-year time frame which includes 2024 GAA revenue of $45 million and 2025 GAA revenue would have to be $455 million to arrive at the $500 million commitment from 2024 to 2026, implying a 10X surge in 2025 GAA revenue.
  • A three-year time frame which includes 2024 GAA revenue of $45 million, 2025 GAA revenue of $90 million and 2026 GAA revenue of $365 million to arrive at the $500 million commitment from 2024 through 2026, which would imply a 2X and then 4X ramp in GAA revenue in 2025 and 2026, respectively.

In the latter three-year scenario, GAA revenue would represent 8.3% in 2024, 13.4% in 2025 and 50.8% in 2026 of total product revenue based on consensus analyst estimates, as GAA adoption surges in 2026. Either scenario underscores the point that GAA is shaping up to be a major catalyst for Nova.

What is Semiconductor Metrology?

Semiconductor metrology uses precise measurement techniques to inspect wafers for defects and contamination, ensuring process control during manufacturing. Its main goal is to identify and locate issues so engineers can address them, thereby maximizing chip yield, which is the percentage of functioning chips produced from a wafer.

Higher Chip Yields are the Holy Grail of Efficient Semiconduction Manufacturing

 Higher yields are essential for reducing costs by producing more functional chips from the same raw materials, which is a critical edge and key driver for Nova. Their metrology tools, including chemical and optical instruments help major customers like Taiwan Semi overcome yield challenges at advanced nodes. Advanced semiconductor packaging technology like chip-on-wafer-substrate (CoWoS), which was crucial to scaling NVIDA’s Hopper and Blackwell production, add further complexity. Taiwan Semi’s 3nm node only hit yields of 50% to 60% in 2023, but has reportedly achieved 3nm yield of over 90% and 2nm yield of 60%. Nova's role in optimizing CoWoS yields becomes a linchpin for success—more chips, lower costs, and higher margins.

As chip features shrink, tighter tolerances increase defect risks, making advanced metrology essential for quality and cost efficiency. While traditional applications face seasonal slumps, secular growth in AI is turbocharging demand for Nova’s metrology solutions across all semiconductor segments.

Nova offers many types of metrology solutions:

  • Dimensional Metrology measures the physical dimensions of semiconductor structures to ensure the accuracy of features on chips. Nova Fit Series utilizes optical techniques to measure critical dimensions (CD), including weight, height and side wall angle in 3D structures.
  • Materials Metrology provides insight into the material properties impacting device functionality. Technologies included X-ray photoelectron spectrometry (XPS), X-ray fluorescence (XRF), and secondary ion mass spectrometry (SIMS). Products include VeraFlex, Elipson and Metrion.
  • Chemical Metrology focuses on the chemical composition and purity of materials focusing on the presence, concentration, and distribution of chemical species, including dopants, impurities, and contaminants. Its technologies often overlap with materials technology through techniques like secondary ion mass spectrometry (SIMS) with platforms like Metrion. The Metrion platform offers SIMS capabilities for in-depth chemical analysis, which is vital for understanding dopant distribution and detecting impurities.
  • Spectral Interferometry uses light interference to measure depth, thickness and properties of thin films and complex 3D structures, especially for probing vertically stacked layers. Nova Prism uses this for optical CD metrology.
  • Optical Scatterometry analyzes how light scatters off the patterned structures to infer dimensions and shapes to measure critical dimensions of periodic structures and wafers. Nove MMSR+ uses this for high-precision measurement of CD and thin films.
  • Advanced Imaging involves capturing high-resolution images to analyze defects, patterns and material properties in combination with other metrology techniques for comprehensive analysis. Nova T600 integrated advanced imaging for better pattern recognition and precision alignment.
  • Hybrid Methodology combines different metrology techniques from various toolsets, including optical CD, atomic force microscopy (AFM), and scanning electron microscopy (SEM) to enhance accuracy. It’s used to measure parameters that are too difficult with just a single method. Nova's Hybrid Metrology Solutions uses a hybrid approach (IE, integrating spectral interferometry with scatterometry) to enhance measurement accuracy for parameters that single methods can't capture.

Taiwan Semiconductor Faces Yield Issues; Advanced Packaging Alleviates the Problem

Taiwan Semi’s CoWoS advanced packaging technology can improve chip yields, but it’s more of a double-edged sword. It boosts yields but introduces new yield challenges during the actual packaging process. CoWoS involves stacking multiple dies like (IE: GPU + HBM memory) on a silicon interposer and mounting that on a substrate. This is key for high-performance chips like NVIDIA’s Hopper and Blackwell GPUs.

CoWoS lets Taiwan Semi use smaller and higher-yielding dies instead of a single giant chip, two dies vs one monolithic chip. If one die fails, it can be swapped out (before stacking), rather than having to replace the whole chip. The challenge is that stacking dies on interposers is more complicated (IE: multi-die stacks heating unevenly causing warping, which can cut yields up to 10%). This is where Nova’s metrology tools, like PRISM II and ELIPSON, step in to enhance CoWoS yields, pushing them above 90% on mature production runs.

We’ve also broken down the importance of CoWoS-L capacity in clearing Blackwell bottlenecks here.

China Generates the Most Revenue, But That Will Be Shrinking with Advanced Nodes

In 2023, China generated 30% of total revenue. In full year 2024, that percentage climbed to 39%. However, growth will come from advanced nodes, so the China share is expected to decline. Nova CEO Gaby Waisman confirms this point.

“Sure. So, in 2024, the China share of our overall sales was 39%. Our strength there is in line with industry peers. And due to the fact that growth this year will come from advanced nodes, we see the share of China declining.”

Since China’s access to advanced nodes is limited due to trade restrictions, they lag behind leading-edge manufacturers like Taiwan Semi and Samsung. This is a positive as it enables more geographic and technological diversification. Advanced nodes are the future and generate strong margins. It reduces Nova’s dependence on any single market or technology, which captures opportunities in higher-growth and higher-value segments.

China's lag in advanced semiconductor nodes is largely due to U.S. trade restrictions that prevent ASML, the sole manufacturer of EUV lithography machines needed for chips at 7nm and below, from selling to China. These restrictions, influenced by U.S. policy, have left China using less advanced deep ultraviolet (DUV) systems, resulting in a technological gap where their most advanced chips are still at the 7nm node, about five years behind the global frontier. Incidentally, TSMC has also halted producing 7nm AI chips for Chinese customers, including Baidu, Alibaba and ByteDance, as of Nov 11, 2024. Any future AI chip production will need U.S. approval.

Financials: Growth Spurt Driven by AI/HPC Demand. Are Analysts Asleep at the Wheel?

Nova had a record 2024 driven by the AI boom. However, indications appear that a flattish 2025 is on the horizon. Nova experienced a growth spurt that started in Q3 2024, peaking out by Q1 2025, as it flattens out in 2025, according to analyst estimates. Nova only provided Q1 2025 guidance. They don't provide full-year guidance. The bump up in Q1 2025 is helped by the accretive nature of the Sentronics acquisition, which generates an estimated $20 million annually. Nova will start to add Sentronics revenue into the Company’s earnings starting in Q1. Nova reported a record Q4 and 2024 revenue powered by record sales of material metrology and dimensional metrology solutions. While analysts still forecast 25.26% YoY revenue and 23.09% YoY EPS growth rate in 2025, QoQ growth indicates a plateau.

Revenues Surge to All-Time Highs, But 2025 Analyst Estimates Indicate a Flattish Year

Q4 revenue grew 45.11% YoY and 9% QoQ to a record $194.77 million, beating consensus analyst estimates by $8.28 million or 4.44%. The revenue beat was attributed to record strength in its materials metrology portfolio driven by robust sales of the VeraFlex, Elipson and Metrion platforms augmented by record sales of their dimensional standalone OCD solutions that saw heightened demand from GAA and advanced packaging solutions.

Management guided Q1 2025 revenue of $205 million to $215 million, with a midpoint of $210 million, representing 48.09% YoY growth. Full year 2024 revenue rose 30% YoY to 672.4 million. Q1 2025 will include Sentronics revenues, which are estimated to be around $20 million annually or an additional $5 million per quarter. The geographic revenue split in 2024 was: China generated 39%, Taiwan had 20%, Korea had 18%, the U.S. had 14% and other territories contributed the remaining 9%. 

Nova’s revenue will have grown for nine consecutive quarters, potentially peaking out in Q2 2025 at $213.8 million, according to consensus estimates. Analyst estimates for FY 2025 indicate flat revenues through Q1 2026 hovering between the $210 to $213.80 million level per quarter, despite 2025 YoY growth estimated to fall to 25.28%, which echoes the sentiment for many other component suppliers like Monolithic Power acknowledging a slow start and potential flattish 2025 as it pertains to AI and data center growth. Other suppliers like Vertiv have issued contradictory guidance indicating a slowdown as the year progresses (perhaps due to a softer Q2).

Revenues Split Between Product and Service Sales

Nova generates dual revenue streams through two segments: Products and Services. The Products segment includes sales of all the platforms, tools and systems. Service revenues include installation, training, maintenance, customization, support and upgrade services.

Products revenue rose 52.3% in Q4 to a record $158.55 million due to the adoption of Nova’s metrology solutions for logic for AI applications, advanced packaging and memory technologies like HBM. Product revenue distribution of 72% from logic and foundry and 28% from memory. Product revenues included three customers and four territories, which contributed each 10% or more. The principal customers come from Taiwan (Taiwan Semi), South Korea (Samsung), China (Semiconductor Manufacturing International Corporation) and the United States (Intel). Growth went from being down (15.7%) in Q4 2023 to consistent quarterly improvements to close the year with record revenues.

Services revenues rose 20.3% YoY to $36.2 million in Q4 driven by the increasing utilization of tools and expansion of Nova’s customer base on ongoing service contracts. Nova has over 6,400 active installed bases at over 400 customer sites. The Service division delivered record results, with 2024 revenues up 19% YoY due to increased capacity demand and yield improvements. The market remains robust, driven by mobile and AI demand and investments in advanced logic, DRAM, and packaging, with wafer front-end (WFE) expected to grow at mid-single digits this year. Management expects 10% to 15% growth in 2025.

Non-GAAP EPS: Solid YoY EPS Growth Peaks by Q1 2025 and Decelerates in 2025

Nova reported Q4 non-GAAP EPS of $1.94, beating consensus estimates of $1.82, by $0.12 or 6.5%. Non-GAAP EPS rose 42.65% YoY. Interest income for the quarter fell 48.4% YoY to $3.76 million, yet GAAP EPS rose 31.67% from $1.20 to $1.58.

Management guided Q1 2025 EPS to $2.00 to $2.16, with a midpoint of $2.08, which would beat analyst estimates by $0.01, indicating 49.6% YoY growth.

Nova experienced a growth spurt that started in Q3 2024, expected to peak by Q1 2025 driven by AI/HPC chip demand as it flattens out in 2025. Non-GAAP EPS peaks at $2.08 in Q1 2025 guide, as analyst estimates call for a sequential drop to $2.03 by Q1 2026, down -1.92% YoY.

Margins: Consistent Gross and Operating Margins

Nova has done a good job holding the line with margins, as they remained mostly flat in 2024. Non-GAAP gross margins for Q4 were 58%, down from 61% in 1H, but still high enough to indicate strong pricing power and operational efficiencies. The target gross margin is greater than 60%. Non-GAAP operating margins in Q4 were 28%, relatively flat throughout 2024.

Improving Cash While Chipping Away at Debt

Nova closed Q4 2024 with $820 million in cash and cash equivalents, up 27.9% YoY, while chipping away at debt close Q4 2024 at $180.6 million, down 8.65% YoY.

Conference Call: Growth Spurt in 2024, Managements Sees It Continuing in 2025

CEO Gary Waisman noted they are encouraged by the broad adoption of Nova’s portfolio across gate-all-around (GAA) and high-bandwidth memory (HBM) processes. Looking forward, Nova is poised to leverage the transition into advanced manufacturing processes and architectures. They expect growing exposure to new market segments and their differentiated portfolio to drive sustained growth into 2025, continuing the momentum from 2024.

Their standalone optical critical dimension (OCD) solutions had a record year, increasing market share as the Nova PRISM platform delivered high double-digit year-over-year growth. This success was driven by the platform's superior productivity and precision, appealing to both front-end and advanced packaging customers. To meet the rising demand for productivity and yield improvements, they launched Nova Velocity, a next-generation dual-chamber platform that offers the highest productivity in the market. Its speed and robustness have already secured a multi-tool purchase from a leading logic manufacturer, highlighting its ability to deliver innovative, high-yield solutions.

The surge in AI-related demand has been a significant driver as it necessitates energy-efficient computing power and accelerates the demand for advanced processing nodes and memory solutions.

Nova’s customers are leading the transition to 3D architectures, which translate into multiple catalysts for the business, including larger and more complex dies that require a growing number of wafers, a higher number of layers and a leap in the number of process steps, at a much smaller tolerance for error.

Leading foundries, logic, and memory manufacturers are increasingly adopting multiple Nova solutions from their optical dimensions, materials, and chemical metrology portfolio. This widespread adoption demonstrates their ability to meet the complex metrology challenges of advanced semiconductor nodes. Their solutions for 2.5D and 3D applications enable customers to achieve the precision and efficiency required for current and next-generation technologies.

Their materials metrology portfolio delivered record quarterly and annual results. The Metrion platform was adopted by a leading global memory customer for advanced DRAM R&D and high-volume DRAM and NAND production thanks to its high sensitivity and precision in full-wafer epitaxial layer measurements. Additional orders from this customer are expected, and two top memory and logic customers are evaluating the platform. Meanwhile, the fourth-generation VeraFlex platform has been widely adopted by several leading foundries and memory customers, and the Nova Elipson platform performed strongly with repeat orders and penetration into two new major customers.

Nova closed the Sentronics Metrology GmbH acquisition deal on Jan 30, 2025. Sentronics develops modular multi-sensor platforms with proprietary sensors and software that expand our solution capabilities. These platforms are critical for advanced packaging, measuring total thickness variation, surface roughness, and wafer bow and warpage.

Nova expects it to be accretive on a non-GAAP net earnings basis within 12 months of closing. Q1 2025 forecast includes the revenues from the relative period Sentronics will report on Nova. Sentronics had about 10% of a $200 million TAM in 2024, which leads to a total revenue of around $20 million annually or $5 million per quarter. This was gathered from CEO Gary Waisman’s comment here:

“So, as I mentioned, the first quarter forecast includes the revenues from the relative period Sentronics will report on the Nova. And you can deduct from the fact that Sentronics had about 10% of a $200 million TAM market last year as to the level of business that we expect in — especially at least in the first quarter.”

When asked about the demand in 2025 for memory versus logic, Waisman responded.

“So, we do expect advanced logic and advanced packaging to lead the pack in 2025 with the growth. We definitely see the HBM segment as part of the advanced packaging growing with the metrology intensity as well. But the bottom line is it's definitely advanced logic and advanced packaging.”

A leading memory manufacturer selected Metrion.

Mark Millar of The Benchmark Company asked where Nova was seeing significant share gains and in which markets. CEO Waisman responded with this:

“So first of all, in terms of the PRISM, standalone OCD, we saw share gains in both advanced manufacturing as well as advanced packaging. We saw an increase in market share on the front-end side of the Chemical Metrology portfolio. We saw obviously high utilization and additional adoption of the XPS tools in Material Metrology. And we also gained some share on the integrated Metrology, especially as we entered into the advanced packaging space, both in the 2.5D architectures and logic as well as in high-bandwidth manufacturing.”

Waisman clarified that memory sales should be looked at by category—specifically DRAM versus NAND—rather than just comparing HBM to NAND. He explained that DRAM sales are significantly stronger than NAND sales, and within DRAM, high bandwidth memory (HBM) makes up the majority of their business. Moreover, HBM is growing at a faster rate compared to overall DRAM.

Charles Shi of Needham asked about the flatline number and 25% YoY growth estimates in 2025, with wafer front-end (WFR) growth in mid-single digits. Shi asked what the reasons are to believe they will continue 2024’s growth in 2025.

Waisman answered, “I think it has to do with two — three major issues. One is the position that we have, especially with the unique value driven by the technologies that we offer. The second one is expanding our position into advanced packaging and seeing higher adoption. And the third one, of course, which drives that as well, is the Sentronics acquisition that gives us an opportunity to expand to additional customers than the ones that we are exposed to-date. I would top it all, of course, by the fact that we have a strong position in advanced logic, and that gives us grounds to believe that we have the fundamentals, the infrastructure in order to drive the growth into 2025.”

CEO Waisman had confirmed committing that gate-all-around (GAA) revenues are expected to grow cumulatively to $500 million from 2024 through 2026. He stated this.

“I'm not sure I can add more to the fact that we are committed to the $500 million from gate-all-around until 2026. We haven't changed our position in that respect. We will try to give more color during the Investor Day on March 17. But I think that the tracking that number gives a lot in terms of our confidence in making it this year as well.”

This implies a solid growth driver as the adoption of GAA is clearly accelerating as it was 8-9% of total 2024 Product revenues equating to $45 million at midpoint. A doubling conservatively implies $90 million-ish in 2025 or 13.4% of Product revenue based on 25% YoY total revenue growth in 2025. Analyst Charles Shi commented on this.

“Based on your latest reporting, it sounds like, it's $40 million-ish. I think most of your peers are guiding gate-all-around revenue doubling, but your guidance seems to suggest that's a little bit more than doubling for you guys”

This leaves 2026 GAA revenues of $365 million (to complete the $500 million cumulative revenue “commitment” from 2024-2026) implying its growth to 50.8% of total Product revenues in year 2026, using the consensus analyst estimates for full year 2026 revenue of $896.75 million.

Valuation:

The flat/plateaued 2025 assumption is based on consensus analyst forecasts, which had dead-on accuracy for Q1 2025 as management’s midpoint revenue guidance of $210 million matched consensus analyst estimates for $210.1 million. The consensus analyst non-GAAP EPS estimates peak at $2.07 in Q1 2025 versus the $2.08 midpoint guide by the Company and progressively decline for the next four quarters to $2.03 in Q1 2026. Estimates reaccelerate in Q2 2026 at $2.18, then $2.25 for Q3 2026 and $2.34 for Q4 2026. While YoY growth analyst estimates forecast 25.26% revenue and 23.09% non-GAAP EPS growth, the growth during the year indicates a plateau with consensus analyst revenue estimates rangebound between $210 million to $213.8 million.

Either the analysts are asleep at the wheel with their growth estimates or are taking a wait and see approach or expect a flattish 2025 and reacceleration in Q2 2026. Metrology systems providers (semiconductor equipment makers) tend to be a leading indicator compared to power and cooling technology providers, although can be off cycle due to the lumpiness of capex spending from foundries.

It could also coincide with what other AI and data center suppliers are hinting at: a slowdown in 2025 and a reacceleration in 2026. Nova shares trade at a P/E of 47.56 compared to its five-year median P/E of 34.88 and a P/S of 13.06 compared its five year P/S of 6.96. Nova has doubled their revenues every five years since 2007. At the current elevated price levels, it may be prudent to wait and see what further guidance the Company provides in its Q1 2025 earnings release during market hours on May 9, 2025, and its Investor Day on May 16, 2025.

Conclusion:

Nova’s solutions are gaining traction, and the growth catalysts are evident. AI chips are power hungry beasts that advanced packaging technology like CoWoS helps to feed by enabling more performance in less space. Higher yields keep costs down, but the bottleneck occurs when packaging yields lag front-end wafer yields choking off supply. Nova’s metrology tools like its Prism 2 optical critical dimension (CD) platform and Elipson chemical profiling improve interposer alignment and film thickness control in CoWoS packaging. This contributes to higher yields potentially hitting up 90% or greater on mature runs at the foundry level. We can surmise their three principal customers and the four top regions coming from Taiwan (Taiwan Semi), South Korea (Samsung), China (Semiconductor Manufacturing International Corporation) and the United States (Intel).

Management was upbeat about the growth momentum continuing into 2025 but only provided Q1 guidance, as they don't provide full-year forecasts. The Q1 guidance indicates revenue growth of 48.09% YoY at the $210 million midpoint. Around $5 million of that is estimated to be from the addition of Sentech revenues. Customer concentration is less of a concern with a company that supplies foundries, as the total number of foundries globally is limited. However, there is an overallotment of clients in Asia, where most of the world’s foundries are located. Nova is highly exposed to Asia, which comprised 77% of total revenue in 2024, led by China at 39% of total revenues. Nova has a handful of large customers. Three customers comprise at least 30% of the Product revenues. Nova has maintained consistent margins, and its debt-to-equity ratio is 0.19. The company has a war chest of $820 million in cash and cash equivalents.

Welcome to the I/O Fund’s new Discovery Tier, where we cover a new stock idea on a weekly or bi-monthly basis. We are excited to bring you more coverage from the I/O Fund team geared toward new idea generation only.

I/O Fund Equity Analyst, Jea Yu, contributed to this analysis

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

Dell Q4: Projects $15 billion in AI shipments this year 

Posted on by io-fund

Dell had the unfortunate timing of reporting earnings at the moment tariffs threats increased, with China to see 20% tariffs starting as soon as next week. Dell is exposed to tariffs as a report from 2023 stated China revenue falls between 8% and 14% of sales.

In addition, Dell is the clearest supplier yet to illustrate the impact of the NVL systems being delayed. This quarter, the company stated they project $15B in AI shipments this year, yet in Q4, AI orders were $1.7 billion, down (53%) QoQ and shipments of $2.1 billion, down (28%) QoQ with $4.1 billion in backlog as customers work through technology changes. If Wall Street didn’t get the point from Nvidia’s fairly vague commentary, then Dell made it abundantly clear by offering shipments and backlog, which Supermicro and others do not offer.

Down the income statement, the company reported mixed results. Revenue grew by 7.2% YoY to $23.93 billion, missing estimates by 2.6%. While adjusted EPS grew by 18.1% YoY to $2.68, beating estimates by 6.4%.

Management guided Q1 revenue in the range of $22.5B to $23.5B, representing a YoY growth of 3.4% at the midpoint, missing estimates by 3% and expects adjusted EPS to grow 25% YoY to $1.65.

Management has guided $15 billion in AI shipments for FY2026. The company’s AI server backlog increased from $4.5 billion in Q3 to $9.0 billion in Q4, primarily driven by the recent deals, including xAI.

Where Dell and Super Micro may both be seeing lower growth than expected likely goes back to the delivery of key Nvidia systems, where the larger systems lead to higher revenue (and you’re aware by now these were delayed by “couple months”). It’s also perhaps due to Nvidia’s partnership with Foxconn, who has seen more news lately than peers Dell and Supermicro in terms of shipping Blackwell systems. According to a news report from Economic Daily the GB200 was shipped by Foxconn in small quantities at the end of Dec and is expected to be shipped in large quantities at the end of January. Nvidia is expected to unveil the GB300 AI server product line at the GTC conference in March.

Ultimately, this is one more supplier that is offering a muted outlook. Per Knox’s webinars, IOF has been preparing for a rout of sorts. Our goal has been to add to AI positions during the rout. With that said, we have must-hold levels and will let you know if Dell breaks our must-hold.

Revenue

Q4 revenue grew by 7.2% YoY to $23.93 billion driven by the 22% YoY growth in the ISG segment. However, the revenue fell short of estimates by 2.6%.

  • Management guided Q1 revenue in the range of $22.5B to $23.5B, representing YoY growth of 3.4% at the midpoint, missing estimates by 3%.
  • Analysts expect revenue to grow 1.8% YoY to $25.5 billion in Q2 and 9.3% YoY to $26.6 billion in Q3.
  • FY2025 revenue grew by 8.1% YoY to $95.57 billion. The management guide for the FY2026 is $101B to $105B, representing YoY growth of 7.8% at the midpoint of $103 billion, slightly missing the consensus estimate of $103.62 billion.
  • Analysts expect FY2027 revenue to grow 6.8% YoY to $110.71 billion.

Key Operating Segments

Infrastructure Solutions Group

ISG revenue grew by 22% YoY and flat QoQ to $11.4 billion. Revenue growth decelerated from 34% in Q3.

In Q4, AI orders were $1.7 billion, down (53%) QoQ and shipments of $2.1 billion, down (28%) QoQ while management mentioned that $4.1 billion is in backlog as customers work through technology changes. The company’s AI server backlog increased from $4.5 billion in Q3 to $9.0 billion in Q4, primarily driven by the recent deals, including the $5 billion from XAI. Management has guided $15 billion in AI shipments for FY2026.

The company reported a record ISG operating income of $2.1 billion, up 44%. This was driven primarily by higher revenue. The ISG operating margin was 18.1% compared to 15.3% in the same period last year.

  • Servers and Networking revenue grew by 37% YoY and down (10%) QoQ to $6.6 billion.
  • Storage revenue grew by 5% YoY and 18% QoQ to $4.7 billion.

Regarding Dell’s AI server solutions, the company highlighted the Power Edge servers and the ability to scale up and scale in the opening remarks:

“We added five platforms to our AI-optimized portfolio, including support of Blackwell architectures, the highlight being the PowerEdge XE9712 supporting NVIDIA's NVL72 GB200, which we were the first to ship in the world. We launched the Dell Infrastructure Rack Scalable System, our IR7000 and 5000 in both 21-inch and 19-inch versions, providing up to 96 GPUs in a rack and 786 GPUs in a scalable unit. We have made significant advancements with CDUs, cold plates, manifolds and power distribution with our IR7000 supporting up to 480 kilowatts per rack.”

The company also benefits from the increase needs of storage due to data-hungry AI models: “We made significant advancements to PowerStore with PowerStore Prime, our mid-range storage solution addressing the fastest-growing portion of the market. And we introduced the PowerScale F910 and F710 in our unstructured portfolio that is primed to support unstructured and AI workloads.”

Client Solutions Group

Client Solutions Group revenue grew by 1% YoY and down (2%) QoQ to $11.9 billion. The CFO said in the earnings call, “We saw some promising signs as we went through November and December with pockets of strength in large deals, but overall saw a slowdown in January. As Jeff mentioned, we saw strength in small and medium business, which is historically a leading indicator.”

The CSG operating margin was 5.3% of revenue compared to 6.2% of revenue in the same period last year and was also down 90 basis points sequentially due to a more competitive pricing environment.

  • Commercial revenue was up 5% YoY and down (1%) QoQ to $10 billion.
  • Consumer revenue declined by (12%) YoY and (5%) QoQ to $1.9 billion.

When AI PCs take off, Dell will be a beneficiary with the following stated on the call regarding their participating in both Arm-based and x86-based PCs: “We introduced the most Copilot+ PCs powered by ARM-based Qualcomm Snapdragon processors and also launched the broadest portfolio of Intel Lunar Lake commercial PCs, furthering our number one leadership position in commercial AI PCs worldwide.”

Operating Margins Stronger than Expected

The company’s margins improved due to the operating leverage and higher profits in the ISG segment driven by the Dell IP storage portfolio. The company also discovered previously unrecognized accumulated credits from suppliers and have revised the margins slightly higher than the previous figures. However, the company’s gross margins are feeling pressure due to increasing competition faced in the CSG segment and also the increase in AI revenue.  

  • The gross margin was 23.7% compared to 24.1% in the same period last year. Adjusted gross margin was 24.3% compared to 24.8% in the same period last year. The lower gross margin was due to a higher AI-optimized server revenue mix and lower profits in the CSG segment due to increased competition. The gross margin rate will be lower sequentially given seasonally lower storage mix and a higher AI-optimized server mix.
  • The operating margin was 9% compared to 6.9% in the same period last year. Adjusted operating margin was 11.2% compared to 9.8% in the same period last year. The improvement in the operating margin was due to higher revenue and lower operating expenses.
  • Net margin was 6.4% compared to 5.4% in the same period last year. The adjusted net margin also showed improvement as it reported 8% compared to 7.4% in the same period last year.
  • FY2025 adjusted gross margin was 22.8% compared to 24.5% in FY2024. Management expects adjusted gross margin to be down 100 basis points in FY2026 due to a higher mix of AI-optimized servers and the current competitive environment.
  • FY2025 operating margin improved 40 basis points to 6.5%. Adjusted operating margin remained the same at 8.9%.
  • FY2025 net margin improved 100 basis points to 4.8% and adjusted net margin remained the same at 6.1%.

Adj.EPS grew by 18.1%

  • Q4 adjusted EPS grew by 18.1% YoY to $2.68, beating estimates by 6.4% driven by operating leverage and improvement in the ISG segment operating margin and partly helped by revisions from the previously unrecognized accumulated credits from suppliers.
  • Management expects Q1 GAAP EPS to be $1.29, down (6%) YoY and adjusted EPS to grow 25% YoY to $1.65.
  • Full-year FY26 GAAP EPS expected to be $7.85, up 23% year over year, and adjusted EPS to be $9.30, up 14%.

When the time comes for AI servers to ship in volume, one thing to keep in mind is Dell’s operating leverage compared to other hardware peers – here is how the CFO described it with double digit growth that exceeds revenue growth: “Now let's look at operating income. We delivered a 22% increase to $2.7 billion or 11.2% of revenue. This was driven by higher revenue and lower operating expenses, partially offset by a decline in our gross margin rate. Q4 net income was up 15% to $1.9 billion primarily driven by stronger operating income. And our diluted EPS was up 18% to $2.68.”

Cash Flow and Balance Sheet

The company’s cash flows were lower due to lower CSG revenue and also higher inventory to support the AI demand. This was clarified by the management in the earnings call Q&A as they expect cash flows to improve this year with higher CSG revenue.

Q: Amit Daryanani (Analyst)

Thanks a lot. I guess I have a question on free cash flow. In fiscal '25, it looks like your free cash flow is down a couple of billion dollars versus '24. Can you just talk about what's driving this contraction in free cash flow? And maybe, Yvonne, you can help us kind of understand how do we think about free cash flow expectations as we head into fiscal '26? What are sort of puts and takes around it? It would be really helpful to kind of get the context, at least for fiscal '26, what's going on?

A: Tyler Johnson (Vice President and Treasurer)

“Amit, look, I think — as I was sitting here last year, I definitely thought cash flow was going to be a little bit stronger. If you look at what played out, one, we didn't see the growth in CSG that we were expecting. And as you know that throw off really good cash. And then two, we invested a lot in our AI business through inventory. And so you can see that our inventory has gone up, and that had a big impact to CCC. Now if I look where I am today and I think about FY '26, I would say I've got a few things working in my favor. So one, we're at a CCC level where historically, we've always shown improvement from here. And that will throw off good cash. We expect good CSG this year, and that will throw off good cash. And if I think about the growth in the P&L, that will throw off good cash. So look, I think we feel pretty good about cash. I do expect it to be greater than 1x and so yes.”

— End of Quote

  • Q4 operating cash flow margin was 2.4% compared to 6.9% in the same period last year.
  • Q4 free cash flow margin was (-0.5%) compared to 3.6% in the same period last year and adjusted free cash flow margin was 2.0% compared to 4.5% in the same period last year.
  • Cash and investments were $5.13 billion and debt of $24.57 billion compared to $6.5 billion and $25.02 billion in Q3.
  • Inventories were $6.7 billion in Q4 compared to $6.65 billion in Q3. However, it was significantly higher than the $3.62 billion in the same period last year to support the AI demand.  
  • The company repurchased shares worth $734 million and paid dividends of $311 million.
  • It also announced an 18% increase in the annual dividend to $2.10 per share. Additionally, the Board of Directors has approved a $10 billion increase in the share repurchase authorization.

Earnings Call:

Blackwell Margins to be Lower than Hopper Margins; Yet Dell Likely Will Win the Margin Battle

Margins are a key issue with hardware suppliers, and Dell stated Blackwell margins will be lower than Hopper margins, which is something to watch as the next architecture ramps: “I mentioned in the last call that the Blackwell margins were lower than the Hopper margins and remains so today. We're still early. The deals are very large upfront. There's more competitors, so it's a more competitive landscape.”

An analyst pointed out that despite growing their mix of AI server revenue, Dell guided for ISG margins to be flat – which is a win since AI servers are supposed to weigh on margins. Dell’s response is they will leverage storage to offset any margin weakness from AI servers.

In terms of Dell’s thesis, I would place margins equal to its AI server backlog, which is why I’m starting with this piece rather that AI in the call summary.

“Michael Ng

Hi good afternoon. Thank you for the question. I just have one on the ISG margin outlook of flat year-over-year for the upcoming year. It's a great outlook, particularly considering AI server revenues growing 50%. So can you talk a little bit about the expectations for margins for some of the components, traditional servers, storage AI servers? I'm just trying to understand the ability to keep ISG margins flat despite presumably the dilutive effect from the AI server margins. Thank you.

Jeff Clarke

I think, Mike, maybe the way to look at this is the, first and foremost, as we think about holding ISG margins flat, I love the way that you asked the question, we're going to do that by growing at least $15 billion in AI servers. I know your question is how we're going to do that. But for us, that's a very important mark that we're going to be able to meet that operating range that we've committed in our long-term framework and we're going to grow at a minimum of $15 billion in AI servers. And we're going to do that by what we've done in traditional servers and what we've done in storage. The storage leverage that Yvonne talked about earlier is front and center. When we grow the storage business and we control our expenses, scale matters, the operating margins improve. When we pivot to Dell IP storage, which we have done, our margins improve. The margins of our own IP are vastly superior than third-party IP. We've been doing that for some time.”

— End Quote

Dell Calms Supply Fears About Blackwell

Dell is the perfect company to corner on the GB200s shipping and if there is a “supply” issue. This is also the perfect question to ask as DeepSeek has greatly obfuscated the issue to where the media is focused on demand (which is def not the problem for 2025!!). Supply is the issue, to where rather than there being supply constraints (like during Covid), there are supply hiccups.

I find there are two kinds of management teams – those that are transparent and take it on the chin for being so, and those who are more vague and hope the market skates over their commentary. Dell is the former.

Samik Chatterjee

Hi, hopefully you can hear me now. Jeff, I just wanted to go back to some of your prepared remarks and — about the $15 billion of AI server revenue that you were highlighting that you at least expect to grow to that level. Just wondering, how much of that is gated by supply, particularly versus the visibility into supply that you're getting? And how much of that commentary around sort of at least growing there is a supply dynamic versus a demand dynamic? And should we be expecting more sort of linear growth for the quarter as we think about the — with visibility on supply? Thank you.

Jeff Clarke

Well, I think clearly, Hopper supply is available today. I believe there is references yesterday that Blackwell is in production and ramping. We're open for business and taking orders. The message that I really wanted to drive in our remarks is on day 27 of the fiscal year, we're trying to communicate that we are at least $15 billion in AI shipments. Our 5-quarter pipeline continues to grow. It's several multiples of our backlog. We are going to pursue every opportunity with the CSPs and in enterprise. These large-scale systems are accelerating and getting bigger. Models are quickly moving to reasoning models, which consume and require more computational capability, i.e., more computers. And the use cases continue to get clearer for enterprise to drive the return on investments they want to see to actually use AI more broadly. Algorithm innovation continues to accelerate. Again, these reasoning models are — will consume more computational capability. They're moving to be multimodal, which even consumes more kind of like where this is going. We're optimistic. I don't see supply as an issue. Clearly, these are about building the right architecture. There's a customer preparation or customer readiness component of this, new data centers getting powered, getting water, getting cooling. There's other materials beyond the GPU, getting the rack, getting the cold plates, getting the CDUs, the PDUs, all of that is what we orchestrate. We have line of sight that is at least $15 billion. We'll continue to update as that might change. And we're all in. I don't know what else I can tell you. I hope that helped.

Conclusion:

Dell’s server shipments QoQ makes it abundantly clear they were impacted by the delay in Blackwell NVL server shipments. How any issues are sorted will be fluid, as any supplier that can alleviate the issues around standing up the systems will be exponentially rewarded while those who falter will be eliminated. Dell’s $15B is lower compared to Supermicro with annual FY estimates of $24B. However, Dell is quickly catching up. What separates the two is the operating leverage, the more trustworthy management team (given SMCI has been in the crosshairs of the SEC lately), and the AI PC story that will materialize eventually.

Given the weak price action AH, we have a must hold level for Dell as the market is panicked right now over tariffs and weak AI supplier commentary (officially across the board now that we have Dell’s more transparent report on server shipments). Our goal is to not get shook out the position, while adhering to risk management. We will keep you in the loop if the must hold breaks.

I/O Fund Equity Analyst Royston Roche contributed to this article.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Readings:

Nvidia Suppliers Send Mixed Signals for Delays on GB200 Systems – What It Means for NVDA Stock

Posted on by io-fund

Nvidia has led the Mag 7 for two years with a return of 853% since Jan of 2023. The second closest Mag 7 return in this time frame is Meta at 495% with all other Mag 7 stocks returning less than 200%, and popular tech ETF QQQ returning 105% in that time frame. All the above is outstanding performance, yet one stock has clearly set itself apart.

Given the spectacular outperformance of more than 7X the QQQs, it goes without saying there is outsized pressure on Nvidia’s stock, whether it’s from short sellers who are still in disbelief, or from investors who are simply wanting to book gains.

Recently, a low-cost large language model (LLM) from China sent the stock down 25% in two days. In the heart of the aftermath, I spoke with Charles Payne on Fox Business News, discussing key reasons investors should stay the course, given cheaper AI development will drive more demand (not less). The stock has nearly recouped its losses, trading 6% below its high before the Deep Seek news broke. Ultimately, Deep Seek is immaterial to AI demand, or perhaps even a net add given AI is too expensive to develop for most developers, resulting in fewer consumer and enterprise applications.

When it comes to Nvidia, there’s no denying the Street gets plenty wrong. Perhaps most egregious was when the Street dropped the stock 60% based on a gaming miss, as Ethereum’s merge to proof of stake in September of 2022 was expected to be the death knell for the AI juggernaut. A month later, the AI GPUs that drove the 7X outperformance shipped. The disconnect was staggering.

The I/O Fund has a strong track record on this stock, discussing every twist and turn publicly for our free stock newsletter readers with documented gains of up to 4,100% as far back as 2018 based on a very-early AI thesis. That gaming miss that resulted in a drop of 60%? We bought that too.

However, it’s important that I pause the exuberance and discuss something stirring beneath the surface. Recently, in this quarterly only, key suppliers are providing mixed guidance on the timing of Nvidia’s Blackwell GB200 systems. The commentary is subtle, and it would require knowing this stock thoroughly to identify the change in tone.

To put it simply, if Blackwell GB200s were ramping, we would see strong sequential growth for Q1. At the very least, there would be some indication Q2 is setting up for strong growth, and yet the commentary is shifting toward a second half discussion. When you add that suppliers do not want to get into the crosshairs with Nvidia, yet are under SEC regulations on how they offer guidance, the language used is incredibly easy to miss.

Ultimately, my spidey senses are up as commentary for Q1 and Q2 should be stronger on the products and solutions that supply Nvidia’s GB200 NVL systems. Meanwhile, Nvidia’s management team is under immense pressure as the company has beat revenue estimates by $1 billion or more for six quarters. This means even a minor delay or minor adjustment in expectations could come under close scrutiny.

As a reminder, we don’t make earnings calls, as many factors can affect stock price. For example, even though I’m getting mixed signals on the timing for GB200 NVL systems, Nvidia’s B200s are ramping, and theoretically can absorb some of this demand.

Instead of making an earnings call, we present quality research so that investors are fully informed to make their own decisions. From there, we take this a step further and publish every single trade we make on our research site. In finance, full transparency is rare, yet through never-ending tenacity, my firm has offered up to 4100% gains on Nvidia alone. We continue this long-standing dedication to our readers in the analysis below.

Nvidia’s Future Hinges on Blackwell – GB200s are the Standout SKUs

Due to the cyclical nature of GPU shipments, the differences in each generation are critical for investors to track. The Hopper generation has driven immense revenue growth over the past two years, while the Blackwell generation is expected to drive revenue that exceeds 2023 and 2024 combined. Hopper brought Nvidia to a $100 billion data center segment – at $26.3 billion in fiscal Q2 2024 – yet I pointed out how Blackwell could drive the data center segment to $200 billion-plus ten months ago.

You can read more about the nuances of each generation of Nvidia’s AI accelerators here, and why Nvidia’s future generations of GPUs can help Nvidia Stock reach $10 Trillion in Market Cap.Nvidia Stock reach $10 Trillion in Market Cap.

The rack-scale GB200 NVL72 features 36 GB200s, or 72 B200 GPUs and 36 Grace CPUs, offering up to 30X inference improvement and up to 5X training improvement versus the HGX H100, with significant improvements in energy efficiency and data processing speeds. The GB200 NVL72 was designed to address and unlock real-time inference for trillion-plus parameter models.

Despite being on an accelerated timeline, Blackwell will deliver the largest leap generationally to date for Nvidia’s AI GPUs.

  • The B200 GPU will deliver a 2.5X training improvement and 5X inference improvement over the H100.
  • Blackwell will see a massive upgrade in chip size, at 208 billion transistors compared to the H100’s 80 billion transistors.
  • The B200 will also have 20 petaflops of FP4 compared to the H100’s 4 petaflops of FP8. I’ve covered the importance of this awhile back, and more recently following DeepSeek.

Blackwell’s pricing power is one of the key factors behind its growth potential, with the GB200 expected to be priced between $60,000 to $70,000, around double the H200’s estimated $32,000 price tag. For rack-scale solutions, the price tags are much heftier – the GB200 NVL36 (featuring 18 GB200s and 18 Grace CPUs) is estimated to carry a $1.8 million price tag, while the GB200 NVL72 is estimated to command an eye-watering $3 million price tag.

Roughly 35,000 NVL72 racks are estimated to be shipped this year, that’s already up to $105 billion in revenue for Nvidia. This volume correlates to shipments of ~2.52 million GB200s, versus Hopper shipments of 2 million-plus in 2024. To note, the revenue and volume do not include the NVL36 racks, B200s, B200As, B300s, and Hopper GPUs still on deck this year.

Nvidia B200s Will Be in High Demand and Likely Ship in Volume in H1

There is certainly a scenario where Nvidia’s GPUs are in such high demand that other SKUs can help make up for a delay on the much-larger GB200 NVL systems. The B200s are expected to be in high demand and companies such as Super Micro are shipping B200 HGX systems in volume. This is one reason Super Micro has seen favorable price action despite lowering their fiscal year guidance; it’s assumed the company can make up for any delays on the GB200s with B200 and B200 HGX systems.

However, the market does not like surprises. As of now, the market is expecting the GB200s to ramp in Q1 and further ramp in volume in Q2. It will be a tall order to meet these expectations (on the dot) with lower priced GPUs and HGX systems.

SECTION TAKEAWAY: Blackwell is a significant undertaking to have 36 GPUs and up to 72 GPUs communicate as one GPU (number depends on the SKU, with complexity increasing if it’s a single rack versus multi-rack configuration). Previously, Nvidia’s AI systems combined 8 GPUs. This significant leap not only increases pricing from roughly $32,000 per GPU to up to $3 million per system, but it also greatly increases the need for new networking architecture and comes with increased power requirements.

Nvidia GB200 Delay Rumors Grow – The I/O Fund’s Take for Q1 & Q2

Dating back six months, there have been rumors that the GB200s are delayed. The first rumor came from The Information, stating machines were sitting idle at Taiwan Semiconductor, where Nvidia’s chips are made. I informed my readers at the time this was unlikely to be the case as TSMC was reporting record high-performance computing revenue. In fact, TSMC continues to report strong HPC revenue.

NVDA stock outlook chart: TSMC reports strong QoQ HPC growth (Q2-Q4), indicating active production.

NVDA stock outlook: TSMC shows no signs of idle machines with strong QoQ growth for the HPC segment during Q2 to Q4.

In January, analyst Ming-Chi Kuo reported there were “Short-term Potential Risks in Nvidia and Related Supply Chain,” stating that GB200 NVL72 shipments in 2025 could reach roughly 25,000 to 35,000 racks compared to 50,000 to 80,000 racks from last year. The same analyst reported earlier in the year that “The biggest challenges in NVL72 development mainly stem from the 132kW thermal design point (TDP) requirement, which makes it the highest-power-consuming server in history. Nvidia and its supply chain need more time to solve unprecedented technology issues.”

This post piqued my interest as the emphasis would be outside the foundry. Instead, this would suggest an entirely new delay with thermal management issues (direct liquid cooling), or perhaps with power management solutions, or even networking related — rather than a material delay related to TSMC and yield issues.

To cross-examine this possibility, below is a brief summary from some of the more well-known direct-liquid cooling suppliers, PMIC suppliers (power management integrated circuits), and a few networking companies outlining commentary that leads to a higher probability there is a delay for the GB200s. If my read-through from these management teams is correct, the delay falls somewhere on the thermal/power solutions or networking components part of the supply chain rather than on TSMC.

Explaining the Lower Rack Shipments of NVL72 Systems

Regarding the statement that the GB200 NVL72 shipments in 2025 could reach roughly 25,000 to 35,000 racks compared to 50,000 to 80,000 racks from last year, I want to take a step back and clarify here that while that would represent a substantial drop, it’s more representative of a significant shift in NVL72 share from last year, and more reflective of what key partner TSMC can handle in terms of CoWoS capacity.

In July of 2024, estimates for GB200 racks were revised 50% higher from 40,000 to 60,000 racks, with the NVL36 at more than 80% share at 50,000 racks with the remaining 10,000 racks to be NVL72.

This would correlate to 2.52 million GPUs shipped – coincidentally, the same amount of GPUs expected to be shipped with 35,000 NVL72 racks, with more on deck from NVL36 configurations.

So, while it may look to be a large decline, GPU shipments at 35,000 NVL72 racks plus additional NVL36 racks would actually be higher than last year’s 60,000 NVL72 and NVL36 estimate due to a large shift towards the NVL72 configuration.

screen shot of Beth Kindig’s tweet

Additionally, the industry’s more recent, most optimistic forecasts for 50,000 to 80,000 NVL72 racks would be unrealistic and practically impossible for TSMC to meet – at the midpoint, this would require ~4.68 million B200 GPUs, or more than 292,000 in CoWoS wafer allocation simply for the NVL72 (not including NVL36).

Current estimates from Morgan Stanley place the B200’s CoWoS allocation at ~220,000 wafers for 2025, or more than 30% of TSMC’s total capacity for the year. This coincides with estimated B200 GPU shipments of 3.52 million, more than enough to handle 35,000 NVL72 racks and similar volumes of NVL36 racks.

SECTION TAKEAWAY:

With what the I/O Fund knows today, the product mix of NVL72 being higher, thus resulting in a lower number of racks being delivered, is not the issue. Rather, it’s the timing that could pose an issue due to Wall Street potentially having to face uncertainty on what quarter NVL systems would ship in volume.

Blackwell Necessitates Direct Liquid Cooling – Super Micro and Vertiv’s Comments

In June, we wrote an analysis on AI Power Consumption: Rapidly Becoming Mission Critical which stated that as the industry progresses towards a million-GPU scale, more emphasis is placed on future AI GPU generations to focus on power efficiency while delivering increasing levels of compute. Data centers are expected to adopt liquid cooling technologies to meet the cooling requirements to house these increasingly large GPU clusters, and the percentage of air-cooled systems shipped versus liquid cooled systems will change (dramatically) with Blackwell’s NVL systems.

Therefore, any changes to Blackwell’s NVL systems timing would appear in commentary from companies that provide direct-liquid cooling solutions and servers, as the previous generations are air cooled.

Super Micro Lowers FY25 Revenue Guide, Pushes Back 30% Liquid Cooling Target

Super Micro recently reported preliminary fiscal Q2 results, in which it cut its fiscal 2025 revenue guide while hinting that Blackwell GPUs could take longer to ramp in the first part of the calendar year.

At Computex in June 2024, CEO Charles Liang stated that he “expects 15 percent of racks [SMCI] sells this year to use DLC, and 30 percent to employ it [in 2025].” In August 2024,  Liang slightly changed his tune, noting that they now expect “25% to 30% of the new global datacenter deployments to use DLC solutions in the next 12 months.”

I remember this comment well, as it was part of the evidence used to debunk the August rumors considering SMCI had moved up their timeline for DLC delivery (from end of year 2025 to mid-year 2025).

Two weeks ago in the February 2025 report, this timeline was shifted back with Liang stating that Super Micro now believes that “DLC or overall liquid cooling market share will grow all the way to 30% or even more in the next 12 months” — it’s not a dramatic shift in tone but rather pushes the 30% target back from mid-year 2025 noted in August, to beyond the original year-end timeline — to now technically being early 2026.

The August earnings call also housed one critical piece of information regarding Blackwell and Super Micro’s 2025 revenue. Wells Fargo’s Aaron Rakers questioned management about Blackwell and guidance for December 2024 and onwards. Liang explained that for the December quarter, Blackwell “will be very small. Engineering sample small volume. So the real volume, I believe, had to be March quarter next year. And that's why we foresee only $26 billion to $30 billion.”

But now, SMCI implied they are not able to ship Blackwell in the March quarter:

“Do you guys have a forecast from NVIDIA? You know when you think you're going to start to see you know supply the GPU, so that you can ship the NVL72 where visibility is still pretty low on availability of the GPUs?

Sign up for I/O Fund's free newsletter with gains of up to 2600% because of Nvidia's epic run – Click hereSign up for I/O Fund's free newsletter with gains of up to 2600% because of Nvidia's epic run – Click hereClick here

Charles Liang
We already proved pretty much everything. And now, just waiting for – and we are in some allocation, some volume, but the volume demand is way much bigger. So we are waiting for more allocation. So hopefully very soon we can ship in much higher volume."

Super Micro revised its FY25 guidance lower at the beginning of last week, now seeing revenue of $23.5 billion to $25.0 billion, or more than 13% lower than the initial guide at midpoint. Q3 revenue was guided between $5 billion and $6 billion, below the $5.95 billion estimate. A weaker-than-expected March and FY25 guide plays into management’s comments of a lack of Blackwell NVL72 availability, with Super Micro only just beginning volume shipments of Blackwell-based racks:

Q: Jon Tanwanteng, CJS Securities: “I was wondering if you could break down the factors …driving the reduction in the 2025 revenue guidance. How much is maybe pricing related? How much do you think is related to delays or availability of Blackwell?”

A: David Weigand, Super Micro CFO: “Yeah, I would say, Jon, that probably the biggest factor was just the delay in new technology because we were, when you think about it, we were all set to go. We were all set to ship with liquid cooling. We were ready. But the problem was that not everything else was.”

This was corroborated by Liang:

“Once Blackwell [is] in volume production, I believe we will have strong growth. And now we are just preparing, diligently preparing all the logistics, including the system enclosure, thermal solution, for sure GPU supply from our vendor Nvidia. So we are well prepared and once logistics ready, we are ready to ramp up our growth.”

Super Micro has walked back the revenue guide that hinged on Blackwell’s NVL systems shipping in volume in the March quarter. Management was also quite clear that the new product was facing a delay as they await more GPU supply.

Vertiv Signals Softer Q1

Vertiv is a provider of thermal and power management solutions and is a leader in direct liquid cooling. The company stated they benefited from a “particularly strong” Q4, with revenue rising nearly 26% YoY as they overdelivered by nearly $200 million as customers wanted products as soon as possible.

Vertiv’s strength in Q4 may be tied to key partner Dell, as its AI server backlog in fiscal Q3 (October quarter) rose more than 18% QoQ to $4.5 billion after being flat QoQ in Q2, and its AI pipeline rising more than 50% QoQ. Analysts placed Dell’s AI server pipeline at $16 to $17 billion, up from $11 to $13 billion previously, a rather large jump for one quarter. However, Dell noted that one factor for its softer Q4 and fiscal year guide was the “unpredictability of the AI shipments” as there are “some more timing differences than what we were anticipating when we gave the guide the last quarter.”

For Q1, Vertiv guided a deceleration to 19% YoY growth, with FY25 revenue growth guided at 16%. Management defended this deceleration into Q1 by saying: “Now of course, Q4 was particularly strong. So we should not look at Q1 as a quarter-to-quarter, really look at the first quarter sales as the acceleration that has taken place. With a 19% organic growth in the first quarter, I feel very, very good about what that tells us about our overall trajectory” and also that it is “higher than what we actually saw in 2024.”

Liquid cooling capacity is not a constraint for Vertiv, as they had said last summer that they were “on track to finish in 2024 with a 45x capacity increase compared to baseline at the end of 2023.”

I agree with the analyst sentiment (and weak price action) following the report that the commentary doesn’t check out exactly. Per I/O Fund numbers, management is contradicting itself by guiding for organic growth of 19% at the midpoint for Q1 but 16% growth for the fiscal year (i.e., slower growth later in the year) while stating shipments will increase QoQ.

Additionally, the QoQ/YoY has to be looked at which is lower than what typical seasonality would account for, as our numbers indicate Vertiv was down (12.1%) due to seasonal sequential growth from Q4-Q1. This year, Vertiv is down (16.9%) from Q4-Q1.

My conclusion (still to be confirmed) is that Vertiv may see a weak Q2 before there is an acceleration into the back half.

One Q&A exchange in Q4’s call voiced these concerns:

Q: Steve Tusa, JP Morgan: “Obviously, there's a lot of focus on orders, I think, for good reason. Everybody's trying to discern the trend relative to these CapEx numbers, the pipelines that are obviously pretty eye-popping. You're now two quarters step down relative to what we see at your customers and the way they're spending in these pipelines. What is that disconnect? Is there some sort of disconnect between you guys and everybody else talking about doubling their data center businesses?”

A: Giordano Albertazzi, Vertiv CEO: “I don't think there is a disconnect, quite honestly. If you look at our orders trajectory last year, if you think about a 60% year-on-year growth in the first half of the last year, that's a lot of growth. When our customers talk about their CapEx, of course, they also talk about a lot of the silicon part of their CapEx, not all the data centers. So, I feel pretty good about our visibility of the market and what we win in the market.”

Networking Companies Shift Tone on Timing for GB200s

Please note: the full research including company names, management statements and stock tickers are provided to our research members We offer a more generalized discussion below.to our research members We offer a more generalized discussion below.

PCIe 6.0 supplier:

PCIe 6.0 is a networking standard expected to initially ship with Nvidia’s Blackwell. A key supplier stated to expect this in the second half of the year, as these products “are driving higher dollar content opportunities […] on a full rack and full accelerator basis and we expect volume deployments to begin in the second half of this year.”

Revealed at GTC in March of 2024, PCIe 6.0 was initially expected to launch with the GB200s. The following quote from this supplier also hints that merchant GPUs (Nvidia) will not be driving H1 – which would be odd if Blackwell was shipping. “Now, if you look into 2025, we see both contributing growth. The first half of the year will be more predominantly the internal AI accelerator programs. But if you get into the back half of the year, the transition on the merchant GPUs will also be very strong for us. This is where you'll see the custom rack configurations start to deploy and that’s where we see a big dollar increase in our contemporary GPU with [our product] starting to ramp.”

PMIC Suppliers:

PMICs (power management integrated circuits) are a critical part of the picture for Blackwell, given that these components were linked to Blackwell’s rumored power management issues. Major PMIC suppliers were unable to confirm volume shipments in the first quarter in recent earnings calls.

One PMIC supplier that we covered on our Advanced site on Feb 5th, is stating that “We know that we're not the only one players going after this, so the share is still to be determined. The final design, everything is still to be finalized as well, too. So we're working closely with the customer. Again, at this point, we can say that we're targeting for a mid-year launch.”

This company is either discussing Blackwell Ultra with the B300s, or they are implying GB200s are delayed.

The competing PMIC supplier stated: “Just to add a little bit of color to how we see the year rolling out, we believe that we will be off to a slower start in the first half of the year. But as the year develops, the customer base is expected to broaden as hyperscalers launch their new products. We have multiple product ramps with both existing customers as well as with these new hyperscalers.”

The takeaway is that neither PMIC supplier can confirm they are shipping in volume in Q1 or Q2, yet meanwhile, both are saying they are still part of the supply chain.

Semtech Suddenly Pulls Guidance on Active Copper Cabling (ACCs)

Perhaps the most drastic commentary to come out of the last few weeks was when a key supplier pulled its Q1 guidance intra-quarter.

Semtech filed an 8-K stating that “net sales from its CopperEdge products used in active copper cables are expected to be lower than the Company’s previously disclosed floor case estimate of $50 million due to rack architecture changes, with no expected ramp-up over the course of fiscal year 2026.”

ACC content had been estimated to be substantially higher with 36×2 configurations – it had been rumored back in October 2024 that Nvidia was halting development of one of the NVL72 configurations, the NVL36x2, which linked two 36 GPU racks together in a side-by-side system. Semtech pulling guidance strengthens this view.

The shift in architecture to discontinue the 36×2 configuration was said to possibly “disrupt the supply chain for assembly and cooling solutions,” by removing dual-rack configurations and focusing solely on single-rack NVL72 and NVL36 configurations. DLC suppliers have been pushing back Blackwell’s ramp later in the year, suggesting that the market may have faced some impacts from this rumored design change late last year.

While shifting architectures is not a big deal in the medium-term (as stated above, NVL72 single rack configurations are expected to see a higher mix); it’s the suddenness that Semtech pulled it’s guidance that is cause for concern as the company had reaffirmed its optimistic revenue floor guidance  based on two factors – the timing of Blackwell’s launch and changes in rack design. This suggests both factors may be in play, as a late-stage design change was expected to have downstream impacts on timing for DLC ramping.

Potential Flat Q1 or Flat Q2

A week after we began covering the topic for our research members, Mizuho’s Vijay Rakesh stated he is expecting a “more flattish” Q1 with data center revenue of $36.7 billion versus $37.4 billion consensus.

To reiterate, if there is an issue, it will appear in Q1 or Q2 according to management commentary provided above. For Q4, the QoQ growth in the data center is expected to be $2.6 billion, compared to $4.5 billion QoQ from Q3-Q4.

What to Look for in Nvidia’s Q4 Report

We have yet to see a DLC supplier come in strong, so we believe the risk is elevated that Nvidia reports a weaker-than-expected Q1 or Q2 as the B200s absorbing a timing delay on the GB200s is a tall order.

In terms of the GB200s shipping on time, I’m open to this – and highly prefer it, given I continue to hold a large Nvidia position. However, I can't find clear evidence in the supply chain that these larger systems are ramping for a strong Q1 performance.

HP Enterprise stated that they’ve only now shipped their first Blackwell based system, while Super Micro walked back its FY25 guidance in part due to a soft fiscal Q3, which they had previously said would be when Blackwell increases in volume.

Combined with comments from Vertiv and Dell, as well as other suppliers noted in this analysis, there appears to be an air pocket, of sorts. A quarter that doesn't blow it out of the water would be unusual for Nvidia as it has consistently smashed expectations since Hopper’s breakout quarter in 2023.

On that note, Nvidia’s data center revenue has consistently exceeded expectations by at least $1 billion or more for six consecutive quarters, with fiscal Q2 ‘24 being the largest at a nearly $2.5 billion beat. Last quarter, fiscal Q3 ‘25, was the second largest at nearly $2 billion.

Nvidia data center revenue: Beat $2.5B (Q2'24), $2B (Q3'25). Q2'25 QoQ growth slowdown to 15.3% led to 6% stock decline.

Nvidia’s Q2 ‘24 saw the largest beat at $2.5 billion, while Q3 ‘25 followed with a nearly $2 billion surprise, driven by strong demand for Hopper

This has driven a string of impressive top-line beats each quarter, with the smallest revenue beat being $1.29 billion in fiscal Q2 ‘25. The shrinking size of revenue beats in that quarter correlated to a deceleration in QoQ revenue growth to 15.3%, one factor behind the more than 6% decline Nvidia felt the day following that report.

Nvidia stock: Q4 revenue forecast $38.1B (+8.6% QoQ), Q1 ‘26 projected $41.98B (+10.2%), driven by AI growth.

For the upcoming fiscal Q4 report, revenue is expected to be $38.1 billion, for QoQ growth of 8.6% while Q1 ‘26 is expected to see revenue of $41.98 billion, for QoQ growth of 10.2%.

Conclusion:

Given market jitters around DeepSeek, which turned out to be a non-issue, something more material related to the GB200s, such as growth slowing below expectations at the start of the new fiscal year, could send the stock below $100 — which we would see as a buying opportunity. We provided a more detailed buy plan before the earnings season began in the article “Where I Plan To Buy Nvidia Stock Next”

There is a scenario where Nvidia’s stock marches higher – perhaps based on the remaining strength from Hopper and the B200s, B200 HGX systems and B200As packaged with CoWoS-S and having a lower thermal design power (TDP). Or perhaps the string of suppliers discussed here are simply stating the GB200s are not ramping in volume in Q1 but will in Q2. Even still, I prefer to wait to see how this resolves as my firm has been tracking key suppliers and AI proxies for six years to confidently build our positions.  

Ultimately, my firm trimmed our Nvidia position (to a 10% allocation) and will happily buy lower should the assumptions in this analysis materialize. Nvidia remains the stock of the decade; however, stock returns – and product launches — are not perfectly linear.

The I/O Fund has recently added five new small and mid-cap positions poised to benefit from the ongoing AI spending war. Join us every Thursday at 4:30 p.m. for our exclusive 1-hour webinar, where we cover market entries, exits, and key insights on the broader market. Take advantage of $50 off our Advanced monthly service, now priced at $99/month. Use Code SAVE50ADV through Feb 28th MidnightJoin us every Thursday at 4:30 p.m. for our exclusive 1-hour webinar, where we cover market entries, exits, and key insights on the broader market. Take advantage of $50 off our Advanced monthly service, now priced at $99/month. Use Code SAVE50ADV through Feb 28th Midnight

Disclosure: The I/O Fund owns Nvidia and a handful of Nvidia suppliers including some of the suppliers listed in this analysis. To view the full portfolio, subscribe here.

Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund own shares in NVDA at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

AOSL Q2: Potential B300 Supplier; Margins Disappoint

Posted on by io-fund

Alpha and Omega is openly being considered as a supplier for Nvidia’s B300 systems, expected to ship mid-year. Given its tiny revenue of $173.2 million, and our ability to combine technicals with fundamentals, we feel it’s a shot worth taking. Should AOSL be confirmed as the supplier over Monolithic Power Systems, then the company would theoretically surge. Should they not be confirmed, the stock would theoretically plummet. It’s not for the faint of heart, yet we continue to like the risk/reward the stock offers.

AOSL supplies the PMICs and MOSFETs powering each GPU, and for the upcoming platform of this customer (widely known to be Nvidia), AOSL will additionally sell the total solution controller and the multiphase controller. The number of PMICs/MOSFETs is expected to triple from 9 to 16 up to 50 MOSFETs per GPU.

With that introduction, the earnings report this quarter matters very little — what matters is if AOSL becomes a confirmed supplier. That announcement, or any supply chain whispers, will likely happen outside of an earnings report as we approach CY Q1-Q2.

For example, this week, a new write-up caused AOSL’s stock to rise 11%, stating AOS’s new power-management IC (PMIC) controller and DrMOS chips helps to “significantly cut transient power demands by several hundred watts during the brief periods when the SoC draws peak power” and was available in production quantities with a lead time of 12 to 16 weeks.

However, last month, there was a third-party analyst that stated AOS was seeing thermal management issues, causing over-heating. The same analyst also indicated AOSL would be a GB200 supplier, yet management is pointing toward the B300s (and associated systems) as the main design they will participate in.

Regardless of the factual accuracy of any single analyst, AOS is certainly in the middle of the action when it comes to qualification testing for the outsized power requirements of Nvidia’s road map. With this small of market cap and revenue base, the news and rumors can quickly change the stock price in either direction. 

Below, we do our due diligence on the earnings report, including what management is saying about their AI systems partner, the weak margins that drove some of the weak price action this quarter, and an idea on timing for when the stock will (or will not) show a sizable impact from being the chosen PMIC and controller supplier for the new server design.

Revenue

Alpha and Omega reported a slight revenue beat in Q2, but that was unfortunately overshadowed by lingering margin weakness.

Looking ahead, management hinted that Q3 would be the bottom for the year, signaling that revenue and margins are expected to recover “beyond the March quarter with incremental growth likely from smartphones, graphic cards and AI.” Management also hinted at increasing contributions from next-gen GPU platforms beginning in the middle of the calendar year.

Q2 revenue increased 4.8% YoY to $173.2 million, accelerating from 0.7% growth last quarter. Management said the results came in ahead of expectations and confirmed that inventory corrections are complete.

For Q3, revenue was guided at $158 million, +/- $10 million, for YoY growth of 5.3%. This was below expectations for $161 million, with management seeing some pricing pressures from a more “subdued” market environment. However, the low guide was also partially due to the wind down of licensing and engineering revenue with the 24-month contract expiring this month – for reference, licensing revenue was $5.4 million in Q4.

Management reiterated on a lack of visibility into 2025, though they did explain that they “expect both revenue and margin to recover beyond the March quarter.”

As stated in the introduction, this all comes down to whether AOS is chosen as the supplier over Monolithic Power, and if they are chosen, to what extent

Margins

This was the weakest part of the report, with margins coming in at the low end of the guided ranges and forecast to contract again across the board. Gross margin is now approaching the low-20% range, while operating margin was guided to drop to the lowest level in two years.

Given Q2’s margins came in low across the board, management is forecasting further contraction in a seasonally slower Q3, partially impacted by the wind down of licensing revenue as the 24-month contract expires this month

  • Gross margin contracted to 23.1% in Q2, coming in at the low end of the 24%, +/- 1% guide. The QoQ decline was once again attributed to “ASP erosion and mix changes.” On a YoY basis, gross margin is down 3.5 points.
  • For Q3, management guided for gross margin to contract further to 21.5%, +/- 1%, with some impacts from the phase-out of licensing and engineering revenue in the quarter.
  • Operating was contracted to (3.4%) in Q2, down from (0.1%) last quarter. The contraction stemmed from R&D expenses coming in slightly above expectations combined with the weaker gross margin.
  • For Q3, operating margin is expected to contract further to (7.9%), with operating loss more than doubling sequentially to ($12.5 million).
  • Net margin declined to ($3.8 million) in Q3.

EPS

Adjusted EPS came in just ahead of estimates, though GAAP EPS missed the mark.

  • Adjusted EPS was $0.09 in Q2, beating estimates for $0.08. GAAP EPS was  ($0.29), missing estimates for ($0.23).

Adjusted EPS estimates have come down significantly due to the recent margin weakness, and likely will come down further given Q3’s weak margin guide. Last quarter, AOSL was expected to report $0.35 in adjusted EPS in the second half of the fiscal year — $0.14 in Q3 and $0.21 in Q4 – though that was seen at just $0.08 combined heading into Q2’s report.

Cash and Balance Sheet

Despite the margin shortfall, operating and free cash flow both improved sequentially.

  • Operating cash flow was $14.1 million in Q2, up from $11.0 million in Q1. OCF margin improved from 6.1% to 8.1% in Q2.
  • Free cash flow was $6.7 million in Q2, up from $4.3 million in Q1. FCF margin improved from 2.4% to 3.9% in Q2.
  • Capex was $7.4 million in Q2, and guided to be $7 million to $9 million for Q3, a slight sequential increase at midpoint.
  • Inventory decreased $1.2 million sequentially to $183.7 million.
  • Cash and equivalents totaled $182.6 million.
  • Debt totaled $32.6 million.

Key Segments

Computing

Computing revenue increased 6.0% YoY but decreased (0.5%) QoQ to ~$76.0 million. This decelerated further from 8.6% growth last quarter. Management said that the segment’s results were slightly better than typical seasonality, though they were “slightly worse than our original expectation for slight sequential growth.”

During the quarter, PC and graphics card strength was offset by seasonal declines in notebooks and tablets. Servers and AI GPUs were “softer as the industry prepares for the next platform transition.” 

Management added in the prepared remarks: “Within AI for large data centers, we are a contender in the middle stages of the design-in phase and we see potential for these products to contribute to revenue in the middle of the calendar year. On graphics cards, the next generation platform is ramping up to mass production. With this new platform, we expect BOM content to increase as more power stage ICs, paired with our controller, are being used to power the GPU.”

For Q3, management said Computing revenue will “will likely decline due to seasonality,” with PCs expected to be flat as “tariff uncertainty is leading to demand pull-in with PC makers.”

Consumer

Consumer revenue declined (3.9%) YoY and (28.8%) QoQ to ~$22.5 million, in line with management’s expectations due to seasonality in gaming and appliances, with wearables normalizing after reaching a record level last quarter.

Management added that they do not expect “gaming to return to meaningful growth until the customer transitions to the next platform.” For Q3, revenue is forecast to see a low-single digit decline QoQ due to continued seasonality for gaming and TVs, with some softness in home appliances.

Communications

Communications revenue rose 14.2% YoY but declined (6.4%) QoQ to $33.3 million. Management explained that this was ahead of expectations for a double-digit QoQ decline as US and China smartphone demand only moderated slightly, with Korea increasing in preparation for product launches.

Management added that the “better-than-expected results are due to combination of market share gains, a mix-shift to higher end phones in China, and generally higher charging currents driving increased BOM content.”

Looking ahead to Q3, management expects a low-teens QoQ decline due to seasonality.

Power Supply and Industrial

Power Supply and Industrial revenue was flat YoY and up 9.6% QoQ to $34.9 million, driven by strength in quick chargers and power tools, with steady demand for e-mobility and AC-DC power supplies. Management said that they see opportunities ahead in 2025 for “quick chargers due to increased BOM content driven by higher charging currents,” while they are working on “leveraging relationships in Taiwan to partner on DC fans for server racks.”

For Q3, management forecast a low-teens QoQ decline due to seasonal declines in quick chargers, offset by e-mobility and AC-DC power supplies.

Earnings Call

AOSL’s Commentary about Being a Potential B300 Supplier

Reading between the lines suggests that AOSL is stating they are in the qualification stages for Nvidia’s upcoming B300s, due mid-year this year, as the company is quite explicit in stating they are currently in the design phase for a release mid-year. The B200s are well beyond their design phase, and beginning to ship.

Here is what management stated: “The second category of projects that we are targeting is the AI data center portion. And this is where our solutions are being used on board to power GPUs and these go into their server solutions.

And this is something that we are very excited to take part in, to be able to design into. We are right now still in the middle of that designing phase and we're closely working with that customer in bringing up boards and working with them. And I'm not going to go into details about the design but we did already indicate that this is something targeting mid-year for the launch.”

When pressed again on where they are in terms of timing and content share, the response was the following – indicating the potential for promising things for AOS over the next few months.

“Yes, we certainly think the potential for our business in data centers to be much bigger than that of the graphics portion simply because the usage goes up much more. As you can imagine in these data centers, the power levels are significantly bigger, so it simply requires more power stages to power each of these GPUs.

I don't want to quantify that at the moment, but basically it is something that is multiples bigger, you can say, in terms of the total attempt that we can go after. We know that we're not the only one players going after this, so the share is still to be determined. The final design, everything is still to be finalized as well, too. So we're working closely with the customer. Again, at this point, we can say that we're targeting for a mid-year launch.”

Notably, the company was asked about thermal issues but politely declined responding, stating they would not discuss the designs of their partner or any rumors. They simply reiterated they are “one of the main contenders.”

Margins:

At one point an analyst asked about the long-term target of $1 billion in revenue with a 30% margin, which was stated previously at an analyst day. The CFO corrected him and answered it was their medium-term target, which is incremental positive.

“Yes, our midterm target model is revenue reaches $1 billion. And then at that time, we expect non-GAAP gross margin to be around about 30% range. So that's our midterm target.”

Bill of Materials (BOM) will go up in any AI-design related wins. We’ve covered this in the past here. Additionally, the margins are expected to bottom next quarter with improvement as AOS goes into the June quarter – below is what the CFO stated.

“Sure. I mean, generally, I mean, yes, for the next 12 to 18 months, I would expect that the product mix probably will improve from this March quarter low point of the margin. I would expect probably in the June quarter, we can expect the gross margin on a non-GAAP basis to get back to the December quarter level. So the March quarter was mainly some product there and also the decrease in license and engineering service revenue, and along with the Lunar New Year period, which is not going to help on the margin side. So then, I mean, I would expect that, yes, we can bounce back and recover from there.”

Average sales prices (ASPs) are currently seeing some erosion, yet management pointed toward the new product line being a way to counter this: “Okay, sure. Yes, I mean, for the whole year, ASP erosion was in the range of mid-to-high single digit range on the same product basis. So going forward for this calendar year 2025, we continue to expect the mid to single digit — mid single digit type of ASP erosion. What we do here is we will accelerate our new product rollout to counter the ASP erosion. So with some good opportunities designed in here, yes, we expect beyond March quarter that all of gross margin can bounce back and recover.”

Conclusion:

AOSL sold off following the report as AI-related design wins are not appearing in the financials yet, and meanwhile, the company is seeing headwinds to its margins. Whether we buy more AOSL or sell AOSL will not be determined by this earnings report, it is nearly meaningless until we get more indication of how the design qualification is going for the B300s.

Fortunately, we use technicals to guide our high beta stocks, which can often show strength long before the fundamentals – especially in the case of supply chain whispers (or leaks). Therefore, we will continue following our plan for AOSL while adhering to our stops, yet also being generous with those stops to make sure the volatility does not prevent us from participating in the upside. Should we get the announcement we are seeking, the upside could be sizable. We want to hang in there and see how this plays out.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

Coherent FQ2: Networking Segment up 56% YoY, Keep an Eye on Co-Optics for H2

Posted on by io-fund

Coherent saw positive price action following its FQ2 earnings report with record revenue that beat consensus and EPS growth of 256% YoY. The top line revenue growth of 26.8% does not tell the full AI story, with the company’s AI-related networking segment up 56% YoY and 7% QoQ. When breaking it down by end markets, Datacom within Communications grew 79% YoY and was up 4% QoQ. There are additional AI-related components helping to drive sequential growth in the telecom segment, which was up sequentially for two quarters at 16% YoY growth and 11% QoQ growth.

To refresh your memory, Coherent supplies components for datacom transceivers and for optical interconnects. Transceivers and optical interconnects convert electrical signals into optical signals for fiber optic networks within the data center. Traditionally, optical links have been used for compute and storage servers, yet AI/ML servers are driving an increase in demand for the optical interconnects and transceivers that Coherent supplies.

As we look at larger AI systems ramping this year, there is some evidence that issues around single-rack Blackwell systems have been resolved yet multi-rack interconnects continue to see issues with overheating. Even though Nvidia has stated they are on schedule with Blackwell, the B300s and next generation Rubin will very likely require more fine-tuning.

Co-packaged optics (CPOs) are gaining widespread attention as a possible solution with rumors Nvidia will announce a CPO switch at the upcoming GTC conference. Although a bit early still, Coherent would be a strong contender in partnering on co-packaged optics. We take a look at this and the company’s strong financial performance in the report below.

Revenue Hits a New Record in FQ2

Coherent refined its methodology to report non-GAAP means starting in FQ2 2025 and recast the prior quarter’s non-GAAP numbers. The updated figures were used in this analysis.

FQ2 revenue grew by 26.8% YoY and 6.4% QoQ to a record $1.44 billion, beating consensus estimates for $1.37 billion by 5.11. Revenue strength was attributed to robust AI-related data center demand and growth in its telecom business, which offset weakness in its industrial segment.

  • Management FQ3 guide is $1.39B to $1.48B, representing YoY growth of 18.7% and flat QoQ, beating estimates by 2.9%. Management FQ3 adjusted EPS guide is $0.85 at the midpoint, representing 123.7% YoY growth and beating estimates by 11.8%.
  • Last quarter, revenue growth was up 28% YoY and up 3% QoQ to $1.35 billion. This beat consensus estimates for $1.32 billion by 2.4%.

When asked about the flat QoQ revenue on the call, the CEO stated it was not due to the AI-related segments: “It's pretty straightforward. We expect datacom and telecom to be up sequentially. And we expect the rest of our industrial related businesses to be down sequentially. And that net at the midpoint to be flat.”

Margins Expands for Fourth Consecutive Quarter

Disciplined operating expense management helped to drive strong improvement in operating margin.

  • FQ2 Non-GAAP gross margin rose 363 bps YoY to 38.2%, compared to 34.6% in the same period last year. Non-GAAP gross profit rose for the fifth consecutive quarter to $548 million.
  • Management guided FQ3 non-GAAP gross margin between 37% to 39%, mid-point of 38%, on non-GAAP gross profit of $545.3 million.
  • Non-GAAP operating margin was 18.5%, up from 13.5% in FQ2 2024 with FQ2 marking the fourth consecutive quarter of expanding margins.
  • The non-GAAP operating margin guide for the next quarter is 17.4%. Management is making progress towards its goal of achieving above 40% non-GAAP gross margin, driven by pricing optimization and product cost improvements.

The goal is to see 40% gross margin as pointed out in our Coherent writeup, and was reiterated in the opening remarks:

“And if you recall back when we — several quarters ago, when we put out our long-term gross margin targets of greater than 40%, we talked about our gross margin expansion strategy and the elements that comprise that. Product cost or cost reductions rather are part of that, but also pricing optimization. And so, you know, those areas continue to be, you know, areas of focus for us as we drive toward the long-term model of greater than 40%. And so pleased to see the results in Q2 and, you know, and we'll continue to focus on that going forward.”

Non-GAAP EPS Grows 256% YoY, Newly GAAP Profitable

FQ2 non-GAAP EPS rose 256% YoY to $0.96, beating consensus estimates for $0.67 by 41.8%. Coherent also generated its first GAAP profit of $0.44 in five quarters.

Management FQ3 adjusted EPS guide is $0.85 at the midpoint, representing 123.7% YoY growth and beating estimates by 11.8%.

Our previous analysis pointed out that Jeffries believes by driving more efficiencies, that Coherent can double adjusted EPS from about $3.00 to $6.40 annual EPS as soon as 2026.

Steady Cash Flow Growth and Chipping Away Debt  

FQ2’s operating cash flow hit a five-quarter high of $187.42 million as the operating cash flow percentage rose steadily for the fifth consecutive quarter to 13.10%.

Free cash hit its highest levels for the fifth straight quarter at $81.7 million, and the free cash flow percentage also improved for the fifth consecutive quarter, hitting a high of 5.7%.

Cash flow growth has been consistent. Coherent closed the quarter with $917.8 million in cash and cash equivalents. Debt has been consistently shrinking for the fifth consecutive quarter as the company has paid down $132 million to $3.86 billion.

Please note, Coherent is looking to divest or shut-down non-strategic product lines and assets, which will help margins and to help reduce debt. Read more here.more here.

Key Metrics/Segments

The data center and communications market has grown 58% YoY and now comprises 57% of total revenues, which offsets the softer industrial market which was flat YoY and accounts for 31% of total revenues. Instrumentation at 7% and electronics round out the smallest market at 5%.

Coherent has three key revenue segments.

Networking segment revenues rose 56% YoY and 7% QoQ to $815.9 million, compared to 61% YoY and 12%QoQ growth to $763 million in the previous quarter. Ongoing AI data center demand and continued recovery in telecom were the key drivers. This segment generates 49% of total revenue.

Lasers segment revenues rose 6% YoY and 8% QoQ to $375.3 million, compared to 4% YoY growth and (2%) QoQ drop to $348 million in the previous quarter. Excimer annealing lasers in the display capital equipment business were the main driver. This segment generates 29% of total revenue.

Materials segment revenues fell (4%) YoY and grew 3% QoQ to $243.5 million, compared to (15%) YoY and (3% ) QoQ drop to $237 million in the previous quarter. The weak automotive and market demand was responsible for the drop. This segment generates 22% of total revenue.

End Markets:

Communications – Datacom:

Communications across both datacom and telecom had growth of 58% YoY and 6% QoQ to $823M. Per the opening remarks: “The sequential and year-over-year increases were driven by another strong quarter of growth in our datacom revenue and a second quarter of sequential growth in our telecom revenue.“

CEO Jim Anderson noted Coherent’s record FQ2 datacom revenue rose 79% YoY and 4% sequentially, driven by ongoing strong data center demand. Management continues to see an expanding number of customers adopting and ramping up its 800G transceivers, and 400G and below transceivers remain strong. Coherent remains on track to ramp up sales in 2025 of its 1.6T datacom transceivers as customer engagements continue to expand. They are developing 3.2T transceivers and have the deepest portfolio of photonic technologies.

Communications – Telecom:

Telecom revenue grew 11% YoY and 16% QoQ, driven by data center interconnect, with some improvement in the traditional transport market, too. Continued ramp of new products, including 100G, 400G and 800G ZR and  ZR+ Coherent transceivers, are expected to continue to ramp in the coming quarters.

CEO Anderson stated they are moving from cautious to cautiously optimistic, with their telecom market up 16% QoQ and 11% YoY in FQ2, marking its second consecutive quarter of growth, “And we are expecting the telecom revenue to be up again sequentially in our fiscal Q3. So given all of that, we are cautiously optimistic that we've moved beyond the bottom of the valley in terms of the telecom recovery and demand, and we're on the upslope within that market.” New products and 100G, 400G, and 800G ZR/ZR+ Coherent transceivers drove growth.

Industrial:

Industrial end markets was up 7% YoY and up 3% QoQ to $437M, including display capital equipment and Excimer lasers for OLED screen manufacturing. Expanding OLED adoption in smartphones and adoption in larger format devices like laptops, tablets and computers is driving display strength.

Instrumentation and Electronics were both down YoY but up nominally QoQ. These are small segments >$100 million in revenue.

Valuation

Coherent trades at forward price-earnings (P/E) of 29.86.

The price/sales (P/S) ratio is 2.75, forward P/S is 2.51.

Earnings Call:

AI Datacom Transceivers and Co-Packaged Optics (CPO):

Coherent offers a range of datacom transceivers, and this is one of the company’s key advantages as AI networking is in flux, and Coherent can serve any request whereas other competitors specialize in only one transceiver technology. These transceiver technologies include VSCELs, EMLs, and CW lasers for silicon photonics.

Along the lines of what we’ve been discussing in our Q1 webinar, Coherent is in agreement that the importance of key networking technologies is going to increase as AI clusters grow in size.

Here is what was stated on the call:

“And we believe pluggable transceivers will continue to grow especially in the scale out domain over the coming years. And so we believe we're well positioned to grow not just, because the TAM is expanding, but because we believe we're well positioned for share gain as well.”

Looking beyond traditional pluggable optics, there is an increasing amount of discussion around co-packaged optics (CPOs), which places the optical transceivers directly on the chip package, rather than using separate optical modules. This results in faster data transmission, reduced latency and higher bandwidth. This may be the best of both worlds: the performance of optical yet with reduced power consumption. Tracking this is especially important as since we last covered copper/Semtech, there have been reports that copper is “causing concurrent issues with overheating and glitching” with rumors Nvidia will launch a CPO switch at the upcoming GTC. That could mean Coherent will be a lead supplier for the anticipated CPO switch – we will be monitoring this closely.

Here is what was stated on the call:

“But the second effect that we think that we're expecting, and this is kind of where CPO plays into, is we're expecting there to be a greater proportion of those interconnects between the computing nodes that switch from what are today electrical connections towards optical connections moving forward. And so think about it as the proportion of electrical versus optical, the proportion of optical connections, we believe increases over the coming years.

And the reason for that is if you look at the bandwidth that's required to support the connections between these computing nodes, and those computing nodes could be GPUs, CPUs, or XPUs, some sort of accelerators, the bandwidth is ramping up significantly. And optical connections can provide a much higher level of bandwidth than an electrical connection. And we think CPO is one of those enabling technologies. There'll be other enabling technologies that help enable the TAM expansion or the replacement of electrical connections with optical connections. And we see this CPO primarily, we believe the biggest sort of area of application for it over the long-term is in scale up. So within the rack or within the box connections between the computing nodes is where we believe we'll see the most prevalence over the long-term. And so the net is we believe CPO is a net accelerator of the overall TAM for optical networking and the data center. “

Here was another quote regarding the importance of CPOs and how Coherent will participate:

“But the second effect that we think that we're expecting, and this is kind of where CPO plays into, is we're expecting there to be a greater proportion of those interconnects between the computing nodes that switch from what are today electrical connections towards optical connections moving forward. And so think about it as the proportion of electrical versus optical, the proportion of optical connections, we believe increases over the coming years.”

Optical Switch Platform

The new data center Optical Circuit Switch (OCS) platform is progressing well, and Coherent has secured its first customer order in FQ2, significantly expanding its addressable market in data centers. Unlike other mechanical MEMS-based solutions, their platform uses field-proven digital liquid crystal technology that provides tremendous advantages for customers. Initial OCS revenue is expected in the calendar of 2025, yet will not be a significant contributor until 2026/2027.

“On the revenue ramp, we remain, our view remains that the revenue, we should start to see first revenue in this current calendar year. We'll probably get better, you know, quantification of that as we move throughout the year, but I think given that we would start to see revenue this year would be more of a contributor in calendar ‘26 and ‘27 and beyond”

Indium Phosphide Capacity Triples

Coherent emphasizes its ability to build many of its components yet also source when needed. They reiterate their supply chain resiliency on earnings calls to stand out against competitors who run into sourcing issues. FQ2 indium phosphide production grew 300% YoY, enabling rapid YoY growth of its 800G transceivers, as it’s the key technology behind EML (electro-absorption modulated laser) and CW (continuous wave) lasers. The U.S. CHIPS Act funding will help the expansion of the in-house indium phosphide platform at the Sherman, Texas facility.  

Coherent is one of the few photonics companies with significant vertical integration. The Company grows its crystals and has had an in-house indium phosphide platform for two decades, which it is expanding.

Semiconductor capital equipment saw healthy sequential YoY growth where Coherent’s lasers and advanced materials are critical enabling technologies.

According to management, there were no significant impacts are expected from tariffs in FQ3, and the guidance already factors in any anticipated effects. Additionally, the company's robust and resilient supply chain further supports its ability to mitigate potential tariff challenges. There wasn’t and pull forward of demand from customers to stockpile ahead of tariffs.

Conclusion:

As you know, we’ve covered big tech capex for four years now to help us cement our AI positions. Although in years past, we positioned with the larger design companies, we are keen to position for the smaller networking component suppliers this year.

Our Members should note that this is all in a state of flux, so we will keep you informed as we go along, yet our goal at the I/O Fund is to strategically participate in Nvidia’s AI dominance throughout 2025. You can expect our portfolio to remain agile as we track these twists and turns. According to Trend Force, co-packaged optics could ship in “mass production” as soon as August. In the meantime, Coherent is putting up strong AI-related growth at exactly the right time, and we will not hesitate to buy more of this quality company with a management team that tripled EPS, has plans to reduce debt, and has transparent discussions on their plans increase their margins.

Jea Yu, Equity Analyst at the I/O Fund, contributed to this article.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

Himax Technologies: A Future Key Player in Silicon Photonics

Posted on by io-fund
  • Himax Technologies produces display imaging processing solutions, including display driver chips, timing controllers, CMOS image sensors, microdisplays and WiseEye AI chips and modules.
  • Himax Technologies is a leading provider of endpoint AI solutions, particularly through its ultra-low-power WiseEye ASIC AI processors and modules.
  • Himax Technologies could be a major player in silicon photonics, a revolutionary semiconductor technology that uses light (photons) instead of electricity (electrons) to transmit data within and between computer chips.
  • Silicon photonics enables faster, more efficient data transfer using less power, generating less heat with more data capacity, which is crucial for AI and HPC applications.
  • Taiwan Semiconductor Manufacturing Company is a leading innovator in silicon photonics with its COUPE (compact optical engine) technology enabling high-bandwidth and low-latency communications between chips, and Himax is rumored to be an exclusive key supplier of micro-lens arrays.

Himax Technologies (NASDAQ: HIMX) is a Taiwan-based semiconductor company specializing in display imaging processing technologies. The fabless company’s product portfolio encompasses display driver integrated circuits (DDIC), touch display driver integrated circuits (TDDI), timing controllers (Tcon), liquid crystal on silicon (LCoS) microdisplays and complementary metal-oxide semiconductors (CMOS) image sensors. Its products cater to a range of applications in laptops, tablets, TVs, monitors, smartphones, automobiles and augmented reality (AR) and virtual reality (VR) devices. 

Their main business is supplying display driver chips (DDIC and TDDI) to original equipment manufacturers (OEMs). DDICs are required in almost every electronic device with a screen and TDDI is required for every device that has a touch screen. Himax has a 50% market share of TDDI in the global automotive market. Himax’s end-user products are primarily used by consumers, which lends to the cyclical nature of its business.

WiseEye Ultralow Power AI Sensing and Applications

Himax’s WiseEye AI sensing technology brings computer vision AI to endpoint devices. Its CMOS image sensors include near-infrared (NIR) and RGB and ultralow power always-on sensor (AoS). Himax’s AI is tailored for endpoint devices. Endpoint AI is not to be mistaken for edge AI:

  • Endpoint AI focuses on implementing AI directly on end-user devices, such as smartphones, tablets, laptops, smart home devices, and wearables. It aims to enhance user interaction and provide personalized, localized experiences. In other words, AI models run on these devices, enabling features like voice assistants, facial recognition or smart recommendations without relying on cloud processing.
  • Edge AI, on the other hand, refers to AI that processes data on devices located near the data source in a network. These devices are typically at the "edge" of the network, such as sensors or cameras. The key idea is that data is processed locally on these devices without needing to be it to the cloud for analysis, which reduces latency and bandwidth usage.

Endpoint AI is designed for local image and AI smart sensing. These are used in devices like smart doorbells, smart home cameras, smart glasses, and fitness trackers for motion sensing, object recognition, activity tracking and gesture base controls. It’s also used in smartphones and tablets for facial recognition, gesture control and augmented reality. Its WiseEye AI chips enable real-time monitoring and image-based decision-making as AI tasks are performed locally and directly on the device.

Rather Than Diversification, Himax is Exposed to Consumers

While Himax's products span across many industries and applications, their diversification is limited because of their product's end-user, consumers. Unlike Amphenol, with balanced end market exposure and true diversification in short cycles (IT datacom, mobile devices, mobile networks) and longer cycle end markets like industrials, commercial aviation and defense, Himax's business is cyclical.

Endpoint AI is targeted at consumers, typically on personal devices. Edge AI is targeted at industries and enterprises, used in applications like automation, industrial systems and smart cities. It’s this end-user component that has made its results volatile, lumpy and cyclical. A weak consumer lends to weak results for Himax. Additionally, 75% of its revenues are derived from China, and a significant portion of its products are manufactured in China. However, there could be a major catalyst to offset the cyclicality of its business; which is silicon photonics.

The Next Frontier of AI Chip Development: Silicon Photonics

Silicon Photonics is a revolutionary technology that uses light (photons) instead of electricity (electrons) to transmit data within and between computer chips. This enables much faster and more efficient data transfer with less power, generating less heat. This technology is ideally suited for AI and high-performance computing (HPC) applications. There are many advantages to photonics, including:

  • Higher Bandwidth: Light can carry much more data than electrical signals thereby enabling faster transmission of data.
  • Lower Latency: There is reduced delay in the data transfers, which is crucial for real-time applications.
  • Improved Power Efficiency: Optical communications can be much more energy-efficient than electrical connections, especially over long distances (IE, optical fiber).
  • Smaller Size: Integrating optical components on silicon chips allows for smaller and even more compact devices.
  • Less Heat: Light doesn’t generate much heat compared to electricity.

Taiwan Semi and Nvidia Team Up on Silicon Photonics

TSMC is a leader in silicon photonics with its COUPE (Compact Universal Optical Photonics Engine) technology. COUPE combines optical interconnects with advanced packaging techniques, such as CoWoS (Chip-on-Wafer-on-Substrate), to facilitate high-bandwidth, low-latency communication between chips. Speculation is that TSMC has been working with Nvidia to develop a silicon photonic-based chip prototype. It’s rumored they’ve created one at the end of 2024. They are also working on optical packaging technologies to improve AI performance and usher in a new packaging architecture for optoelectronic chip integration.

Here’s How Himax May Become a Key Player in Silicon Photonics

What Does This Have to Do with Himax? According to TF International Securities analyst Ming-Chi Kuo, Himax exclusively supplies their micro-lens arrays to TSMC for the first and second generations of COUPE FAUs. Himax exclusively supplies their micro-lens arrays to TSMC for the first and second generations of COUPE FAUs. FAUs are fiber array units, which are essential components of optical engines, helping them to focus and direct light for efficient data transmission. Since their components are vital for advanced packaging technology to enable high-performance computing for AI, they are a key upstream player in the AI ecosystem. 

Ming-Chi Kuo said via X, "Himax Emerging as a Key AI Upstream Winner in TSMC's COUPE (Silicon Photonics), Significantly Boosting Growth Visibility for 2026-2028." He further stated, "I previously shared my prediction that Himax might be a potential supplier for TSMC. My latest supply chain survey indicates that TSMC's COUPE (silicon photonics) development and ecosystem visibility have improved markedly. Furthermore, it's confirmed that Himax is the exclusive supplier of micro-lens arrays for the first and second generations of COUPE FAUs).” 

Kuo estimates Himax revenue will reach $1.16 billion vs consensus estimates of $1.11 billion in 2026, $1.42 billion in 2027 and $2.4 billion in 2028. Kuo estimates EPS of $1.00 vs consensus estimates of $0.80 for 2026, $1.60 in 2027 and $3.40 in 2028. Kuo also states that the first generation of COUPE is fully developed with mass production validation on the way and receiving the highest priority among TSMC’s developing technologies. Second-gen mass production validation is expected in the first half of 2026. Kuo believes Nvidia's Rubin GPUs could use COUPE upon mass production.

Himax’s WLO Technology is Key to Developing Silicon Photonics

Himax is involved with silicon photonics through its strategic partner Fiber Optic Communications Inc. (FOCI), a leading provider of optical communications solutions, which also owns a 5.3% equity stake. WLO technology is essential for manufacturing the miniature and precise optical components needed for silicon photonics like waveguides, diffraction gratings and lenses.

Wafer-level optics (WLO) is a key technology from Himax that’s being used in collaboration with FOCI to leverage their involvement in co-packaged optics (CPO), which involves packaging optical components directly onto semiconductors like CPUs and GPUs, reducing the distance that data needs to travel, improving bandwidth, latency and power efficiency. They also collaborate to develop laser-packaged optics (LPO), which uses lasers for even faster transmission. Himax and FOCI are collaborating to develop and manufacture CPO and LPO solutions, combining Himax's WLO expertise with FOCI's optical interconnect technology.

Himax, with its partnership with FOCI, could also be a key supplier of WLO components for TSMC's COUPE platform. These components would be essential for enabling the optical interconnects within CPO-based chips.

In its Q3 2024 earnings presentation, Himax listed WLO as a growth opportunity and noted, “Collaborating with the world's leading AI semiconductor and foundry partner in LPO/CPO, incorporating FOCI's proprietary LPO/CPO connector technology with Himax's nano-scale WLO to create an industry-leading optical transmission solution for Generative AI and HPC” The “foundry partner” is referring to TSMC.

Himax’s Three Subsidiaries

The Company operates three subsidiaries under the Himax umbrella. While each division operates with some degree of independence, they aren’t completely separate entities, and the Company doesn’t report revenues by division. These subsidiaries are:

  • Himax Technologies LTD: This is the main division that produces its display driver integrated circuits (DDIC), for which they have an 8% global market share. They produce thin-film transistor liquid crystal display (TFT-LCD), electrophoretic display (EPD), organic light-emitting diode (OLED), and LED display drivers. It provides technologies for touch sensor displays, including pure in-cell touch TDDI as well as 3D decoder processors, ASIC service and IP licensing, wafer level optics (WLO), power management chips, WiseEye AI processors and modules. They have in-house WLO fabs and color filter fabs. End markets include 4K/8K TVs, gaming monitors, smartphones, tablets, smart speakers and automobile infotainment systems.
  • Himax Display Inc.: This division focuses on microdisplay products. They produce the liquid crystal on display (LCoS) modules for head-mounted display, heads-up display (HUD) and pico projector applications. They have in-house LC and module display facilities.
  • Himax Imaging LTD: This division specializes in CMOS image sensors and ultralow power always-on sensors (AoS) CMOS image sensors. CMOS image sensors record light data to capture images and video, which are often used in digital cameras, webcams and scanners. Its end-user applications include monitors, smart TVs, ADAS, smart home devices, augmented reality (AR)/virtual reality (VR) headsets, digital cameras, webcams, facial recognition and advanced driver assistance systems (ADAS). This division also produces their micro-lens arrays.

Financials: Cyclical, But Still Beat Top and Bottom-Line Consensus

A weak consumer base, especially in China, and an inventory glut caused YoY sales to decline. Q3 2024 QoQ sales also declined due to the stronger Q2 2024 ramp-up from consumer shopping holidays in China. A surge in Q2 also comes from the highly publicized, overhyped run-up to the Chinese government’s stimulus efforts to bolster its fledging economy. Markets were propped up in anticipation of a major stimulus package, which was released at the end of September 2024 with monetary easing through interest rate cuts and easing of the reserve requirement ratio (RRR) for banks to inject liquidity into the economy.

While there was a short-term boost to China's stock market and some economic indicators, the longer-term effectiveness remains to be seen. Himax’s revenues are highly dependent on the Chinese economy as they derive 75% of their revenues from China. The recovery of their automotive market was a boon for their earnings as their higher margin automotive TDDI and in-cell touch displays got a boost, and its latest technology went into mass production in Q3. 

Himax reported Q3 2024 EPS of $0.07, beating the consensus estimate of $.06 by a penny. Revenues fell 6.75% YoY to $221.41 million but still beat the consensus analyst estimates for $219.96 by 1.11%.

Revenue Fell in Q3 Due to Softer Consumer Market

Q3 revenue fell (6.75%) YoY and (7.18%) QoQ to $222.41, beating consensus analyst estimates for $219.96 million by $2.45 million of 1.11%. The YoY revenue decline was attributed to the softer Chinese economy, but order momentum was strong in automotive, tablet and Tcon products.

Three Categories of Revenue Reported in Q3 2024  

While Himax has three subsidiaries, as mentioned earlier, they only report revenue based on three specific categories:

Large-sized display drivers: $30.7 million in revenues, down 21.2% QoQ due to weaker monitor and TV IC sales attributed to customers' de-stocking amid challenging market conditions following substantial Q2 replenishment for shopping festivals. 

Small and Medium-sized display drivers: $155.4 million in revenues, down 2.2% but beating guidance of being down in the low-teens attributed to stronger-than-expected automotive and tablets markets. Automotive drive sales, including TDDI and DDIC, had mid-single digit declines, which beat estimates of decline in the high teens.

Non-driver products: $36.3 million in revenues was a 13.1% QoQ decline, primarily from the double-digit sequential decline in Tcon sales. Again, due to customers pulling forward inventory purchases in the prior quarter in anticipation of strong sales during the shopping festivals.

This better-than-expected result was primarily fueled by rush orders from Chinese panel customers shortly after Himax's last earnings call on the backdrop of the Chinese government's renewed trade-in stimulus announcement made in mid-August as part of their efforts to boost automobile consumption further. Himax's automotive business, comprising drivers, Tcon, and OLED 3 sales, remained the largest revenue contributor in the third quarter, representing nearly half of total sales.

The small and medium-sized driver IC segment accounted for 69.9% of total sales for the quarter, compared to 66.3% in the previous quarter and 67.6% a year ago. Sequential declines were due to customers pulling forward their inventory purchases in the prior quarter, anticipating strong sales during shopping festivals.

EPS: Q3 EPS Beats After a Lumpy Pull-In Q2

Himax reported Q3 EPS of $0.07 vs $0.06 consensus estimates, a penny beat. The Company guided Q4 EPS between $0.93 to $0.11, mid-point $0.10. Q2 2024 had an unusual high EPS of $0.17 due to the extra inventory customers ordered in anticipation of Chinese shopping holidays including the “520 Festival” and “618 Shopping Festival”.

Margins: Consistent Gross Margins and Annual Bonuses Drop Operating Margins in Q3

Himax remained consistent on its gross margin, reporting 30% for Q3 2024. The operating margin dropped to 2.6% due to the annual spike in operating expenses. Operating expenses rose 28.4% QoQ to $60.8 million, again primarily due to employee awards and bonuses, which are paid out at the end of September each year. Himax grants annual bonuses at the end of September every year.

Cash and Debt Levels Drop Sequentially From Annual Bonuses

The sequential drop in cash from $253.8 million in Q2 to $205.5 million in Q3 2024 was primarily due to $50.7 million paid in annual dividends. The ($1.3 million) in operating cash flow was primarily due to the $30.1 million paid out in employee bonuses. Himax had $36 million in debt, of which $6 million was the current portion.

Conference Call: More Hints on Himax’s Future WLO Prospects

CEO Jordan Wu stated that the macro environment remains challenging. Their panel customers are reducing production to stabilize panel prices in response to current market conditions. End brands are simultaneously taking a cautious approach to procuring panels and are maintaining lower inventories. These factors have compressed chip demand, causing Himax to take a conservative outlook for Q4. The global economy is still uncertain, but they are confident in its business outlook for automotive, AI, wafer level optics (WLO) and organic light emitting diode (OLED), which are its key growth drivers. 

  • The automotive display market is its primary revenue contributor and continues to expand. New and cutting-edge technologies LCD, TDDI, OLEN and Tcon will continue to provide sustainable long-term growth. There have been significant fluctuations in the automotive market demand in the Chinese market, which accounts for over 30% of global vehicle sales.
  • WiseEye PalmVein provides high security and reliability with low false acceptance and rejection rates, using unique internal vein patterns to prevent replication or spoofing. With local identification processing to enhance privacy, Himax anticipates strong sales growth and expanding applications for the WiseEye PalmVein module starting next year.
  • Large-panel display driver chips are expected to have double-digit sales decreases in Q4 due to soft holiday shopping demand expectations and ongoing customer de-stocking since Q2. Panel makers are strategically reducing production.
  • CEO Wu commented on presumably TSMC and its WLO business, “Moreover, Himax and FOCI, along with world-leading AI semiconductor companies and foundry partners, have begun new technology development for future generation products. We believe this will create new revenue streams for Himax and make a significant contribution to our total revenue and profit in the coming years.”
  • AI PCs are prompting display upgrades for notebooks. Wu commented, “In the third quarter, our newly introduced in-cell touch TDDI successfully entered mass production for a prominent brand’s first AI PC. Several projects are also in progress with other brands for their upcoming notebook models.”
  • Himax, in collaboration with FOCI, is advancing through the small-scale production phase of their first-generation solution for laser-packaged optics (LPO) architecture and has begun developing next-generation technologies for more complex co-packaged optics (CPO) architectures. WLO and CPO revenue is not expected in Q4.
  • Himax has created nano-scale precision optical systems for LPO/CPO designed to meet the demands of high-speed computing and are key to the success of LPO/CPO optical solutions. Additionally, Himax's WLO expertise is gaining recognition, with increasing engineering collaborations in AR/VR and other applications, positioning WLO to contribute significantly to future revenue and profit.

CEO Wu provided more elaboration on their technology roadmap with WLO. The company is working on next-generation technologies focused on increasing the number of optical fiber lines in advanced multi-chip modules to boost data transmission rates significantly. By bundling multiple chipsets into a single module, the goal is to improve bandwidth, utilizing optical fibers to replace traditional metal wiring for high-speed data transfer.

The Company is focused on packing more optical fiber lines into limited space to achieve higher bandwidth. It requires advancements in optical design and manufacturing, such as improved waveform integrity and precise coupling with photonic ICs. While the timeline is challenging, there is strong customer demand to accelerate the migration from first-generation LPO to more advanced CPO and prepare next-generation products to meet increasing data transmission needs.

CEO Wu concluded the Q&A with, “Without specifics, we believe the same existing capacity will generate substantially more revenue and profit for us as the products for LPO/CPO demand much more sophisticated optical design and manufacturing, compared to those used for our earlier products, which, as I mentioned, is for consumer electronics. So, I hope all address your questions regarding this WLO business.”

Conclusion: Himax has a stable yet cyclical business with its optical display driver chips and various products used by consumers. The Company has been consistent with its gross margins despite falling YoY revenues stemming from an uncertain macroeconomic climate, especially in China. The potential for its future WLO and silicon photonics business is what’s been keeping shares elevated.

Silicon photonics could be the next-generation technology essential for advancing semiconductors and continuing the progress of Moore's Law, particularly for AI and high-performance computing (HPC) applications. This could be the true growth driver that Himax needs to offset lumpiness with the rest of its business, which is overly concentrated in China.

They derive 75% of revenues from China and are highly dependent on their consumers. This poses geopolitical risks also especially with Trump's tariff threats, which are expected to start at 10% on Chinese imports to the U.S. on February 1, 2025. This could further hurt demand for Himax products. While most of their customer base is concentrated in China, a significant number of their OEMs are in China and presumably export their products to the United States.

The speculation that Himax is the exclusive supplier of micro-lens arrays to TSMC for their COUPE platform has added a 30% premium to shares, as it was trading around $6.50 before the December 12, 2024, TF International Securities analyst Ming-Cho Kuos’ blog post.

As for the future of the technology, CEO Wu stated, “We are certainly very excited about the prospect because, you know, if we look at our partners or customers’ projected capacity expansion as well as their projected growth of such high-end 2.5D modules or XPUs, even if we to fully utilize our existing capacity, we can only meet a small fraction of their projected demand.” Wu also clarified that WLO revenues would not be seen in Q4 2024.

The Company is profitable and has $206.5 million in cash with very little debt at $36 million. While cash flow has been decelerating, Q4 would have been the turning point at $27 million or 12% of revenue if backing out the annual worker’s bonus payouts. The stock trades at 10.99X forward earnings and has a 3% annual dividend yield. If the TSMC rumor turns out to be valid, then it may be the magic bullet to smooth out the cyclical nature of its current business model.

Welcome to the I/O Fund’s new Discovery Tier, where we cover a new stock idea on a weekly or bi-monthly basis. We are excited to bring you more coverage from the I/O Fund team geared toward new idea generation only.

I/O Fund Equity Analyst, Jea Yu, contributed to this analysis

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

 Recommended Reading: Recommended Reading:

TSMC Q4 2024 Earnings: Record Profit Led by Robust AI Demand

Posted on by io-fund

TSMC once again reported strong results led by robust AI chip demand. Q4 revenue grew by 37% YoY and 14.4% QoQ to $26.88 billion, beating the midpoint guide of $26.5 billion. The bottom line was even stronger, as EPS grew by 55.6% YoY to $2.24, beating analysts’ estimates by $0.01. Management also expects 2025 to be a strong year and expects revenue to grow close to mid-20s percent, led by strong AI demand.

Revenue

Q4 revenue grew by 37% YoY and 14.4% QoQ to $26.88 billion, driven by strong AI demand. Revenue came at the higher end of the guidance of $26.1 billion to $26.9 billion, beating the midpoint guide of $26.5 billion.

  • Management guide for the next quarter is $25 billion to $25.8 billion, representing YoY growth of 34.6% and down (-5.5%) QoQ at the midpoint. The sequential decline in the first quarter is due to smartphone seasonality.

“Our business in the fourth quarter was supported by strong demand for our industry-leading 3nm and 5nm technologies,” said Wendell Huang, CFO of TSMC. “Moving into first quarter 2025, we expect our business to be impacted by smartphone seasonality, partially offset by continued growth in AI-related demand.”

  • 2024 revenue grew by 30% YoY to $90 billion, driven by robust AI-related demand. Revenue from AI accelerators, which they define as AI GPU, AI ASICs and HBM controllers for AI training and inference in the data center, accounted for close to mid-teens percent in 2024.
  • AI revenue tripled in 2024 and management expects it to double in 2025, as strong AI-related demand will continue in 2025. For the full year 2025, management expects total revenue to grow close to mid-20s percent in US dollar terms.
  • Management expects AI accelerators to grow mid-40% CAGR for the next five years and expects AI accelerators to be the strongest driver of the HPC platform growth and the largest contributor in terms of the overall incremental revenue growth in the next several years.
  • For the long term, management expects total revenue growth to grow 20% CAGR, driven by growth in all four platforms: smartphone, HPC, IoT, and automotive. 

Margins

Margins continue to expand due to cost controls, higher capacity utilization rates, economies of scale, and better price negotiation with customers.

  • Q4 gross margin was 59%, up from 53% in the same period last year and 57.8% in Q3. The strong gross margin was driven by a higher capacity utilization rate and productivity gains, partially offset by the dilution of 3-nanometer ramp-up.
  • Management has guided gross margin to decrease 100 bps sequentially to 58% at the mid-point, primarily due to ramp-up costs associated with N2 and CoWoS expansion, and the start of dilution from the overseas fabs. However, gross margin is expected to be up from 53.1% in the same period last year.
  • Operating margin improved to 49% from 41.6% in the same period last year and 47.5% in Q3. It beat the midpoint guide of 47.5% driven by operating leverage. Management guide for next quarter is 47.5%, up from 42% in the same period last year.
  • Net income grew by 54.7% YoY to a record $11.6 billion or 43.1% of revenue compared to 38.2% in the same period last year.
  • Return on equity improved to 36.2% from 28.1% in the same period last year.
  • 2024 gross margin improved to 56.1% from 54.4% in 2023, driven by cost controls, improvements in overall capacity utilization, and partially offset by 3-nanometer dilution and higher electricity costs.
  • 2024 operating margin improved to 45.7% from 42.6% in 2023, which was helped by operating leverage.
  • Management expects inflationary cost pressures and higher electricity prices in Taiwan will negatively impact the gross margin by at least 1% in 2025. In addition, the ramp-up costs associated with N2 and further conversion of N5 to N3 capacity will negatively impact the gross margin by about 1%. Also, the overseas fabs are expected to have 2% to 3% margin dilution every year. Management is working with its customers to negotiate better prices to improve the margins; along with it the dilution from the N3 ramp is expected to gradually reduce in 2025 and the overall utilization rate to moderately increase in 2025.
  • Over the long term, management has reiterated that a 53% or higher gross margin is achievable.

EPS

EPS grew by 55.6% YoY to $2.24, beating estimates by $0.01, driven by cost controls, higher capacity utilization rate, and economies of scale.

  • Analysts expect Q1 EPS to grow 49.5% YoY to $2.06 and Q2 EPS to grow 44.6% YoY to $2.14.
  • Looking further out, analysts expect 2025 EPS to grow 32.1% YoY to $9.27 and 16.1% YoY to $10.76 in 2026.

Cash Flows and Balance Sheet

The company’s financial stability is evident in its stable cash flow generation.

  • Q4 operating cash flow was $19.2 billion or 71.4% of revenue compared to 63.1% in the same period last year.
  • Q4 free cash flow was $8.0 billion or 29.8% of revenue compared to $7.05 billion or 35.9% in the same period last year. Capex increased 114.3% YoY to $11.23 billion.
  • The company spent $29.8 billion in capex in 2024, down (-2.3%) YoY. However, due to strong AI demand, capex is expected to increase in 2025 to $38 billion to $42 billion, up 34.4% at the midpoint.
  • Cash and marketable securities were $73.9 billion and debt of $30.1 billion compared to $68.5 billion and $30.6 billion at the end of Q3.

Revenue by Platform

As the leading foundry for AI accelerators, TSMC is riding the enormous wave of demand from Big Tech. The chipmaker’s high-performance computing (HPC) revenues rose 19% QoQ to a record $14.25 billion and accounted for 53% of revenue, surpassing the 50% mark for the third time. Management expects AI accelerators to be the strongest driver of the HPC platform growth in the next several years.

The chart below also shows that HPC revenue reached a record $14.25 billion – its largest sequential increase to date at ~$2.26 billion.

Smartphone grew by 17% sequentially to account for 35% of revenue from 34% of revenue in Q3.

The IoT decreased (-15%) sequentially to account for 5% of revenue. Automotive increased 6% to account for 4% of revenue. Digital Consumer Electronics decreased (-6%) to account for 1% of revenue. Others increased 2% and accounted for 2%.

Revenue by Technology

TSMC’s Advanced nodes are defined as 7-nanometer and below. These nodes accounted for 74% of wafer revenue compared to 69% in Q3.

  • In Q4 2024, 3-nanometer process technology contributed to 26% of wafer revenue, while 5-nanometer and 7-nanometer accounted for 34% and 14%, respectively.
  • In Q3 2024, 3-nanometer process technology contributed 20% of wafer revenue, while 5-nanometer and 7-nanometer accounted for 32% and 17%, respectively.

Earnings Call: AI Accelerator Contribution, Margin Headwinds Discussed

TSMC’s management provided optimistic long-term growth and AI revenue forecasts on the earnings call, while also updating on international expansion efforts, some margin headwinds and capex.

Commenting on the broader industry, CEO C.C. Wei said that he expects the foundry industry to  “grow 10% year-over-year in 2025, supported by robust AI-related demand and a mild recovery in other end market segments.” He believes TSMC can once again outperform the industry’s growth due to its broad customer base and technological leadership, forecasting revenue growth in USD close to the mid-20% range. This came in line with some analyst forecasts for 20% to 25% growth in 2025. Management provided some thin details on the segment breakdown for the year, saying that 2025 is “still a mild growth 2026 for PC and smartphone, but everything is AI-related.”

AI accelerators are playing an increasingly large role in TSMC’s long-term growth – TSMC forecast AI accelerator revenue growth at a 40% CAGR from 2024 to 2029, expecting it to be the strongest driver of HPC growth as well as the “largest contributor in terms of our overall incremental revenue growth in the next several years.”

Though margins were arguably very strong in Q4 – gross margin at the high end of the guided range and operating margin above the guided range – management discussed margin headwinds as Q1 is set to see some slight sequential contraction.

Management explained that they have guided Q1 gross margin to decrease 100 bp to 58% at midpoint due to “ramp costs associated with N2 and CoWoS expansion, and the start of dilution from our overseas fabs.” They clarified that these are a few of six significant factors that determine margins: tech development and ramp-up, cost reduction efforts, platform mix, pricing, capacity utilization, and forex.

For the full year, management said there were a few “puts and takes.” One of the primary headwinds includes 2% to 3% dilution impacts from ramping up production at the Arizona and Kumamoto fabs throughout the year, while other headwinds include at least a 1% impact from electricity price hikes in Taiwan, and a 1% impact from retooling from 5nm to 3nm capacity and some additional 2nm ramp. However, offsetting these headwinds are gradual reduction in dilution from the 3nm ramp this year (3nm now accounts for 18% of revenue), and utilization rate moderately increasing in 2025.

Capex was also discussed, with TSMC seeing 2025’s capex slightly ahead of estimates to support demand and capacity expansion plans. For 2024, capex totaled $29.8 billion, with management guiding for capex of $38 billion to $42 billion, or ~34% growth at midpoint; this compares to the $38 to $40 billion estimated by analysts. Of this budget, “about 70% of the capital budget will be allocated for advanced process technologies, about 10% to 20% will be spent for specialty technologies, and about 10% to 20% will be spent for advanced packaging, testing, mask making and others.”

TSMC was rather tight lipped about capacity expansion plans for CoWoS despite multiple analysts asking for more details about its capacity, ramp and revenue contribution. Management simply said that they have “very tight capacity and cannot even meet customers' needs” and they are working “very hard” to increase CoWoS capacity.

On the advanced node front, the 2 nanometer node is “well on track” for volume production in the second half of this year as previously scheduled, with TSMC expecting it to ramp similarly to the 3nm node. However, management added that the tape out in the first two years is expected to be higher than both 3nm and 5nm in their respective ramps due to the strong customer interest and demand for the new node.

Additionally, management shared some updates on international fab expansion, noting that its first fab in Arizona entered high volume production in Q4 with yields comparable to Taiwan. They expect the 4nm node there to ramp smoothly, with construction on the second and third fab in Arizona (set to produce 3nm, 2nm and A16), on track. Japan’s Kumamoto fab began volume production with “record yield”, while the second specialty fab in Japan is expected to break ground this year. TSMC also shared that they are smoothly progressing with plans to build a specialty automotive and industrial-focused fab in Dresden, Germany.

Conclusion

The company’s strong results showed its resilience in capturing the demand for advanced chips due to its technological leadership. Management was optimistic on the long-term forecast for AI revenue growth over the next few years supported by high demand from customers. Strong top-line growth, along with the bottom-line strength, distinguishes TSMC from other companies that are struggling in the current tough macro environment.

 I/O Fund Equity Analysts Royston Roche and Damien Robbins contributed to this article.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

TSMC Q4 2024 Earnings: Record Profit Led by Robust AI Demand

Posted on by io-fund

TSMC once again reported strong results led by robust AI chip demand. Q4 revenue grew by 37% YoY and 14.4% QoQ to $26.88 billion, beating the midpoint guide of $26.5 billion. The bottom line was even stronger, as EPS grew by 55.6% YoY to $2.24, beating analysts’ estimates by $0.01. Management also expects 2025 to be a strong year and expects revenue to grow close to mid-20s percent, led by strong AI demand.

Revenue

Q4 revenue grew by 37% YoY and 14.4% QoQ to $26.88 billion, driven by strong AI demand. Revenue came at the higher end of the guidance of $26.1 billion to $26.9 billion, beating the midpoint guide of $26.5 billion.

  • Management guide for the next quarter is $25 billion to $25.8 billion, representing YoY growth of 34.6% and down (-5.5%) QoQ at the midpoint. The sequential decline in the first quarter is due to smartphone seasonality.

“Our business in the fourth quarter was supported by strong demand for our industry-leading 3nm and 5nm technologies,” said Wendell Huang, CFO of TSMC. “Moving into first quarter 2025, we expect our business to be impacted by smartphone seasonality, partially offset by continued growth in AI-related demand.”

  • 2024 revenue grew by 30% YoY to $90 billion, driven by robust AI-related demand. Revenue from AI accelerators, which they define as AI GPU, AI ASICs and HBM controllers for AI training and inference in the data center, accounted for close to mid-teens percent in 2024.
  • AI revenue tripled in 2024 and management expects it to double in 2025, as strong AI-related demand will continue in 2025. For the full year 2025, management expects total revenue to grow close to mid-20s percent in US dollar terms.
  • Management expects AI accelerators to grow mid-40% CAGR for the next five years and expects AI accelerators to be the strongest driver of the HPC platform growth and the largest contributor in terms of the overall incremental revenue growth in the next several years.
  • For the long term, management expects total revenue growth to grow 20% CAGR, driven by growth in all four platforms: smartphone, HPC, IoT, and automotive. 

Margins

Margins continue to expand due to cost controls, higher capacity utilization rates, economies of scale, and better price negotiation with customers.

  • Q4 gross margin was 59%, up from 53% in the same period last year and 57.8% in Q3. The strong gross margin was driven by a higher capacity utilization rate and productivity gains, partially offset by the dilution of 3-nanometer ramp-up.
  • Management has guided gross margin to decrease 100 bps sequentially to 58% at the mid-point, primarily due to ramp-up costs associated with N2 and CoWoS expansion, and the start of dilution from the overseas fabs. However, gross margin is expected to be up from 53.1% in the same period last year.
  • Operating margin improved to 49% from 41.6% in the same period last year and 47.5% in Q3. It beat the midpoint guide of 47.5% driven by operating leverage. Management guide for next quarter is 47.5%, up from 42% in the same period last year.
  • Net income grew by 54.7% YoY to a record $11.6 billion or 43.1% of revenue compared to 38.2% in the same period last year.
  • Return on equity improved to 36.2% from 28.1% in the same period last year.
  • 2024 gross margin improved to 56.1% from 54.4% in 2023, driven by cost controls, improvements in overall capacity utilization, and partially offset by 3-nanometer dilution and higher electricity costs.
  • 2024 operating margin improved to 45.7% from 42.6% in 2023, which was helped by operating leverage.
  • Management expects inflationary cost pressures and higher electricity prices in Taiwan will negatively impact the gross margin by at least 1% in 2025. In addition, the ramp-up costs associated with N2 and further conversion of N5 to N3 capacity will negatively impact the gross margin by about 1%. Also, the overseas fabs are expected to have 2% to 3% margin dilution every year. Management is working with its customers to negotiate better prices to improve the margins; along with it the dilution from the N3 ramp is expected to gradually reduce in 2025 and the overall utilization rate to moderately increase in 2025.
  • Over the long term, management has reiterated that a 53% or higher gross margin is achievable.

EPS

EPS grew by 55.6% YoY to $2.24, beating estimates by $0.01, driven by cost controls, higher capacity utilization rate, and economies of scale.

  • Analysts expect Q1 EPS to grow 49.5% YoY to $2.06 and Q2 EPS to grow 44.6% YoY to $2.14.
  • Looking further out, analysts expect 2025 EPS to grow 32.1% YoY to $9.27 and 16.1% YoY to $10.76 in 2026.

Cash Flows and Balance Sheet

The company’s financial stability is evident in its stable cash flow generation.

  • Q4 operating cash flow was $19.2 billion or 71.4% of revenue compared to 63.1% in the same period last year.
  • Q4 free cash flow was $8.0 billion or 29.8% of revenue compared to $7.05 billion or 35.9% in the same period last year. Capex increased 114.3% YoY to $11.23 billion.
  • The company spent $29.8 billion in capex in 2024, down (-2.3%) YoY. However, due to strong AI demand, capex is expected to increase in 2025 to $38 billion to $42 billion, up 34.4% at the midpoint.
  • Cash and marketable securities were $73.9 billion and debt of $30.1 billion compared to $68.5 billion and $30.6 billion at the end of Q3.

Revenue by Platform

As the leading foundry for AI accelerators, TSMC is riding the enormous wave of demand from Big Tech. The chipmaker’s high-performance computing (HPC) revenues rose 19% QoQ to a record $14.25 billion and accounted for 53% of revenue, surpassing the 50% mark for the third time. Management expects AI accelerators to be the strongest driver of the HPC platform growth in the next several years.

The chart below also shows that HPC revenue reached a record $14.25 billion – its largest sequential increase to date at ~$2.26 billion.

Smartphone grew by 17% sequentially to account for 35% of revenue from 34% of revenue in Q3.

The IoT decreased (-15%) sequentially to account for 5% of revenue. Automotive increased 6% to account for 4% of revenue. Digital Consumer Electronics decreased (-6%) to account for 1% of revenue. Others increased 2% and accounted for 2%.

Revenue by Technology

TSMC’s Advanced nodes are defined as 7-nanometer and below. These nodes accounted for 74% of wafer revenue compared to 69% in Q3.

  • In Q4 2024, 3-nanometer process technology contributed to 26% of wafer revenue, while 5-nanometer and 7-nanometer accounted for 34% and 14%, respectively.
  • In Q3 2024, 3-nanometer process technology contributed 20% of wafer revenue, while 5-nanometer and 7-nanometer accounted for 32% and 17%, respectively.

Earnings Call: AI Accelerator Contribution, Margin Headwinds Discussed

TSMC’s management provided optimistic long-term growth and AI revenue forecasts on the earnings call, while also updating on international expansion efforts, some margin headwinds and capex.

Commenting on the broader industry, CEO C.C. Wei said that he expects the foundry industry to  “grow 10% year-over-year in 2025, supported by robust AI-related demand and a mild recovery in other end market segments.” He believes TSMC can once again outperform the industry’s growth due to its broad customer base and technological leadership, forecasting revenue growth in USD close to the mid-20% range. This came in line with some analyst forecasts for 20% to 25% growth in 2025. Management provided some thin details on the segment breakdown for the year, saying that 2025 is “still a mild growth 2026 for PC and smartphone, but everything is AI-related.”

AI accelerators are playing an increasingly large role in TSMC’s long-term growth – TSMC forecast AI accelerator revenue growth at a 40% CAGR from 2024 to 2029, expecting it to be the strongest driver of HPC growth as well as the “largest contributor in terms of our overall incremental revenue growth in the next several years.”

Though margins were arguably very strong in Q4 – gross margin at the high end of the guided range and operating margin above the guided range – management discussed margin headwinds as Q1 is set to see some slight sequential contraction.

Management explained that they have guided Q1 gross margin to decrease 100 bp to 58% at midpoint due to “ramp costs associated with N2 and CoWoS expansion, and the start of dilution from our overseas fabs.” They clarified that these are a few of six significant factors that determine margins: tech development and ramp-up, cost reduction efforts, platform mix, pricing, capacity utilization, and forex.

For the full year, management said there were a few “puts and takes.” One of the primary headwinds includes 2% to 3% dilution impacts from ramping up production at the Arizona and Kumamoto fabs throughout the year, while other headwinds include at least a 1% impact from electricity price hikes in Taiwan, and a 1% impact from retooling from 5nm to 3nm capacity and some additional 2nm ramp. However, offsetting these headwinds are gradual reduction in dilution from the 3nm ramp this year (3nm now accounts for 18% of revenue), and utilization rate moderately increasing in 2025.

Capex was also discussed, with TSMC seeing 2025’s capex slightly ahead of estimates to support demand and capacity expansion plans. For 2024, capex totaled $29.8 billion, with management guiding for capex of $38 billion to $42 billion, or ~34% growth at midpoint; this compares to the $38 to $40 billion estimated by analysts. Of this budget, “about 70% of the capital budget will be allocated for advanced process technologies, about 10% to 20% will be spent for specialty technologies, and about 10% to 20% will be spent for advanced packaging, testing, mask making and others.”

TSMC was rather tight lipped about capacity expansion plans for CoWoS despite multiple analysts asking for more details about its capacity, ramp and revenue contribution. Management simply said that they have “very tight capacity and cannot even meet customers' needs” and they are working “very hard” to increase CoWoS capacity.

On the advanced node front, the 2 nanometer node is “well on track” for volume production in the second half of this year as previously scheduled, with TSMC expecting it to ramp similarly to the 3nm node. However, management added that the tape out in the first two years is expected to be higher than both 3nm and 5nm in their respective ramps due to the strong customer interest and demand for the new node.

Additionally, management shared some updates on international fab expansion, noting that its first fab in Arizona entered high volume production in Q4 with yields comparable to Taiwan. They expect the 4nm node there to ramp smoothly, with construction on the second and third fab in Arizona (set to produce 3nm, 2nm and A16), on track. Japan’s Kumamoto fab began volume production with “record yield”, while the second specialty fab in Japan is expected to break ground this year. TSMC also shared that they are smoothly progressing with plans to build a specialty automotive and industrial-focused fab in Dresden, Germany.

Conclusion

The company’s strong results showed its resilience in capturing the demand for advanced chips due to its technological leadership. Management was optimistic on the long-term forecast for AI revenue growth over the next few years supported by high demand from customers. Strong top-line growth, along with the bottom-line strength, distinguishes TSMC from other companies that are struggling in the current tough macro environment.

 I/O Fund Equity Analysts Royston Roche and Damien Robbins contributed to this article.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

TSMC Q4 2024 Earnings Preview: Revenue beat estimates

Posted on by io-fund

TSMC recently released its December monthly revenue, which grew by a robust 57.8% YoY and 0.8% MoM to NT$278.16 billion. The fourth quarter revenue grew 38.8% YoY to NT$868.5 billion, beating the LSEG estimates of NT$854.7 billion.

In US dollar terms, Q4 revenue grew 36.9% YoY to $26.85 billion using the average exchange rate of $1=NT$32.34. We will get the official USD figures when the company releases its full results on January 16th. This suggests that the company’s revenue will come at the high-end guidance of $26.1 billion to $26.9 billion and beat the mid-point guide of $26.5 billion.

The strong December quarter revenue suggests that AI demand will continue in 2025. Recently, Foxconn also beat estimates with record fourth-quarter revenue driven by AI demand. Revenue grew by 15.2% YoY to NT$2.13 trillion, beating estimates of NT$2.1 trillion. Foxconn management said, “Robust AI server demand led to strong revenue growth for its cloud and networking products division.”

The revenue beat of TSMC and Foxconn provides further hope to investors that AI demand is not slowing down in 2025. TSMC is a major beneficiary of surging demand for AI chips, holding over 90% market share in manufacturing advanced AI processors. Increased investments from Big Tech is further expected to drive AI growth; for example, Microsoft recently announced plans to invest approximately $80 billion in AI-enabled data centers in FY2025, a significant increase from the $55.7 billion capex in FY2024.

Investors will be closely watching the comments from the management on the broader semiconductor market for 2025 since TSMC has better insights than any other company as the world’s leading foundry.

Revenue

Strong revenue growth is expected in the coming quarters due to robust demand for AI and smartphone chips. Over the past few quarters, revenue estimates have risen — 4.9 percentage points for Q4 2024 and 6.1 percentage points for Q1 2025 — indicating analysts’ confidence in the company’s ability to maintain growth.

  • Management’s guide for the fourth quarter is between $26.1 billion to $26.9 billion, representing YoY growth of 35.1% and 12.8% QoQ at the midpoint. Based on the monthly TWD figures, Q4 revenue grew 36.9% YoY to $26.85 billion using the average exchange rate of $1=NT32.34. We will get the official USD figures when the company releases its full results on January 16th.
  • Q3 revenue grew 36% YoY and 12.9% QoQ to $23.5 billion, driven by strong demand for AI and smartphone chips.

C.C. Wei, Vice Chairman and CEO, said in the earnings call, “Moving into fourth quarter, we expect our business to continue to be supported by strong demand for our leading-edge process technologies. We continue to observe extremely robust AI-related demand from our customers throughout the second half of 2024, leading to increasing overall capacity utilization rate for our leading-edge 3-nanometer and 5-nanometer process technologies.”

Due to its technological leadership, TSMC can capture robust demand for the most advanced AI chips. The revenue contribution from server AI processors is expected to triple this year and will account for a mid-teens percentage of 2024 revenue. Management guided for full year 2024 revenue to grow close to 30% in U.S. dollar terms and calculating using the recent monthly figures suggest revenue to grow 29.9% YoY to $90 billion, rebounding significantly from a decline of (-8.7%) for 2023.

Morgan Stanley expects TSMC to provide a conservative annual sales growth guide of low 20% in dollar terms for 2025 at the beginning of the year, with a potential to raise throughout the year as the company did in 2024.

Margins

While many companies face the issue of rising costs, TSMC has demonstrated resilience by effectively managing expenses. Through a combination of cost control measures, negotiating better prices with its customers, and leveraging economies of scale, TSMC has maintained strong profitability. The company's leadership position in the foundry industry further solidifies its competitive advantage.

According to an Economic Daily News report, TSMC plans to increase prices for advanced nodes by 5% to 10% in 2025. The price increase to customers will help to allay investor fears on rising costs from shifting of manufacturing processes outside of Taiwan due to geopolitical risks, increasing electricity prices in Taiwan, inflationary cost pressures, and initial higher costs involved in the ramp-up phase of advanced node manufacturing.

  • Management guided Q4 gross margin to increase 20 basis points sequentially to 58% at the midpoint, up significantly from 53% in the same period last year, driven by higher capacity utilization and cost controls, partially offset by dilution from N3 ramp, higher electricity prices in Taiwan, and N5 to N3 tool conversion cost.
  • Operating margins are expanding, helped by operating leverage. Management’s guide for Q4 is 47.5% at the midpoint compared to 41.6% in the same period last year.
  • Q3 net income grew by 52.4% YoY to $10.06 billion or 42.8% of revenue compared to 38.6% in the same period last year.

EPS

Q3 EPS grew by 50.4% YoY to $1.94, driven by better capacity utilization, cost improvement, and operating leverage. EPS is projected to rise significantly in the coming quarters and analysts have raised estimates indicating greater confidence in the company’s ability to grow its bottom line.

  • Analysts expect Q4 EPS to grow 54.9% YoY to $2.23, up from 35.4% expected growth in mid-October, and Q1 EPS to grow 49.5% YoY to $2.06, up from 26.1% expected growth in mid-October.
  • Looking further out, analysts expect 2025 EPS to grow 32.1% YoY to $9.27 and 16.1% YoY to $10.76 in 2026.

Cash Flows and Balance Sheet

The company’s financial stability is evident due to its strong cash flows, with operating cash flows increasing 30.3% and free cash flows by 166% in Q3. The foundry industry is capital-intensive, and this is why you will notice a wide difference between operating cash flows and free cash flows for the company.

  • Q3 operating cash flow was $12.13 billion or 51.6% of revenue compared to $9.31 billion or 54% of revenue in the same period last year.
  • Free cash flow grew by 166% YoY to $5.72 billion or 24.4% of revenue compared to 12% of revenue in the same period last year. Capex was down (-9.9%) YoY to $6.4 billion.
  • Management expects capex to be slightly higher than $30 billion for 2024, revised down slightly from the previous guide of $30 billion to $32 billion. About 70% to 80% will be allocated for advanced process technologies, 10% to 20% for specialty technologies, and 10% for advanced packaging, testing, mask making, and others.
  • As the company continues to invest due to the strong expected AI growth, capex is likely to increase in 2025, and management mentioned that they will provide more details during the January earnings call.
  • Cash and marketable securities were $68.5 billion and debt of $30.6 billion, compared to $63.05 billion and $30.4 billion in Q2.

Revenue by Platform

As the leading foundry for AI accelerators, TSMC is riding the enormous wave of demand from Big Tech. The chipmaker’s high-performance computing (HPC) revenues rose 11% QoQ to $11.99 billion in Q3 and accounted for 51% of revenue, surpassing the 50% mark for the second time.

TSMC is immensely benefiting from the AI spending of Big Tech companies.  The company has more than 90% market share in manufacturing advanced AI chips. Microsoft also recently outlined its plan to invest about $80 billion primarily in AI-enabled data centers in FY2025, up from $55.7 billion capex in FY2024. This substantial investment underscores the growing importance of AI and bodes well for TSMC's continued dominance in the advanced semiconductor manufacturing sector.

As seen in the chart below, HPC revenue has grown sequentially each quarter in 2024 to reach a record $11.99 billion in Q3.

Smartphone grew by 16% sequentially and accounted for 34% of revenue from 33% of revenue in Q2.

Internet of Things revenue grew by 35% sequentially and accounted for 7% of revenue.

Automotive revenue grew by 6% sequentially and accounted for 5% of revenue. Digital Consumer Electronics decreased by 19% and accounted for 1% of revenue; others grew by 8% to account for 2% of revenue.

Revenue by Technology

The most advanced node in production today is the 3-nanometer process technology. Volume production for 2-nanometer is expected in 2025 and should have a meaningful revenue contribution in the first half of 2026.

According to the MoneyDJ report, TSMC has initiated the setup of a pilot production line for its 2-nanometer process at its Hsinchu Baoshan fab in Taiwan, with an initial monthly capacity of 3,000 to 3,500 wafers. By the end of 2025, TSMC expects to significantly expand its 2nm production capacity to over 50,000 wafers per month, including the contributions from its Kaohsiung fab in Taiwan. By the end of 2026, TSMC is expected to further increase its 2nm monthly capacity to 120,000 to 130,000 wafers, suggesting strong demand for the 2nm chips. TSMC's 2nm node is expected to provide 15% better performance and 35% better energy efficiency than its 3nm node. Apple is expected to be the first customer to use 2-nanometer process technology.

Advanced nodes are defined as 7-nanometer and below. These nodes accounted for 69% of wafer revenue in Q3 compared to 67% in Q2.

  • In Q3 2024, 3-nanometer process technology contributed 20% of wafer revenue, while 5-nanometer and 7-nanometer accounted for 32% and 17%, respectively.

Other Points to Watch

Advanced Packaging

Advanced packaging demand is strong and will account for high single-digit revenue this year. According to Economic Daily News, TSMC’s monthly CoWoS capacity could reach 75,000 wafers in 2025, nearly doubling 2024 levels. Due to the strong demand, primarily from Nvidia, the construction of CoWoS facilities time has been shortened to 1.5 years from the previous three to five years.

Overseas Fabs

The company is expanding its fabs overseas to reduce geopolitical risks. TSMC began producing 4-nanometer chips in Arizona fab, United States Secretary of Commerce Gina Raimondo told Reuters. Earlier in November the US government had also finalized the Chips Act incentive with the company and it is expected to get up to $6.6 billion in grants and $5 billion in loans. TSMC’s second Arizona fab is expected to feature the N3 and N2 processes and is expected to be operational in 2028.

According to TrendForce, the first fab in Arizona will have an initial capacity of 20K wafers per month by year-end. TSMC is also in talks with Nvidia about producing its Blackwell chips at the Arizona plant; these chips are currently manufactured in Taiwan. However, these Blackwell chips will still need to be shipped back to Taiwan since the Arizona facility does not have CoWoS packaging capability. If the contract is finalized, Nvidia will be the third customer for the Arizona plant after Apple and AMD

According to Nikkie report, TSMC’s first plant in Japan will start mass-producing chips by the end of the year. The first fab is expected to produce less advanced chips. The company is also planning a second fab that could produce 6-nanometer and 12-nanometer chips, and is also discussing to build a third fab in Japan that could make advanced 3-nanometer chips.

Trade Wars

Due to its limited China exposure (about 12% of revenue) and the strong global demand for AI chips, TSMC is better positioned to mitigate any potential disruptions from the US-China trade war.

Valuation

The company trades at a P/E ratio of 32.3 and a forward P/E ratio of 22.3. The P/S ratio is 12.6 and a forward P/S ratio of 9.5. In the last five years, the P/E ratio peaked at 41.8 in February 2021 and hit a low of 10.3 in November 2022. The stock is now trading above its five-year average P/E ratio of 24.4.

Conclusion

TSMC plays a crucial role in the success of many leading AI companies, including Nvidia, Apple, AMD, Marvell, and Qualcomm, as it is their primary supplier of advanced semiconductor chips and has more than 90% market share manufacturing advanced AI processors. Given the continued growth of the AI economy, TSMC's importance in the semiconductor industry is expected to further increase.

Royston Roche, Equity Analyst at the I/O Fund, contributed to this article.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading: