AMD heads into its fiscal Q1 report with rather high expectations, after management raised FY2024 GPU revenue forecast by 75% last quarter to $3.5 billion plus. Earnings reports last week from Microsoft, Meta and Google reaffirmed a bullish outlook on AI infrastructure spending for 2024, with all three combining for at least $135 billion in capex this year, with management commentary signaling a bulk of that spend will go to data center infrastructure and GPUs.

Therefore, data center revenue and GPU commentary are in focus this report. Yet, it’s key to be objective, and to note for our Members that some analysts are toning down estimates on the data center. Given the channel checks that analysts can do on a company like AMD, the polarized nature of the commentary going into the print tomorrow is interesting. We include analyst commentary below.

As a reminder, just last quarter analysts were setting a very high bar with some pushing for a $6 billion GPU revenue guide for the full year. We are seeing as low as $4 billion and as high as $8 billion. This wide of a range on a key metric is highly unusual, — and investors should be aware, it’ll be nearly impossible for AMD to beat the high-end of the FY expectations for the data center in Q1.

Therefore, it’s important we come up with our own expectations, of sorts. Broadcom is in second place with $1.5B in AI revenue, or $6B annual run rate. Due to analysts muddying the water a bit here on what is reasonable for a Q1 discussion on the call tomorrow, I think a decent goal (and win) would be for AMD to have GPU revenue equal to 2023 data center revenue by the time we exit the year, which was $6B. This number is also in line with AVGO’s current AI revenue and is the midpoint of analyst estimates.

Revenue and EPS:

• Q1 revenue was guided to be $5.4 billion, +/- $300 million, for YoY growth of approximately 0.9% and a QoQ decline of (12.5%). Analysts are expecting slightly higher revenues of $5.45 billion for YoY growth of 1.9%. As you can see above, Q1 should mark the bottom with a strong ramp into Q1 of next year.
• Q1’s adjusted EPS is estimated to be $0.61, for YoY growth of 1.3%. Adjusted EPS growth is expected to accelerate to 90% by Q1 of next year.

Q2’s guide will be important to track. Susquehanna noted Monday, after lowering its price target, that it is “expecting in-line to slightly weaker guidance as Server/PC/XLNX/Gaming continue to weigh.” Q2 revenue is currently estimated to be $5.69 billion for YoY growth of 6.2%.

Key Segments:

Data Center: AMD guided for Q1 data center revenue to be approximately flat QoQ, implying revenues of $2.3 billion for YoY growth of ~77%. This is due to “a seasonal decline in server sales offset by a strong Data Center GPU ramp.”

Data center revenue, and more importantly, MI300 revenue is likely to be the most critical aspect of Q1’s report and Q2’s guide, as analysts are simultaneously resetting and increasing expectations for MI300 revenues in April. This will be discussed in more detail in the section “MI300 GPU Sales in Focus” below.

For the full year, management said “the largest incremental revenue opportunities are going to come from Data Center between both the server side gaining more share, and Data Center GPU side with the significant ramp up of our MI300.”

Client, Embedded, Gaming: AMD guided for Client, Embedded and Gaming segment sales to “decline sequentially, with semi-custom revenue expected to decline by a significant double-digit percentage.”

Margins:

Adjusted gross margin is guided to be 52% in Q1, a 120 bp QoQ expansion as data center mix increases.

Adjusted operating margin is expected to be 20%, a ~300 bp QoQ contraction and the lowest level in three quarters. However, data center operating margin has expanded significantly over the past two quarters, from 19.1% in Q3 to 29.2% in Q4. For context, DC generated $666 million in operating income in Q4, up 118% QoQ and the most of any of AMD’s segments.

CFO Jean Hu shed more light on both the trajectory of margins and DC margin in Q1 and the rest of the year:

“We guided the Q1, 120 basis points higher than Q4 sequentially, primarily because the higher Data Center contribution actually more than offset the decline of Embedded business in Q1. Going forward, the way to think about it is as you said is the major driver is going to be Data Center business is going to grow much faster than other segment. That mix change will help us to expand the gross margin nicely. I think you also are spot on, the Embedded coming back in second half, which will be a tailwind. With the Data Center GPU, we are at the very early stage of ramp. We are improving testing time yield and continue to expand gross margin and we expect to be accretive to corporate average. So, those are all the tailwinds coming in the second half. I would say the headwinds side continue to be in the first half where we see Embedded business not only Q1 we see sequential decline, Q2 probably are going to be sequentially flattish versus Q1.”higher Data Center contribution actually more than offset the decline of Embedded business in Q1. Going forward, the way to think about it is as you said is the major driver is going to be Data Center business is going to grow much faster than other segment. That mix change will help us to expand the gross margin nicely. I think you also are spot on, the Embedded coming back in second half, which will be a tailwind. With the Data Center GPU, we are at the very early stage of ramp. We are improving testing time yield and continue to expand gross margin and we expect to be accretive to corporate average. So, those are all the tailwinds coming in the second half. I would say the headwinds side continue to be in the first half where we see Embedded business not only Q1 we see sequential decline, Q2 probably are going to be sequentially flattish versus Q1.”

Cash and Debt:

AMD reported operating cash flow of $381 million in Q4 for a margin of 6.1%, and FY23 OCF was $1.67 billion for a 7.4% margin. Free cash flow was $242 million in Q4 for a margin of 4%, and FY23 FCF was $1.12 billion for a 4.9% margin.

Operating cash flow growth is expected to unfold as one of the larger fundamental recoveries in 2024. Current estimates point to nearly 277% YoY growth in OCF to $6.28 billion, or a margin in the 24% range based on current revenue estimates of $25.7 billion for the year.

AMD has $5.77 billion in cash and equivalents on hand, and total debt of $2.47 billion.

MI300 GPU Sales in Focus

As noted earlier, AMD’s MI300 GPU revenue outlook for the full year will be the most important data point coming out of Q1’s report, after AMD increased its outlook by 75% last quarter.

Analysts are hinting towards possible weakness in Q1 due to rumors of Microsoft cutting some orders, though other analysts are expecting full year GPU revenue of $8B, more than double AMD’s guide. This is setting up an interesting scenario in which even if AMD surprises with better-than-expected GPU revenue in Q1, the full year picture may still disappoint against outsized expectations for $4.5B+ all the way to $8B.

Here’s some recent analyst commentary and updated expectations for GPUs:

• Susquehanna analyst Christopher Rolland said it is likely that “upward revisions to the MI300 are ‘necessary’ for the stock to move higher, especially as investor sentiment has cooled” following Nvidia’s GTC conference. “Buy-side expectations for the MI300 are as high as $8B, while Rolland estimates revenue from the MI300 for this year at around $5B.”
• Deutsche Bank “believes the most anticipated aspect of the quarter will be any update to the company's 2024 outlook for MI300 revenues. On this metric, it thinks buy-side expectations have recently fallen on the back of suspected order cancellations by Microsoft, which are yet to be substantiated, but still are likely at a minimum of $4B.”
• TD Cowen is expecting strength in the data center and “increased its MI300 2024 revenue estimate to $4.5B from $4B, saying ramps at several customers will continue to happen more quickly than typical.”
• HSBC says that “market expectations for the company's MI300 2024 and 2025 revenue have been reset,” but thinks AMD “has enough supply capacity and demand to surpass management's artificial intelligence revenue guidance” for the year.
• Baird believes “MI300X orders have been cut by a U.S. hyperscaler recently,” saying that the “magnitude of the initial order suggests it was a multi-year agreement, but it does not know whether the cut is due to market share shift or the hyperscaler scaling down to numbers more in line with shipment expectations.” Baird also “continues to believe there is ‘comfortable upside’ in AMD's artificial intelligence revenue guidance for this year, based on high-bandwidth memory order visibility.”
• Wells Fargo analyst Aaron Rakers says the “focus on AMD is squarely on upside related to its MI300X accelerator chips,” and sees “a path towards AMD generating $8B in revenue from the MI300 (up from $3.5B to $4B), and wonder if there is a recovery in the traditional server market and an ‘underappreciated’ story in market share gain.” Rakers adds that concerns over Microsoft’s order cuts are “too narrowly cited."

While rumors for Microsoft’s order cuts are yet to be substantiated, there are further notes discussed on the site Tom’s Hardware that Microsoft is reportedly able to purchase MI300 GPUs at a 33% discount, at approximately $10,000 per GPU compared to a $15,000 price tag for other customers. This could present a margin headwind in Q1 should Microsoft account for a majority of GPU shipments in the quarter due to the pricing discrepancies.

In the bigger picture, to meet the lowest end of analysts’ GPU revenue estimates of $4 billion and $4.5 billion, AMD would need to boost its forecast by 15% to 30% — the question here is whether management has enough visibility after one quarter to confidently raise its full-year outlook to that extent after raising it by 75%.

Big Tech Capex Commentary

Capex commentary from Big Tech last week was directionally bullish, with Microsoft, Meta and Alphabet expecting to spend upwards of $135 billion this year, predominantly on AI infrastructure. This sets up a positive long-term picture for AMD to increase market share against Nvidia among the major hyperscalers, given AMD’s GPUs are available, can compete on performance, and undercut on price.

Microsoft increased its capex 80% YoY to $14 billion this quarter, and for the entire fiscal year, capex will increase approximately 50% YoY to more than $50 billion. Demand for Azure’s AI services is outpacing its capacity in the near-term, hence the need for Microsoft to accelerate spending to boost GPU supply.

Meta boosted its full year capex range to $35-40 billion, up from $30-37 billion, to build out AI infrastructure and support its internal AI roadmap. However, Meta’s Q1 capex was only $6.7 billion, implying that the bulk of this spend will hit in the second half of the year and accelerate into 2025. By the end of 2024, Meta is aiming to have 350,000 H100 GPUs and 600,000 total GPUs including H100 equivalents, leaving ~250,000 GPUs split between its custom processors and AMD’s MI300. Assuming AMD can capture 50% of that remaining 250K units, MI300 revenue to Meta may surpass $1.8 billion this year.

Alphabet’s capex rose 91% YoY to $12 billion in Q1, primarily for technical infrastructure – this capex spend was led by servers and followed by data centers. Management is expecting quarterly capex “to be roughly at or above the Q1 level,” implying a full-year capex around $50 billion.

Overall, the planned capex outlays and management commentary from the trio is ultimately bullish. Meta’s Q1 spend declined ~$300 million YoY, and was low compared to its full year forecast, and rumors for Microsoft’s order cuts can’t entirely be shrugged off. As such, there is a chance that Q1’s data center and MI300 sales come in light before finding strength in the back half of the year.

AMD Count and Game Plan

By Knox Ridley

This is the bullish count we’re following, and it’s the best interpretation higher given the price information. Here’s what we would guess based on the current technicals: we get a final drop into the $138 region, which will get bought, and then we push higher. If this is going to happen, we must hold $128. Below here and we could see a bigger drawdown take hold, as the below path higher will get invalidated.

Look at the red arrow in the chart above. That’s indicating a 3-wave bounce, which is usually corrective (suggesting lower). This is happening on momentum making a higher high with price making a lower high (common in downtrends). However, I think that what is missing is the final 5th wave lower before reversing. This is a very stretched downtrend pattern, so we should see a reversal soon.

Look for us to buy around $138 on a pullback or around $174 if we get a breakout. If price breaks below $128, we will risk manage the position. (Note: real-time trade alerts and weekly webinars reviewing IOF positions are offered for Advanced Members)

Insiders Activity

Since February of 2024, we’ve seen about $95M of insider sells. Lisa Su was about $50M of this total. Forrest Norrod and Victor Peng also had sales above $10M. The CTO sold about $8M. This is the largest cluster of sales since December of 2021. The synchronicity of four top tier execs, two on the same day, is not my favorite thing to see. We collect these details to help inform our decision if the stock should break a key level.

Conclusion

AMD’s Q1 report is one of the most highly anticipated reports of the quarter as the company is battling both lowered near-term estimates and heightened long-term expectations for GPU revenue. Analyst estimates on MI300s range from $4 billion to $8 billion.

AMD’s fundamentals are expected to improve throughout the year as data center sales accelerate, with operating cash flow growth of nearly 277% on top of an acceleration in EPS growth to the 90% range by the first quarter next year. We’re continuing to track Big Tech’s capex for signs on AI spending, and so far, there’s indications that spend will continue to increase for multiple quarters, leaving time for AMD to gain share.

Analyst estimates are showing a sizable rebound in H2. Let’s see what management says. Stay tuned for our post-earnings report in your inboxes tomorrow night.

Damien Robbins, Equity Analyst for the I/O Fund, contributed to this analysis

Recommended Reading:

• Q2 2024 Earnings Kickoff Webinar Replay
• Super Micro Q3 Pre-Earnings: Puts and Takes for the AI Bullet Train
• Lam Research Fiscal Q3 Earnings: A Tad Early to the 2025 Rebound
• AI's Opportunity: Growth, Investment, and the Future

Micron delivered an exceptional fiscal Q2, with revenue rising nearly 58% as strong AI demand led to pricing power coupled with tight supply dynamics to accelerate its return to profitability this quarter. Q3 was guided 10% above consensus to $6.6 billion at midpoint, representing 76% YoY growth, pointing to an impressive rebound from declining growth just three quarters ago.

Margins were significantly ahead of expectations, driving a strong shift to profitability. Micron was initially expected to return to profitability next quarter, but reported a solid 13.6% GAAP net margin this quarter as operating margin expanded nearly 20 percentage points QoQ.

CEO Sanjay Mehrotra said Micron’s “preeminent product portfolio positions us well to deliver a strong fiscal second half of 2024,” as he believes the company “is one of the biggest beneficiaries in the semiconductor industry of the multi-year opportunity enabled by AI.”

There were many strong, bullish statements on the call: “AI server demand is driving rapid growth in HBM, DDR5 and data center SSDs, which is tightening leading-edge supply availability for DRAM and NAND. This is resulting in a positive ripple effect on pricing across all memory and storage end markets. We expect DRAM and NAND pricing levels to increase further throughout calendar year 2024 and expect record revenue and much improved profitability now in fiscal year 2025.”

For more information regarding the importance of HBM3 and HBM3e in Nvidia and AMD’s 2024 product road map for AI Accelerators, please reference our past analysis noted at the end of this analysis.

Revenue and EPS:

Revenue shows a clear and obvious rebound in the memory market and EPS was a blowout:

• Revenue of $5.82 billion beat estimates by ~9%, and represented YoY growth of 58% and QoQ growth of 23%.
• Fiscal Q3 revenue was guided at $6.6 billion, +/- $200 million, for YoY growth of 76% and QoQ growth of 13%.
• GAAP EPS was $0.71, compared to estimates for ($0.38). This compares to GAAP EPS of ($1.12) in Q1 and ($2.12) in the year ago quarter.
• Adjusted EPS was $0.42, compared to estimates for ($0.24). This compares to adjusted EPS of ($0.95) in Q1 and ($1.91) in the year ago quarter.
• GAAP EPS was guided at $0.17 +/- $0.07, compared to estimates for $0.08.
• Adjusted EPS was guided at $0.45 +/- $0.07, compared to estimates for $0.20.

Margins:

• GAAP gross margin was 18.5%, an expansion of 5120bp YoY from (-32.7%) and 1920bp QoQ from (0.70%). Management had guided for a gross margin of 12%.
• Adjusted gross margin was 20.0%. Gross margins “benefited from $382 million associated with selling the remainder of previously written-down inventories.”
• GAAP operating margin was 3.3%, an expansion of 6570bp YoY from (-62.4%) and 2720bp QoQ from (-23.90%). Management had guided for (-8.2%). Adjusted operating margin was 3.5%.
• GAAP net margin was 13.6%, an expansion of 7620bp YoY from (-62.5%) and 3970bp QoQ from (-26.1%). Adjusted net margin was 8.2%.

•  For Q3, GAAP gross margin was guided at 25.5% +/- 1.5%, an expansion of 700bp QoQ at midpoint. Adjusted gross margin was guided at 26.5% +/- 1.5%. Despite the rather large benefit in Q2 from selling written-down inventories, strong increases in DRAM and NAND pricing are driving this sequential expansion.
• For Q3, GAAP operating margin is implied to be 8.7% at midpoint, an expansion of 540bp QoQ. Adjusted operating margin is implied to be 11.5%, an expansion of 800bp QoQ. Micron is forecasting continued operating income through the rest of FY24.

Management made it crystal clear that HBM3 is accretive to margins. This has been a concern since it’s 3X more expensive to manufacture. The strength in the margin is due to pricing power.

“So with respect to the accretive nature of HBM, look, HBM carries a higher cost, but it also carries a significantly higher pricing because it brings such great value in the applications in terms of its performance and power. And we are executing well. Our yield ramp is going well as well according to plan.”

“And therefore, we are pleased that in this quarter, when we have begun our production shipments, we will be having it accretive to our gross margins in the quarter. And of course, this momentum will continue to build in the quarters ahead.”

“Answer
Mark Murphy (Executives)

Yes. Brian, it's Mark. We won't break it out specifically, but maybe just to give you a sense of the trajectory of gross margins. The increase from first quarter of 1% to 20% in the second quarter was dominantly price. And obviously, a lot of other things going on, but the dominant feature of that increase was price. 

Likewise, in the 20% second quarter actuals to the 26.5% guide, price remains the largest contributor. And offsetting part of that is, of course, what CJ mentioned on the benefit of those lower cost inventories fade away. So — but price is still the largest factor.”

Cash and Debt:

• Cash and short-term investments totaled $9.0 billion.
• Debt totaled $13.7 billion.
• Operating cash flow was $1.22 billion, an increase of 256% YoY but a decrease of (13% QoQ). The sequential decrease may have been impacted by strong pre-payments in the prior quarter from customers aiming to secure supply. Management commented last quarter there were $600 million in prepays but declined to comment on prepays this quarter.
• Adjusted free cash flow was ($29 million), compared to adjusted FCF of ($333 million) in Q1 and ($1.81 billion) in the year ago quarter. Micron is expecting to generate positive adjusted FCF in both Q3 and Q4.

Key Metrics:

• DRAM revenue was $4.2 billion, an increase of 21% QoQ. DRAM pricing increased by the high-teens QoQ. DRAM had increased 24% QoQ in the previous quarter, so this was the second quarter of strong DRAM growth which we covered here.
• NAND revenue was $1.6 billion, an increase of 27% QoQ. NAND pricing increased by more than 30% QoQ, offsetting a low single-digit QoQ decrease in bit shipments.
• Compute and Networking (CNBU) revenue was $2.19 billion, representing an increase of 26% QoQ and 59% YoY. Per mgmt comments: “Data center revenue grew robustly, and cloud more than doubled sequentially.”
• Mobile (MBU) revenue was $1.6 billion, representing an increase of 24% QoQ and 69% YoY. Per management comments: “an expected decline in volume was more than offset by improved pricing” and management confirmed mobile will recover this year: “Smartphone unit volumes in calendar 2024 remain on track to grow low to mid-single digits.”
• Embedded (EBU) revenue was $1.1 billion, representing an increase of 7% QoQ and 28% YoY.
• Storage (SBU) revenue was $905 million, representing an increase of 39% QoQ and 79% YoY. Per management comments: “Data center SSD revenue more than doubled from a year ago driven by share gains from Micron's products.”

Revenue Acceleration Strongly Underway

Fiscal Q2 reaffirmed that Micron’s revenue acceleration is strongly underway, as revenue and Q3’s guide came in well above expectations. Micron added that they are expecting to generate record revenue with “much improved” profitability in fiscal 2025. 

Fiscal Q3’s guidance would mark the highest quarterly revenue in seven quarters, coming in above $6 billion for the first time since the fourth quarter of fiscal 2022. This is driving the fastest acceleration that we have seen for Micron since late 2017.

Q3 is expected to see ~76% YoY revenue growth at midpoint, a 18 percentage point acceleration from Q2 and a 61 percentage point acceleration from when revenue inflected back to positive growth in Q1. However, it’s important to note that these YoY growth rates are viewed against extremely weak comps – the real test for the strength and scale of this acceleration will be fiscal 2025’s growth rates; for example, how close each quarter can stay to the 60% expected revenue growth in fiscal Q1 2025.

An improved pricing environment driven by AI server demand is aiding the revenue growth story. Micron said it was able to drive “robust price increases as the supply-demand balance tightened.”

In particular, AI server demand was seen “driving rapid growth in HBM, DDR5 (D5) and data center SSDs, which is tightening leading-edge supply availability for DRAM and NAND.” Micron said this is causing “a positive ripple effect on pricing across all memory and storage end market.” As a result, Micron is expecting prices to continue to increase through 2024 and into 2025.

Management expressed how unusual the demand for HBM3 is: “And 2024 volume as well as pricing is all locked up. 2025, as I mentioned, the volumes are largely allocated. A vast majority of our production supply is allocated, and some of the pricing is already firmed up. Keep in mind, this has never happened before, right, that we are talking about 2025, and we are sitting in CQ1, and we already have so much discussion around supply and pricing for 2025 getting locked up here as we speak.”

Tight Supply:

Once semiconductor segments are aligned in terms of a rebound, the impact from AI will be more evident. Inventory helps to foreshadow the strength of the rebound.

This is what management stated: “Inventories for memory and storage have improved significantly in the data center, and we continue to expect normalization in the first half of calendar 2024. In PC and smartphone, there were some strategic purchases in calendar Q4 in anticipation of a return to unit growth. Inventories remain near normal levels for auto, industrial and other markets.”

The words “tight supply” were repeated 7 times, which tends to translate to strong pricing power. Here are a few of the comments, which are important to note as the tone of the call was that this pricing power should only increase:

“We anticipate strong HBM demand due to AI, combined with increasing silicon intensity of the HBM road map, to contribute to tight supply conditions for DRAM across all end markets.”

“The trade ratio of 3:1, increasing demand in HBM, increased profitability of HBM is putting a non-HBM part of the memory in tight supply. This is why we say that leading-edge nodes are in very tight supply. And as a result, we would fully expect that D5 as well as other DDR products will improve in their profitability picture as well, given they're very much tight supply there.”

“And so, I mean, this overall tight supply environment bodes well for our ability to manage the pricing increases as well as keep an eye on demand-supply balance and remain extremely disciplined in driving the growth of our business in revenue and profits while continuing to execute our strategy of maintaining stable bit share.”

Note on HBM3e Progress

Micron’s HBM3e was a core part of our multi-faceted AI-driven growth thesis in December, and the company has provided positive updates on HBM3e development and revenue generation.

Management said “we commenced volume production and recognized our first revenue from HBM3E in fiscal Q2 and now have begun high-volume shipments of our HBM3E product.” The company is “on track to generate several hundred million dollars of revenue from HBM in fiscal 2024.”

Micron is expecting these HBM revenues “to be accretive to our DRAM and overall gross margins starting in the fiscal third quarter.” This is an important quote – Micron has already driven tremendous improvement in gross and operating margins in Q2, and this implies that HBM pricing power provided a tailwind to margins. Moving beyond fiscal Q3 and Q4 and into fiscal 2025, margins are expected to continue to expand at a fairly strong rate as HBM revenues ramp significantly.

Micron shed light on customers and capacity, noting that while its HBM3e will be a part of Nvidia’s H200 Tensor Core GPU, it is “making progress on additional platform qualifications with multiple customers.”

Micron’s upcoming 12-high HBM3e has been sampling to customers, and Micron said it will begin ramping the cube in high volume production throughout 2025: “Earlier this month, we sampled our 12-high HBM3E product, which provides 50% increased capacity of DRAM per cube to 36 gigabytes. This increase in capacity allows our customers to pack more memory per GPU, enabling more powerful AI training and inference solutions. We expect 12-high HBM3E will start ramping in high-volume production and increase in mix throughout 2025.”

Nvidia’s H200 win is major win for Micron, as competition in the HBM landscape remains stiff. Per management: “NVIDIA announced its next-generation Blackwell GPU architecture-based AI systems, which provides a 33% increase in HBM3E content, continuing a trend of steadily increasing HBM content per GPU. Micron's industry-leading high-bandwidth memory HBM3E solution provides more than 20x the memory bandwidth compared to standard D5-based DIMM-server module.”

Market leader SK Hynix, who had shipped HBM for Nvidia’s H100, is investing at least $1 billion this year to improve stacking and yields for HBM3/3e, while Nvidia just confirmed that it is qualifying Samsung’s HBM for next-gen GPUs.

While it is not certain that Samsung will pass the qualification stage, it raises questions whether this qualification is for the B200 or another upcoming GPU, or whether Nvidia is seeking to qualify HBM products from all three manufacturers in order to secure ample supply in 2025 and 2026 (given that Micron’s capacity is nearly booked and SK Hynix just commenced HBM3e mass production).

Per Micron, the following sets them apart: “Customers continue to give strong feedback that our HBM3E solution has a 30% lower power consumption compared to competitors' solutions. This benefit is contributing to strong demand.”

Though it is rumored that SK Hynix is shipping to Nvidia’s Blackwell lineup, Micron raises a critical point: the architecture “provides a 33% increase in HBM3E content, continuing a trend of steadily increasing HBM content per GPU.” This trend for higher memory content to support larger and faster GPUs is likely to continue especially as chipmakers such as AMD work quickly to encroach on Nvidia’s share with comparable or faster GPUs.

Additional Growth Opportunities

HBM3e is stealing the spotlight but it’s worth mentioning a few additional growth opportunities for Micron:

• The company is releasing a 128-gigabyte server DRAM module that will provide high bandwidth D5 capability and greater than 20% energy efficiency with 15% better latency compared to Samsung’s 3D TSV solutions. This product has “strong customer pull” with “several hundred million dollars of revenue in the second half of fiscal 2024.”
• Micron reported record revenue share in the data center SSD market last year. In the current quarter, MU grew revenue by 50% QoQ for the 232-layer based 6500 30 terabyte SSDs. These are used for AI data lake applications.
• Edge AI – PCs will be a growth market for Micron. As stated above, mgmt expects PCs to return to growth in CY2024 in the “low single-digit range.” The neural processing units (NPU) chipsets that AI PCs require will see 40% to 80% more DRAM content than non-AI PCs.
• Edge AI – AI phones will require 50% to 100% more DRAM content than non-AI phones.

Conclusion:

A company that is supplying Nvidia (and likely AMD, perhaps Broadcom) on critical memory components for GPUs this year, HBM3E, plus will afford us an early entry for Edge AI with spring-loaded margins and strong pricing power that is expected to increase? Yes, please.

I’m quite positive you will see a new buy alert on Micron tomorrow and we are also looking at entering Lam Research (see below for LRCX analysis).

The report tonight has many implications for a thesis we have been carefully building on a memory rebound, which with some careful risk management, should have a long runway with Edge AI up to bat next (2025).

A special thank you to my team of analysts – Damien, Royston and Knox — who have worked diligently to identify this thesis. Go team go.

Resources:

• 2024 Trend: Memory and PC Rebound
• Memory and PC Stocks Review
• Micron: AI Offers a Multifaceted Secular Growth Tailwind
• Micron Q1: The Memory Rebound has Arrived Fueled by HBM3e – notes on anticipated NVDA partnership
• Micron Q2 Pre-Earnings: Signs of Rebound
• Lam Research: Wafer Fab Equipment Leader & HBM/DRAM Memory
• AMD is Ready to Rival on AI Acceleration – notes on upcoming GPUs with HBM3 memory

Nvidia’s much-anticipated Q4 earnings report saw the AI GPU leader post another large beat and raise as it reported revenue growth of 265% YoY. Nvidia guided fiscal Q1 revenues nearly $2 billion above consensus on top of its almost $2 billion revenue beat in Q4, mirroring what we saw in Q3 as demand for its H100 Hopper GPUs remains elevated.

The consistency and magnitude of the top-line beats is impressive, with Q1’s guide signaling three quarters in a row of revenue growth above 200%. However, it’s the growth further down the income statement where Nvidia’s report truly shines. Nvidia’s stronghold grip on the data center market at the moment combined with pricing power and elevated demand for its H100 GPU has allowed substantial growth in operating income and has generated robust earnings.

Read our pre-earnings write-up here.pre-earnings write-up here.

Revenue and EPS:

• Q4 revenue was $22.1 billion, beating estimates by 7.56%. This represented YoY growth of 265%, a 60 percentage point acceleration from 205% YoY in Q3.
• FY24 revenue was $60.92 billion, an increase of 126% YoY. 
• Q1 revenue was guided at $24 billion, +/- 2%, ahead of estimates for ~$21.9 billion. This represents YoY growth of 235%, or a 30 percentage point deceleration from Q4’s growth rate. We had covered in our pre-earnings write up that revenue growth will peak in Q4 for now at the 265%. This seems to still be the case unless next quarter comes in at $2.2 billion over the current guide. As we have seen these past few quarters, it’s not out of the question that Nvidia beats by this much next quarter. However, it’s looking less likely that Nvidia can sustain this peak growth as we move into the second half of the year.
• Q4 GAAP EPS of $4.93 beat estimates by 16.8%, representing YoY growth of 765%.
• Q4 adjusted EPS of $5.16 beat estimates by 11.2%, representing YoY growth of 486%.

Margins:

Nvidia’s Q4 report highlighted the incredibly strong leverage and margin expansion that the rapid growth in the data center is driving.

• GAAP gross margin was 76% in Q4, and adjusted gross margin was 76.7%, an expansion of 1270 and 1060 bp YoY respectively. 
• GAAP operating margin was 61.6% in Q4, and adjusted operating margin was 66.7%, an expansion of 4080 and 2990 bp YoY respectively.
• GAAP net margin was 55.6% in Q4, and adjusted net margin was 58.1%, an expansion of 3320 and 2220 bp YoY respectively.

Notably, Nvidia is guided “Beyond Q1, for the remainder of the year, we expect gross margins to return to the mid-70s percent range.” It was mentioned on the call that the slightly softer gross might be caused by the higher cost of HBM3.

• For FY24, GAAP gross margin was 72.7% up from 56.9% in FY23. Adjusted gross margin was 73.8% up from 59.2% in FY23.
• For FY24, GAAP operating margin was 54.1% up from 15.7% in FY23. Adjusted operating margin was 60.9% up from 33.5% in FY23.
• For FY24, GAAP net margin was 48.9% up from 16.2% in FY23. Adjusted net margin was 53% up from in FY23. 

Cash Flows:

• Cash on hand was $26.0 billion, an increase from $18.3 billion in Q3 and $13.3 billion in the year ago quarter.
• Operating cash flow was $11.5 billion in Q4, an increase of 411% YoY. FY24 operating cash flow increased 416% YoY to $28.1 billion. For FY24, operating cash flow more than doubled to 46.1%, compared to 20.9% in FY23.
• Free cash flow was $11.2 billion in Q4, an increase of 546% YoY as FCF margin topped 50%. For FY24, free cash flow increased 618% YoY to $26.9 billion. Free cash flow margin more than tripled to 44.2% from 13.9% last year.
• Debt totaled $10.95 billion.

Key Segments:

Data Center:

Data center revenue dazzled again, with Nvidia attributing the growth to “higher shipments of the NVIDIA Hopper GPU computing platform” alongside growth for InfiniBand. Revenues rose 409% YoY and 27% QoQ to $18.4 billion – in other words, a $3.9 billion increase from Q3. Nvidia generated $47.5 billion in data center revenues in FY24, up 217% YoY from $15 billion in FY23.

This is what they mean by “hockey stick” growth:

To put just how rapid this ascent in data center revenues has been, this year’s $47.5 billion in revenue is 18% more than total revenues in the segment for the past five years combined – Nvidia generated $40.2 billion in data center revenue between FY18 through FY23.

According to the CFO commentary on the call for next quarter: “We expect sequential growth in data center and ProViz, partially offset by seasonal decline in Gaming.” As our pre-earnings writeup pointed out, a few analysts were modeling $25 billion data center quarters (for $100 billion per year), so it makes sense that we will see sequential growth in the data center into the foreseeable future.

The CFO also stated that 40% of data center revenue is from inference. This is the first I remember management discussing the percentage that is from inference, and I believe that’s because AMD is pushing hard on the narrative that the MI300s will specifically outperform on inference.

Regarding China, the following was stated: “Growth was strong across all regions except for China, where our Data Center revenue declined significantly following the U.S. government export control regulations imposed in October. Although we have not received licenses from the U.S. government to ship restricted products to China, we have started shipping alternatives that don't require a license for the China market. China represented a mid-single-digit percentage of our Data Center revenue in Q4, and we expect it to stay in a similar range in the first quarter.”

Gaming:

Gaming revenue in Q4 was $2.9 billion, representing a 56% YoY increase against a softer comp and flat growth QoQ. FY24 revenue was $10.4 billion, up 15% YoY.

Pro Viz

Pro Visualization revenue in Q4 was $463 million, up 105% YoY and 11% QoQ. FY24 revenue in the segment was $1.6 billion, up 1% YoY.

Automotive:

Automotive revenue was $281 million in Q4, up 8% QoQ but down 4% YoY. FY24 revenue was $1.1 billion, up 21% YoY as more automakers in China adopt Nvidia’s Drive platform for autonomous driving capabilities.

Additional Notes:

Nvidia’s rapid top-line growth is the primary eye-catching statistic, as no other companies in tech can report such blistering revenue growth at a rate above 200% for multiple quarters at an annualized revenue rate near $90 billion. However, the strengths of Nvidia’s report lie within the operating leverage that this growth is driving.

Operating income in Q4 increased 983% YoY to $13.6 billion, driving a 769% increase in net income to $12.3 billion.

For the full year, operating income of 681% to nearly $33.0 billion, up from $4.2 billion in FY23, while net income rose 581% YoY to $29.8 billion from $4.3 billion in FY23. FY24’s GAAP EPS of $11.93 was nearly 6x higher than FY23’s $1.74.

Cash flow generation surged, with OCF margin more than doubling and FCF margin tripling in FY24. OCF and FCF have increased sequentially each quarter this year, as top-line growth is flowing directly through to the bottom line.

Earnings Call:

There wasn’t much to dissect in the earnings call as what was delivered was another blowout quarter. However, there were some questions on supply that I want to note here. It’s no secret that demand is greater than supply, hence these blowout quarters. It did seem analysts were poking holes at what the timing could be as to when supply won’t be able to continue to afford this extraordinary growth. The answers to the questions were not very informative, rather I’m noting that this seems to the be predominant concern among the analysts even if management chose to remain vague.

Question
Stacy Rasgon (Analysts)

I wanted to — Colette, I wanted to touch on your comments that you expected the next generation of products, so that black well [B100s] to be supply constrained. Can you dig into that a little bit? What is the driver of that? Why does that get constrained as Hopper is easing up? And how long do you expect that to be constrained? Like do you expect the next generation to be constrained like all the way through calendar '25? Like when do those start to ease?

Answer
Jensen Huang (Executives)

Yes. The first thing is overall, our supply is improving. Overall, our supply chain is just doing an incredible job for us. Everything from, of course, the wafers, the packaging, the memories, all of the power regulators to transceivers and networking and cables, and you name it, the list of components that we ship […] The supply chain is really doing fantastic supporting us. And so overall, the supply is improving. We expect the demand will continue to be stronger than our supply provides, and through the year and we'll do our best. The cycle times are improving and we're going to continue to do our best. However, whenever we have new products, as you know, it ramps from 0 to a very large number, and you can't do that overnight. Everything is ramped up. It doesn't step up. And so whenever we have a new generation of products and right now, we are ramping H200s, there's no way we can reasonably keep up on demand in the short term as we ramp […] So we'll — with all new products, demand is greater than supply. And that's just kind of the nature of new products, and we work as fast as we can to catch up with the demand. But overall, net-net, overall, our supply is increasing very nicely.”

Here was another question on supply that was shrugged off, so to speak, yet helps our Members to understand the Q&A had a few analysts focused on figuring out the supply constraints:

Question
Timothy Arcuri (Analysts)

I wanted to ask about how you're converting backlog into revenue. Obviously, lead times for your products have come down quite a bit. Colette, you didn't talk about the inventory purchase commitments, but if I sort of add up your inventory plus the purchase commits and your prepaid supply, sort of the aggregate of your supply, it was actually down a touch. How should we read that? Is that just you saying that you don't need to take as much of a financial commitment to your suppliers because the lead times are lower? Or is that maybe you're reaching some sort of steady state where you're closer to filling your order book and your backlog?

Answer
Colette Kress (Executives)

Yes. So let me highlight on those three different areas of how we look at our suppliers. You're correct. Our inventory on hand, given our allocation that we're on, we're trying to, as things come into inventory, immediately work to ship them to our customers. I think our customer appreciates our ability to meet the schedules that we've looked for.

The second piece of it is our purchase commitments. Our purchase commitments have many different components into it, component that we need for manufacturing but also often we are procuring capacity that we need. The length of that need for capacity or the length of the components are all different. Some of them may be for the next 2 quarters but some of them may be for multiple years. I can say the same regarding our prepaids. Our prepaids are predesigned to make sure that we have the reserve capacity that we need as several of our manufacturing suppliers as we look forward.

So wouldn't read into anything regarding approximately about the same numbers as we are increasing our supply. All of them just have different lengths as we have sometimes had to buy things in long lead times or things that need a capacity to be built for us.”

There was an important question about that pertains to our thesis that Nvidia will become a predominant player for AI software. Right now, software is at a $1 billion run rate. The comment below was the first that I can recall where the CEO was more detailed as to how Nvidia will become a force in AI software. I’m quoting it in full here as a follow up to our deep dive on AI software in July of 2022:

Answer
Jensen Huang (Executives)

Let me take a step back and explain the fundamental reason why NVIDIA will be very successful in software. […] If you don't have software, you can't open new markets. If you don't have software, you can't open and enable new applications. Software is fundamentally necessary for accelerated computing. This is the fundamental difference between accelerated computing and general-purpose computing that most people took a long time to understand. And now people understand that software is really key.

And the way that we work with CSPs, that's really easy. We have large teams that are working with their large teams. However, now that generative AI is enabling every enterprise and every enterprise software company to embrace accelerated computing, and when it is now essential to embrace accelerated computing because it is no longer possible, no longer likely anyhow, to sustain improved throughput through just general-purpose computing, all of these enterprise software companies and enterprise companies don't have large engineering teams to be able to maintain and optimize their software stack to run across all of the world's clouds and private clouds and on-prem.

So we are going to do the management, the optimization, the patching, the tuning, the installed base optimization for all of their software stacks. And we containerize them into our stack called NVIDIA AI Enterprise. And the way we go to market with it is think of that NVIDIA AI Enterprise now as a run time like an operating system. It's an operating system for artificial intelligence. And we charge $4,500 per GPU per year. And my guess is that every enterprise in the world, every software enterprise company that are deploying software in all the clouds and private clouds and on-prem will run on NVIDIA AI Enterprise, especially obviously, for our GPUs. And so this is going to likely be a very significant business over time. We're off to a great start. And Colette mentioned that it's already at $1 billion run rate and we're really just getting started.

Conclusion:

The I/O Fund portfolio is on fire right now. Our audited results from last year will be out soon, and those results will put us in the 90th percentile of all funds in the world for 2023 and also on a 4-year cumulative basis. From there, the first two months of 2024 have been extraordinary as we positioned for Q1 with a high allocation to many year-to-date winners. However, we do not think it will always remain this way – tech cannot remain in favor forever.

As you know from our pre-earnings writeup, we think Nvidia’s hitting peak growth is “tricky” for investors while acknowledging the valuation is eerily low still — it is very unusual for a stock to be up 250% in a year and yet be cheaper than it was at its bottom (Oct 2022 for Nvidia). The valuation is what makes it a buy on any dips. However, we also won’t be shy about taking gains if we reach predefined price targets. We have one in mind for Nvidia, let’s see if we get there for our next trim.

Too many investors ride high on paper gains, and subsequently lose those gains. We don’t want to choose between holding a high conviction stock and making money. Instead, we want it all – put some money in the bank, lock-in gains, yet hold the stock for the long haul at a high allocation and hedge if tech falls out of favor.

We will keep doing our very best to bring you quality winners alongside risk management with the ultimate goal of answering the million-dollar or billion-dollar question, which is how to safely participate in the life changing gains tech has to offer. We do not believe this question has been satisfactorily answered. Which is why if you see us hit our price target on Nvidia … and trim our high conviction stock … but buy aggressively on dips — then, you’ll know we are working hard to answer this question for our Members.

Recommended Reading:

• Nvidia Earnings Preview: 239% is the Revenue Growth Peak (for now)
• Crypto and AI Opportunity: Real Vision Video Interview
• AMD Q4 2023 Earnings: 75% GPU Raise, Separating the Wheat from the Chaff
• Super Micro Q2 2024 Earnings: The AI Bullet Train