Nvidia is the global AI compute leader and there is not a distant second. Therefore, reporting (1%) for Q2’s compute segment should have tanked the stock. Instead, we are seeing a mild reaction because Q3 was quite strong and spells good things to come for Nvidia in the second half of the year.
Given we are on the cusp of Blackwell starting to drive the stock’s narrative (c’mon already), it’s a good time to pause and talk about where Nvidia could go from here. I revisited the medium-term and longer-term forecast on a few pre-earnings discussions and was pleasantly surprised that Jensen Huang did the same on the earnings call.
Also important for I/O Fund members who hold AI networking positions, the report was much stronger than expected with networking up a whopping 46% QoQ and 78% YoY to $7.25 billion. This foreshadows what’s to come in the AI networking space; and should be a boon for stocks like Astera Labs and Credo which supply Nvidia.
Despite Q3 putting Nvidia right on track, I would not be surprised if the market offers soft price action on the stock in the near-term. The China narrative is confusing (and beaten to death), Q2 should mark a bottom and semiconductors don’t do well at their cyclical bottoms, plus the AI market is running on fumes until Nvidia can carry the market (Q3-Q4).
First, we will run through the financials, and then in the second section I’ll break down what I think you can expect on this stock for the remaining part of this year, into calendar year 2026 and by the end of the decade.
Slight Revenue Beat in Q2, Solid Q3 Guide
Nvidia reported $46.74 billion in revenue in Q2, slightly ahead of estimates for $46.13 billion. This corresponded to growth of 55.6% YoY, decelerating more than 13 points from 69.2% in Q1; on a QoQ basis, revenue increased just 6.1%, slowing from 12% in Q1.
This is expected to be a temporary lull in sequential growth, as Nvidia guided for $54 billion in revenue, +/- 2%, for Q3, corresponding to 53.8% YoY growth and a rebound to 15.5% QoQ growth. This was ahead of estimates for $52.7 billion, even with Nvidia noting that this guidance assumed no H20 sales in the upcoming quarter.
For FY26, Nvidia is currently expected to report 55.4% revenue growth to $202.8 billion, though it is likely that this figure is revised higher, towards $205 billion, accounting for the slight beat in Q2 and Q3’s guide.
Key Segments: First Sequential Decline in Compute Revenue, No China Sales
Nvidia’s data center revenue increased 56% YoY and 5% QoQ to $41.1 billion, a marginal miss versus estimates for ~$41.2 billion in the quarter. This is also the smallest sequential increase since Hopper’s breakout quarter at just ~$2 billion.
Notably, Q2’s data center result would mark the segment’s first miss since Q4 FY23 as Q1 was essentially in-line.
While Compute revenue was up 50% YoY, the segment reported an unusual (1%) QoQ decline to $33.8 billion, a sharp shift from double-digit growth in late FY25 and 5% in Q1. Nvidia said the sequential decline was driven by a $4 billion QoQ reduction in H20 sales, as it reported no H20 sales to China this quarter and $0.65 billion in H20 sales to an unrestricted customer outside of China.
On a more positive note, Nvidia said that it is continuing to ramp Blackwell and now Blackwell Ultra GPUs, with growth of 17% QoQ. Based on comments from Q1 placing Blackwell revenue in the $23.5-24B range, this would place Blackwell revenue approaching $28 billion.
Networking growth was also robust in Q2, helping offset the softness in Compute as revenue rose 46% QoQ and 78% YoY to $7.25B. This marked a 22 point acceleration from 56% YoY growth in Q1. Nvidia said the strong performance in Networking was driven by “growth of NVLink compute fabric for GB200 and GB300 systems, the ramp of XDR InfiniBand products, and adoption of Ethernet for AI solutions.”
- Gaming revenue was $4.29 billion, rising 14% QoQ and accelerating 7 points to 49% YoY. Nvidia said this was driven by sales of Blackwell-based GPUs.
- Automotive revenue rose 69% YoY and 3% QoQ to $586 million.
- Pro Viz revenue rose 32% YoY and 18% QoQ to $601 million.
- OEM and other revenue rose 97% YoY and 56% QoQ to $173 million.
Margins Outperform in Q2 with Strong Expansion
Despite the softness in data center compute, Nvidia outperformed on margins, beating its guidance across the board and delivering a notable uplift in operating margin. Nvidia guided for more sequential improvement in operating margin in Q3, to the highest level since Q1 FY25.
- GAAP gross margin was 72.4% in Q2, slightly ahead of guidance for 71.8%. Adjusted gross margin was 72.7%, or 72.3% excluding $180M in H20 inventory releases, ahead of guidance for 72%.
- For Q3, Nvidia guided for GAAP gross margin to improve nearly 1 point to 73.3%, +/- 0.5%, and adjusted gross margin to improve to 73.5%, +/- 0.5%.
- GAAP operating margin was 60.8% in Q2, improving nearly 12 points QoQ and coming in 1.7 points ahead of guidance for 59.1%. Adjusted operating margin was 64.5%, also up nearly 12 points QoQ and 1.4 points ahead of guidance for 63.1%.
- For Q3, Nvidia guided for GAAP operating margin to expand 1.6 point QoQ to 62.4%, signaling continuing operating leverage tailwinds in the quarter. Adjusted operating margin was guided at 65.7%, up 1.2 points QoQ.
- GAAP net margin was 56.6%, up 14 points QoQ and more than 6 points ahead of guidance for 50.2%. Adjusted net margin was 55.2%, up 10 points QoQ.
- For Q3, Nvidia’s guidance implies a GAAP net margin of 52.9%, moderating 3.7 points QoQ.
Slight Adjusted EPS Beat in Q2
Nvidia reported just a 3.9% adjusted EPS beat in Q2, reporting $1.05 in earnings versus estimates for $1.01. This corresponds to growth of 54.4% YoY, rebounding substantially from Q1’s H20-affected 32.8% growth. Adjusted ESP growth is expected to remain strong through the rest of the fiscal year, at 47.2% and 53.2% in Q3 and Q4.
For FY26, analysts are currently expecting Nvidia to earn $4.36 in adjusted EPS, up 45.8% YoY. This has improved $0.20 since May’s estimate of $4.16, or 39.2% YoY growth. While still early, Nvidia’s earnings growth is expected to remain quite strong in FY27 at 38% YoY to $6.02.
Cash Flow Margins Drop to Lowest Level since Q4 FY23
Nvidia’s cash flow margins dropped to the lowest level since Q4 FY23, as sharp QoQ growth in accounts receivable and inventories weighed on both operating and free cash flow.
- Operating cash flow was $15.37 billion, down from $27.41 billion in Q1. OCF margin was 32.8%, down nearly 30 points QoQ and more than 15 points lower YoY.
- Free cash flow was $13.45 billion, down from $26.14 billion in Q1. FCF margin was 28.8%, again down more than 30 points QoQ and more than 16 points lower YoY.
- Accounts receivable surged $5.7 billion QoQ, or nearly 26%, to $27.8 billion. Nvidia said this was driven by timing of cash collections and Blackwell Ultra ramping late in the quarter.
- Inventories rose more than $3.6 billion QoQ, or 32%, to $14.96 billion, to support Blackwell Ultra’s ramp.
- Cash and equivalents totaled $56.8 billion, while debt remained steady at $8.47 billion.
$200 Billion Run Rate (and why it matters)
The Q3 guide was key as if we assume a similar mix of data center to other segments, Nvidia’s Q3 guidance puts it on the brink of achieving a $200B data center run rate with Q3 data center at $48.5B DC at midpoint.
We had outlined more than a year ago in our free analysis that analyst estimates were too low, Nvidia Q1 Earnings Preview: Blackwell And The $200B Data Center stating: “These are the current estimates, yet if the analysts are correct, then the far right of the graph will end in $50B quarterly revenue. The difference between the current consensus and this much higher trajectory can be summarized in one word: Blackwell.” At the time, consensus was that we end the year with $33B in revenue.
In fact, if we grow 10% QoQ in Q4 (it’s likely to be higher), then the data center will have quarterly revenue of $53.4 billion or about 62% higher than where analyst estimates were in May of 2024. This is a substantial disconnect for the world’s most valuable company (and a profitable one for investors who track this).
The reason this matters is that if we nail the $50B quarter (very likely we do at $53.4B) then we are on track to hit a $75 billion data center quarter by the end of fiscal 2027. This assumes a 50% CAGR over 6 quarters and 10%-11% QOQ data center growth. Keep in mind, capex just grew 23% QoQ – which helps illustrate why 10% to 11% QoQ should be doable.
If we do reach a $75 billion data center quarter compared to the $47 billion quarter that was recently reported, then that’s tracking about 60% growth in the AI segment. When you layer-in that we like to buy low when we can, and the market likes to sell semiconductors stocks at cyclical lows (when it’s really the best time to buy) we may see even more upside.
Note: I elaborated on this pre-earnings in various interviews this week including Bloomberg, Schwab and Fox Business. Bloomberg, Schwab and Fox Business.
$6 Trillion Market Cap Prediction – Revisited
Nvidia has a 30 forward PS 3-year median but even if we go with something more reasonable like a 20 forward PS, then the market cap would be $6 trillion for $300 billion data center revenue.
From my perspective, it’d be better to get the stock lower if there is high confidence in reaching this market cap.
$10 Trillion Market Cap – Revisited
To get to a $10 trillion market cap at a 20 forward PS, Nvidia’s revenue would have to be $500 billion. Analysts have the company reaching this a little after the year 2033 whereas I believe it will reach $10 trillion before 2030.
The reality is that with Nvidia’s trajectory, we could see this market cap as early as 2028. That’s because we have not factored in software to the equation, which is expected to be an equal size market as hardware. The timing on software is tricky, and thus the I/O Fund has primarily preferred lower risk, yet also hypergrowth AI hardware stocks.
Nvidia essentially has to grow the data center at a 30% CAGR for two years from 2026-2028 to reach my prediction of $10 trillion market cap two years early.
That would mean analyst estimates are five years off as analysts have Nvidia reaching this revenue into the first quarter of 2034.
Note that I'm not stuck on this happening – we will shift if the market shifts. However, it leads back to why Q3 puts us right on track as the $50B data center quarter is achievable this year (the first milestone for these predictions).
What Jensen Huang said in the Earnings Call
I’d call this the total addressable market (TAM) earnings call as that was the predominant theme.
In the opening remarks, JH stated: “We see $3 trillion to $4 trillion in AI infrastructure spend in the — by the end of the decade.” This was followed up by commentary that he sizes the market at $600 billion today. The math there is 5X growth.
These are strong comments and thus it was followed up on during the Q&A where JH expanded his comments to say:
“And so over the next couple of years, you're going to — well, you asked about longer term. Over the next 5 years, we're going to scale into with Blackwell, with Rubin and follow-ons to scale into effectively a $3 trillion to $4 trillion AI infrastructure opportunity. The last couple of years, you have seen that CapEx has grown in just the top 4 CSPs by — has doubled and grown to about $600 billion. So we're in the beginning of this build-out, and the AI technology advances has really enabled AI to be able to adopt and solve problems to many different industries.”
This was followed up on by an analyst who asked for more specifics about what Nvidia’s share of that would be. Frankly, my guess would have been about 30% to 50% for Nvidia but management stated it would be closer to high 50% to 70%.
“Benjamin Alexander Reitzes
Jensen, I wanted to ask you about your $3 trillion to $4 trillion in data center infrastructure spend by the end of the decade. Previously, you talked about something in the $1 billion range, which I believe was just for compute by 2028. If you take past comments, $3 trillion to $4 trillion would imply maybe $2 billion plus in compute spend. And just wanted to know if that was right and that's what you're seeing by the end of the decade. And wondering what you think your share will be of that. Your share right now of total
infrastructure compute-wise is very high, so I wanted to see. And also if there's any bottlenecks you're concerned about like power to get to the $3 trillion to $4 trillion.
Jen-Hsun Huang
[…] United States represents about 60% of the world's compute. And over time, you would think that artificial intelligence would reflect GDP scale and growth and so — and would be, of course, accelerating GDP growth.
And so our contribution to that is a large part of the AI infrastructure. Out of a gigawatt AI factory, which can go anywhere from $50 billion to plus or minus 10%, let's say, $50 billion to $60 billion, we represent about $35 billion plus or minus of that and $35 billion out of $50 billion per gigawatt data center.”
My thoughts are that this could be “CEO math” which is often too high. For example, JH had said 1K racks were shipping per week yet the reality is this print didn’t confirm that statement was true. The CFO did state that in this earnings call, and naturally, CFOs tend to be more accurate so let’s see if the 1K rack statement comes to fruition in the next earnings report.
Regarding what was stated above, if the current number is $600B for AI infrastructure and we have close to a $200B run rate, then 33% is a good solid number to work with, especially given energy will take front and center on costs as we go along. At $3-$4 trillion, JH is basically stating Nvidia could have a $1 trillion data center segment if Nvidia sees 33% of that by the end of this decade. That would be up from a $184B data center segment today and up from a $200B data center segment in Q3.
Additional Commentary:
Management drew attention to the magnitude that demand outstrips supply: “Right now, the buzz is — I'm sure all of you know about the buzz out there. The buzz is everything sold out. H100 sold out. H200s are sold out. Large CSPs are coming out renting capacity from other CSPs. And so the AI-native start-ups are really scrambling to get capacity so that they could train their reasoning models. And so the demand is really, really high”
In terms of why Rubin can stand its own against previous generations of GPUs and continue to drive forth demand, management expanded on the advent of reasoning agentic AI:
“At the highest level of growth drivers would be the evolution, the introduction, if you will, of reasoning agentic AI. Where chatbots used to be one shot, you give it a prompt and it would generate the answer, now the AI does research. It thinks and does a plan, and it might use tools. And so it's called long thinking; and the longer it thinks, oftentimes, it produces better answers.
And the amount of computation necessary for 1 shot versus reasoning agentic AI models could be 100x, 1,000x and potentially even more as the amount of research and basically reading and comprehension that it goes off to do.”
Lastly, keep an eye on networking as it is a subtle clue that the NVL72s are starting to move the needle – above all else, we want to see these larger systems shipping as these massive SKUs are already at $28B this quarter and have a long runway to go. As you know, we are positioned to capture the AI networking tailwinds from these larger systems:
“Networking delivered record revenue of $7.3 billion, and escalating demands of AI compute clusters necessitate high efficiency and low latency networking. This represents a 46% sequential and 98% year-on-year increase with strong demand across Spectrum- X Ethernet, InfiniBand and NVLink. Our Spectrum-X enhanced Ethernet solutions provide the highest throughput and lowest latency network for Ethernet AI workloads. Spectrum-X Ethernet delivered double-digit sequential and year-over-year growth with annualized revenue exceeding $10 billion. At Hot Chips, we introduced Spectrum-XGS Ethernet, a technology design to unify disparate data centers into giga-scale AI super factories. [ CoreWeave ] is an initial adopter of the solution, which is projected to double GPU-to-GPU communication speed.”
Conclusion:
I wouldn’t be surprised if the market continues to sell this report, as the market likes to do at cyclical lows for semiconductor companies. What’s quite impressive is that Nvidia has continued to beat even at its cyclical low (defined as being in-between GPU generations) and despite the curveball with China’s H20 revenue.
Our fundamentals process favors QoQ acceleration and the 15% QoQ growth for Q3 helps put our estimates right on track. The strong networking growth spells good things to come as it indicates it’s the larger systems that are propping up the growth (Blackwell was stated to have grown 17% QoQ).
Now is the perfect time to pause and look a the medium-term and longer-term picture to answer for ourselves if the world’s most valuable company can continue to grow – or are we in a bubble? Each Member will need to decide that for themselves although my answer is “yes, for many AI stocks we are in a bubble – but for Nvidia, we are not.” Hopefully, this post-earnings writeup helps to substantiate why I continue to believe there is room in this stock.
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: