NVDA Q2 Earnings | NVIDIA Keeps Breaking the Rules of ScaleNVIDIA CorporationBATS:NVDAmoonyptoNVIDIA just crossed $96 billion in quarterly revenue, with growth accelerating to more than 100%. Management now expects revenue to grow another 70% next year, which would add more than $200 billion in annual revenue. At this size, those numbers sound almost ridiculous. Yet NVIDIA says demand is still being limited by supply, not a lack of customers At the same time, the story is becoming more complicated. NVIDIA is helping finance the infrastructure its customers need, while some of those same customers are investing heavily in their own competing chips That leaves investors with two big questions: How long can the AI infrastructure boom continue to compound, and how much of that spending can NVIDIA keep capturing? 1. NVIDIA Q2 FY27 NVIDIA’s fiscal year ends in January, which means the July quarter was Q2 FY27 The Data Center business continues to operate at an extraordinary level Income Statement Revenue jumped 106% year over year to $96.2 billion, beating expectations by $4.1 billion. Data Center revenue rose 117% to $89.0 billion, while Edge Computing increased 27% to $7.2 billion Gross margin came in at 75%, up 3 percentage points year over year. Operating margin reached 66%, up 5 percentage points. NonGAAP EPS was $2.22, beating estimates by $0.13 Cash Flow Operating cash flow increased 57% year over year to $24.1 billion, while free cash flow climbed 59% to $21.3 billion Balance Sheet NVIDIA ended the quarter with $99.4 billion in cash and marketable securities against $33.4 billion of debt Q3 FY27 Guidance Management expects revenue of $108 billion, representing 12% sequential growth and 89% year over year growth. That was $3.4 billion above expectations Gross margin is expected to come in at 74%, down one percentage point from Q2. The guidance also assumes zero Data Center compute revenue from China What Does It All Mean MoonMaster? 🚀 Growth is accelerating again: Revenue growth increased from 85% in Q1 to 106% in Q2. Data Center growth accelerated from 92% to 117%. NVIDIA added almost $15 billion in revenue sequentially, and its Q3 outlook calls for another $12 billion increase ☁️ Demand is becoming more diversified: Hyperscale revenue reached $48.7 billion, while AI Clouds, Industrial, and Enterprise revenue reached $40.3 billion. NVIDIA is increasingly benefiting from neoclouds, enterprises, and sovereign AI projects, in addition to the traditional Big Tech customers 🔄 Rubin is arriving without a slowdown: Blackwell Ultra is still ramping, but Rubin shipments have already begun. Instead of seeing demand pause between architectures, NVIDIA appears to be stacking one generation on top of the next 📉 Margins are finally coming under pressure: Gross margin is expected to fall from 75% in Q2 to 74% in Q3 and roughly 71% to 72% in Q4. Higher memory costs are the main reason. So far, this looks more like a supply issue than a sign that demand or pricing power is weakening 🧱 NVIDIA is locking down supply: Supply and capacity commitments jumped from $119 billion to $279 billion in just three months, largely because NVIDIA wants to secure memory. That is a huge bet on future demand, but it also creates greater exposure if the AI spending cycle eventually slows 🔮 The outlook remains extraordinary: NVIDIA expects roughly 70% revenue growth in FY28, well above previous Wall Street expectations. If that forecast holds, the AI infrastructure cycle is still expanding at an incredible pace despite NVIDIA already operating at enormous scale The big picture: Growth is accelerating, demand is spreading across more types of customers, and Rubin is arriving before Blackwell has even started to slow down. The biggest new issue is the cost of maintaining that pace. Memory prices are pressuring margins, while NVIDIA’s supply commitments are rising rapidly 2. Business Highlights ⚡ Rubin Is Making Every Gigawatt More Valuable NVIDIA is generating more revenue from each new generation of AI infrastructure Management estimates that its revenue opportunity per gigawatt has increased from roughly $18 billion with Hopper to $25 billion with Blackwell. With Vera Rubin, that figure rises to approximately $40 billion per gigawatt. Rubin combines GPUs, CPUs, networking, and software, delivering 30x higher throughput per megawatt and 35x lower token costs than Grace Blackwell Ultra. Production shipments have already started, with purchase orders from every major hyperscaler, AI cloud provider, and system OEM. 🏦 NVIDIA Is Becoming an AI Financier NVIDIA has invested nearly $50 billion in frontier AI labs and has worked with major financial institutions to help raise more than $500 billion in third party capital for AI infrastructure. The company is also using credit support and take or pay agreements to help AI labs and neoclouds finance additional capacity Critics describe this as circular financing. NVIDIA's argument is simpler: its customers have growing demand, but their balance sheets cannot always fund the infrastructure required to meet it Either way, NVIDIA is doing more than selling the infrastructure. It is increasingly helping customers finance and build it There is also a strategic advantage. By helping fund infrastructure today, NVIDIA can potentially secure deployments before competing chips have enough scale to challenge it. 🤗 NVIDIA Agrees to Buy Hugging Face According to The Information, NVIDIA has agreed to acquire Hugging Face for $12.9 billion, nearly three times the company's 2023 valuation..Hugging Face is one of the largest platforms for developers to discover, share, and deploy open AI models. It is often described as the "GitHub of AI" With only around $150 million in annual revenue, NVIDIA is clearly paying for strategic positioning rather than immediate profits The deal also fits with Jensen Huang's broader argument that NVIDIA benefits as AI models spread across the industry. Open models are especially important because startups and enterprises typically do not build their own chips. They rely on existing computing infrastructure. Owning Hugging Face would bring NVIDIA closer to developers and strengthen the connection between open-source model adoption and NVIDIA's broader platform. The risk is that NVIDIA ownership could undermine the neutrality that helped make Hugging Face so valuable in the first place. 3. Key Quotes From the Earnings Call CEO Jensen Huang on Custom Chips Jensen's argument against the custom-silicon threat is not that NVIDIA's customers will stop building their own chips. He acknowledges that they will. His point is that custom chips are usually designed around specific workloads, while NVIDIA offers a broader platform covering training, inference, networking, CPUs, and multiple cloud environments The real question is whether that flexibility and breadth remain valuable enough to justify NVIDIA's premium economics as customers develop more specialized silicon On the Shift Toward Agentic AI Huang believes AI is moving toward a world where millions of agents could operate continuously and interact with one another Today, much of AI usage still starts with a human request. In an agentic world, AI systems could be working constantly in the background If that happens, inference demand would no longer depend as directly on how often humans interact with AI. The amount of computing power required could rise dramatically His broader argument is that AI is already doing useful work and generating profitable output. If additional compute allows these systems to produce more valuable output, companies have a direct financial incentive to keep buying more computing power. 4. What to Watch Next NVDA is up roughly 20% year to date and continues to outperform the S&P 500 Even after the post earnings rally, the stock trades at roughly 20x forward earnings. That is well below the multiple NVIDIA commanded earlier in the AI boom and below many other major U.S. technology companies But valuation alone does not settle the debate The bigger question is whether today's extraordinary earnings can continue. In semiconductors, a low P/E ratio can sometimes mean earnings are near a peak rather than that the stock is cheap. Investors may be assigning a lower multiple because they are questioning how sustainable current profitability really is So far, that skepticism has repeatedly been wrong Here’s What I’m Watching Custom silicon: OpenAI has published early results for Jalapeño, its custom inference chip. The company says it delivered 1.5x to 1.9x higher throughput per watt and materially lower latency than the NVIDIA systems tested across several models OpenAI still plans to use NVIDIA extensively, but Jalapeño is another sign that NVIDIA's biggest customers have strong incentives to move specialized inference workloads onto their own chips Circular financing: NVIDIA says AI labs receiving some form of balance-sheet support could represent roughly one-quarter of its business next year That does not mean the demand is fake. But it does increase NVIDIA's exposure if AI labs eventually struggle to generate enough revenue from the infrastructure they are building. Margins and supply: Memory shortages are expected to push gross margins lower before pricing and other factors help margins recover next year. The key question is whether NVIDIA can continue securing enough supply to satisfy demand without giving up too much profitability 📉 The bear case: AI infrastructure spending eventually grows faster than the profits it can generate, while custom chips gradually take more of the inference market 📈 The bull case: Agentic AI creates a new wave of computing demand that grows faster than efficiency improvements and competition can bring it down For now, NVIDIA is still winning that race