🤖 NVIDIA: Full Throttle
🤖 NVIDIA: Full Throttle
Demand for Blackwell is "incredible" but questions remain
Welcome to the Free edition of How They Make Money.
Over 140,000 subscribers turn to us for business and investment insights.
Heads up! Stay tuned for our PRO coverage on Saturday and the Premium August report with 130+ visuals on Sunday. Monday is Labor Day in the US, so there will be no Premium article on Tuesday. We’ll be back with our usual schedule next Friday.
In case you missed it:
NVIDIA’s earnings have become Wall Street’s Super Bowl, capturing the Zeitgeist with watch parties and heavy interest from investors and traders. NVDA is the most traded stock—a title previously owned by TSLA.
The stock was down after the latest earnings report on Wednesday. I’m here to remind you that NVDA tripled in 2023 and doubled in 2024 (so far). Stocks don’t go up in a straight line, even less so when the valuation assumes stellar execution before earnings.
Shipments of Blackwell, the next-generation platform, slipped by a quarter, shifting revenue from Q3 to Q4 (and beyond). That’s merely a timing issue.
But for the long-term? Jensen Huang had the same message:
“Global data centers are in full throttle to modernize the entire computing stack with accelerated computing and generative AI.”
In our review of Meta’s earnings, we discussed how Big Tech plans to spend more on AI infrastructure in 2025. The $1 trillion question is: How long will it last?
To answer this, Huang shared five critical points about the opportunity ahead.
Let’s unpack it all.
Today at a glance:
NVIDIA’s Q2 FY25.
Five critical points.
Key quotes from the call.
What to watch looking forward.
1. NVIDIA Q2 FY25
NVIDIA’s fiscal year ends in January, so it was Q2 FY25. I’m focusing on sequential growth (quarter-over-quarter), a better representation of the momentum.
Income statement:
Revenue jumped +15% Q/Q to $30.0 billion ($1.3 billion beat).
⚙️ Data Center grew +16% Q/Q to $26.3 billion.
🎮 Gaming grew +9% Q/Q to $2.9 billion.
👁️ Professional Visualization grew +6% Q/Q to $0.5 billion.
🚘 Automotive grew +5% Q/Q to $0.3 billion.
🏭 OEM & Other grew +13% Q/Q to $0.1 billion.
Gross margin was 75% (-3pp Q/Q), in line with guidance.
Operating margin was 62% (-3pp Q/Q), 1 point ahead of guidance.
Non-GAAP operating margin was 66% (-3pp Q/Q).
Non-GAAP EPS $0.68 ($0.04 beat).
Cash flow:
Operating cash flow was $14.5 billion (48% margin).
Free cash flow was $13.5 billion (45% margin).
Balance sheet:
Cash and cash equivalents: $34.8 billion.
Debt: $8.5 billion.
Q3 FY25 Guidance:
Revenue +8% Q/Q to $32.5 billion ($0.8 billion beat).
Gross margin 74.7% (-0.5pp Q/Q).
So what to make of all this?
NVIDIA delivered a revenue beat of 5%. If you recall, the revenue beat was 13% in Q3, 8% in Q4, and 6% in Q1. So, the beats are getting smaller as the market sets expectations higher and the growth normalizes. The Q2 FY25 revenue guidance was $0.8 billion ahead of consensus. That compares to a $1.2 billion beat in the previous quarter. Despite resetting expectations higher every quarter, NVIDIA still finds a way to surprise. But the extent of the overperformance has waned.
⚙️ Data Center was 88% of overall revenue (+1pp Q/Q). It grew 154% year-over-year and 16% sequentially. As part of Data Center:
⚡ Compute: Up 162% year-over-year and 17% sequentially to $22.6 billion. Like previous quarters, the driver was strong demand for the Hopper GPU computing platform used for training and inferencing of large language models (LLMs), recommendation engines, and Gen AI apps.
🔌 Networking: Up 114% year-over-year and 16% sequentially to $3.7 billion. The Spectrum-X Ethernet networking solution optimized for AI doubled sequentially, and Infinitiband remained a crucial driver.
🎮 Gaming grew 9% sequentially, partially boosted by back-to-school season. GeForce RTX PCs have an installed base of over 100 million.
👁️ Professional Visualization grew 6% sequentially, driven by AI and graphic use cases, including model fine-tuning and Omniverse-related workloads.
🚘 Automotive grew +5% sequentially, driven by the ramp of AI cockpit solutions and self-driving platforms.
Gross margins compressed sequentially, as expected. Management expects a gross margin in the mid-70s for FY25 (unchanged), implying a contraction in the second half, explained by a higher mix of new Data Center products.
Operating margin worsened in the FY25 outlook, with operating expenses now expected to grow in the mid-to-high 40% compared to low 40% previously, primarily due to investment in the next generation of products.
Management announced a $50 billion stock buyback program. That’s not necessarily noteworthy, considering that NVIDIA is now a $ 3 trillion company generating nearly $15 billion in free cash flow per quarter.
Guidance was ahead of expectations ($0.8 billion beat or less than 3%), a relatively modest beat compared to the past few quarters. Why? Several billion dollars of Blackwell revenue slipped from Q3 to Q4.
2. Five critical points
Jensen Huang gives the elevator pitch
Huang wrapped the earnings call with his five top points about NVIDIA:
Accelerated Computing Tipping Point: The demand for accelerated computing is surging due to slowing CPU scaling. NVIDIA is expanding its CUDA-X libraries to open new markets and support diverse applications like data science, 5G, and genomics.
Blackwell AI Infrastructure Platform: Blackwell represents a significant leap forward in AI infrastructure, integrating a powerful GPU with a full-stack system designed for building AI factories. It delivers 3-5x more AI throughput than its predecessor, Hopper.
NVLink Game-Changer: NVLink's all-to-all GPU switch enables unprecedented connectivity, with up to 144 GPUs in a single rack. This accelerates the training of large models and enhances inference capabilities.
Generative AI Momentum: Gen AI is rapidly evolving, with developers pushing model capabilities and applications expanding beyond chatbots and image generators. Enterprises and governments are investing heavily in AI infrastructure.
Enterprise AI Wave: The demand for enterprise AI solutions is rising. NVIDIA AI Enterprise platform provides tools for customizing and deploying AI models, offering substantial value for businesses.
Blackwell is coming
Blackwell, NVIDIA’s next-generation platform, has been delayed by a quarter, pushing some revenue out of Q3 and impacting the headline guidance.
At this stage, NVIDIA is shipping functional samples of Blackwell. Shipments were supposed to ramp in Q3 (initially a quarter earlier than expected). However, a delay pushed the plan to Q4. Colette Kress explained:
“We executed a change to the Blackwell GPU mask to improve production yield. Blackwell production ramp is scheduled to begin in the fourth quarter and continue into fiscal 2026. In the fourth quarter, we expect to ship several billion dollars in Blackwell revenue. Hopper demand is strong, and shipments are expected to increase in the second half of fiscal 2025.”
So Blackwell would be additive to continued growth from Hopper, NVIDIA’s current generation. While the delay has a short-term impact, Huang noted that “the demand for Blackwell is incredible.” In short, this is a supply-constrained product, with demand expected to outpace supply in FY26.
3. Key quotes from the earnings call
CFO Colette Kress:
On Data Center:
“Customers continue to accelerate their Hopper architecture purchases, while gearing up to adopt Blackwell. Key workloads driving our data center growth include generative AI, model training, and inferencing. Video, image, and text data pre and post-processing with CUDA and AI workloads, synthetic data generation, AI-powered recommender systems, SQL, and vector database processing as well.”
She provided updates on the three major customer categories:
☁️ Cloud service providers (CSPs) contributed 45% of data center revenue. All hyperscalers (Amazon, Microsoft, Google) are NVIDIA customers.
💬 Consumer Internet companies are also a critical vertical. Meta’s Llama 3 (which powers Meta AI) was trained on a cluster of 24,000 H100 GPUs.
🗄️ Enterprise includes thousands of companies building gen AI apps for consumers, advertising, education, enterprise, healthcare, and robotics.
NVIDIA’s four largest customers represented 46% of Q2 FY25 revenue.
Inference workloads contributed over 40% of NVIDIA's Data Center revenue in the past year. As more generative AI applications make their way into consumer products, inference is poised to become a massive market, offering customers a significant return on their data center investments.
As a reminder, AI systems operate through two core stages:
🎓 Training: AI learns from vast amounts of data, developing intelligence and pattern recognition. NVIDIA's powerful GPUs dominate this phase.
🧠 Inference: AI applies its knowledge to real-world tasks and decision-making. While facing stiffer competition here, NVIDIA is making significant progress.
On Sovereign AI:
“Our sovereign AI opportunities continue to expand as countries recognize AI expertise and infrastructure at national imperatives for their society and industries. […] We believe sovereign AI revenue will reach low-double-digit billions this year.”
Sovereign AI refers to a nation's ability to produce artificial intelligence using its infrastructure, data, workforce, and business networks. Management often calls out Japan. This soundbite implying $10+ billion in revenue from Sovereign AI is critical. That’s nearly 10% of FY25 data center revenue.
On the impact of export restrictions:
“Our data center revenue in China grew sequentially in Q2 and is a significant contributor to our data center revenue. As a percentage of total data center revenue, it remains below levels seen prior to the imposition of export controls. We continue to expect the China market to be very competitive going-forward.”
The US regulations affect the highest performance levels.
On the products driving demand:
“A range of applications are fueling our growth, including AI-powered chatbots, generative AI copilots, and agents to build new monetizable business applications and enhance employee productivity.”
Nay-sayers often mention a lack of “revenue-bearing AI” for NVIDIA’s customers (and their customers). But that’s not true anymore:
Hyperscalers are already seeing billions in AI revenue.
Enterprise software giants offer (and charge for) co-pilot solutions.
Meta sees improvement in recommendations and ad performance.
Chatbots like Perplexity are just getting started on ad monetization.
On software:
“We expect our software, SaaS and support revenue to approach a $2 billion annual run rate exiting this year, with NVIDIA AI Enterprise notably contributing to growth.”
Software reached a $1 billion annualized run rate in FY24. So, the implication is a 100% growth in FY25. While this revenue remains small today, it could creep up fast. Huang pointed to the significant addressable market for NVIDIA AI Enterprise as the CUDA-compatible GPU installed-based grows from millions to tens of millions. Software is the hidden bullcase underneath the headline story focused on Hopper and Blackwell.
CEO Jensen Huang:
Huang has described a ‘generative AI wave’ worth repeating here. AI is moving from one category to the next:
→ Startups and CSPs.
→ Consumer Internet.
→ Software platforms.
→ Enterprise and government.
“Everyone is getting an LLM” is my way of summarizing it.
On the two gen-AI dynamics:
“The first one is, there are larger frontier models trained on more modalities and surprisingly, there are more frontier model makers than last year. And so you have more on more on more. That's one of the dynamics going on in generative AI. The second is below the tip of the iceberg. What we see are ChatGPT, image generators, we see coding. […] What's below the iceberg are the largest computing systems in the world today, which are are recommender systems moving from CPUs to generative AI.”
Huang named examples like recommendation algorithms, ad customization, search, and user-generated content. These applications are still in their infancy but will likely offer critical use cases for gen AI. In turn, new AI startups create demand for cloud services, which will be a boon for CSPs—NVIDIA’s largest customers.
On explaining the continued demand for Hooper:
“The first reason is, if you just look at the world's cloud service providers and the amount of GPU capacity they have available, it's basically none. […]
The second reason for Hopper demand right now is because of the race to the next plateau. The first person to the next plateau gets to introduce a revolutionary level of AI. The second person who gets there is incrementally better or about the same. The ability to systematically and consistently race to the next plateau and be the first one there is how you establish leadership.”
Huang is doubling down on the idea of AI leadership. In his view, not only do CSPs need to have the capacity for AI workloads, but they also need to be first to offer state-of-the-art capabilities to demonstrate their leadership. So even though Blackwell is around the corner, being the “world’s best” in the current environment remains paramount.
4. What to watch looking forward
Buybacks continue
NVIDIA maintained an aggressive buyback pace in Q2:
$9.5 billion in FY24.
$7.7 billion in Q1 FY25.
$7.2 billion in Q2 FY25.
While this could signal management's belief in the company's prospects, some investors may question if the current cash influx would be better spent on research and development, M&A, or other growth initiatives.
Is it too late to buy?
NVDA was notably absent from high-profile hedge funds’ top picks in the latest 13F filings, illustrating that hedge funds have become more cautious in 2024.
According to data from YCharts:
NVIDIA's forward PE ratio is 41.
Apple’s forward PE is 34.
Microsoft’s forward PE 31.
Based on the expected earnings growth in the coming years, NVIDIA appears cheaper than the rest of big tech. With nearly $17 billion in net profit in Q2, NVIDIA’s valuation is based on the fundamentals of the underlying business. This isn’t Cisco during the dot-com bubble. The profits are very real and cashed in every quarter.
Here’s the catch: NVIDIA’s forward earnings include the expected surging demand for Blackwell. But what comes after? The semiconductor industry is known for its cyclicality. How will the competitive landscape evolve over a multi-year time horizon?
Jensen Huang has made his pitch very clear:
The shift to accelerated computing, combined with gen AI tailwinds for consumers and businesses, will lead to surging demand well beyond 2025 or 2026. In addition, with a gigantic install base in the making, CUDA could be the next big growth avenue—with repeatable revenue and software margins to boot.
That said, as the market shifts, many questions remain:
Could CSPs look for internal solutions to save money?
How long can NVIDIA keep its market share in the GPU market?
What happens once AMD catches up with differentiated offerings?
Answering these unknowns is impossible today.
But that’s the beauty of investing.
That’s it for today!
Stay healthy and invest on!
Get Your Business a Custom Visual
Interested in custom visuals for your organization or brand? Complete the form here, and we'll get in touch.
Disclosure: I am long AAPL, AMD, AMZN, GOOG, and NVDA in App Economy Portfolio. I share my ratings (BUY, SELL, or HOLD) with App Economy Portfolio members.
Author's Note (Bertrand here 👋🏼): The views and opinions expressed in this newsletter are solely my own and should not be considered financial advice or any other organization's views.
NVIDIA spends $50bn on buy-backs when it has a total cash balance of $35bn. Most of its earnings are being allocated to buy-backs. Buy-backs are occurring at valuations (25x revenue) which many would argue are well above intrinsic value based on unit economics, so buy-backs will be destroying shareholders equity. To make matters more curious, Jensen is selling about $20m of his own holding on a daily basis and then buying it back with shareholder capital. Try to square that circle. He is effectively propping up the price with shareholder equity while he cashes in. It is like compelling all other shareholders to buy his stake at inflated prices. The guy is a genius, no doubt, but ethics need to be called into question here. And if he is selling down his own stake at these levels, do you think that he knows it is over valued?
It kinda looks a lot like Cisco and Microsoft in the year 2000. Check out what happened to both in the decade or more that followed!
NVIDIA is a great company, no question, but even the best company at the wrong price makes for a poor investment. Cisco and Microsoft are evidence of that.
Added to which, the semi-conductor business is cyclical and we are peak cycle right now. Once all the AI infrastructure spending is done, including phenomenally expensive NVIDIA GPUs, what then?
When investing, shareholder gains are brought about by one of two things. One is value created by the business, which is arguably a small amount of the Market Cap, and the remainder is brought about by irrational speculation. Those that gain from the latter, do so at the expense of those who make big losses on the way back down when the correction comes. There is no such thing as free money, one man's gain will be another's loss.
Caveat Emptor