Why NVIDIA Has Become More Than a GPU Company

NVIDIA still makes the most sought-after chips in the world. But describing it as a GPU company now misses most of what it does. In 2026 it sells whole data center systems, finances the labs that buy them, backs governments building national AI capacity, and supplies the software and simulation tools for robots. The company has become something closer to an infrastructure platform for the AI economy. For more on the industry around it, see our AI chip and semiconductor coverage.

The scale of the business

Start with the numbers. In its second-quarter fiscal 2027 results, announced August 26, 2026 for the quarter ended July 26, NVIDIA reported revenue of $96.2 billion, up 106 percent from a year earlier. Data center revenue was $89.0 billion, up 117 percent, and gross margin was 75 percent.

The mix is changing as well. Futurum’s breakdown shows hyperscalers contributed $48.71 billion while AI clouds, industrial, and enterprise customers contributed $40.31 billion. That second group is growing faster: Fifth Person reports its revenue rose 138 percent, and says it is expected to approach half of data center revenue. In other words, NVIDIA is no longer just a supplier to a handful of cloud giants.

A platform, not a chip

When analysts asked CEO Jensen Huang on the earnings call how he balances investing in AI labs that are also designing their own chips, he answered that NVIDIA is building a platform, not a single chip, spanning the entire AI life cycle and running in every cloud.

We described the technical side of that platform in why NVIDIA is expanding beyond GPUs: CPUs, networking, data processors, storage, inference accelerators, and software all sold together. NVIDIA’s own summary of its edge, as reported from the call, is the combination of its full-stack AI factory and the CUDA ecosystem, which lets it extend AI into new markets. CUDA, the software layer developers have built on for years, is a big reason customers who try other chips still keep NVIDIA hardware in the mix.

NVIDIA as financier

The biggest change is that NVIDIA now helps pay for the demand it serves. According to the call summary, the company has invested nearly $50 billion in frontier AI labs and is working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR on financing platforms expected to raise more than $500 billion of third-party capital. Futurum adds that this is subject to definitive agreements, so it’s a goal, not a closed deal.

CNBC reported that this includes a $30 billion stake in OpenAI earlier this year, and that Huang said his only regret was not investing more and sooner. It also noted that NVIDIA is increasingly providing backstops and other arrangements that let new data centers get funded and built. One summary says NVIDIA committed credit support for AI-cloud infrastructure, including up to $105 billion tied to an Ohio OpenAI campus. On the call, NVIDIA said OpenAI’s existing and planned commitments represent roughly 12 gigawatts of NVIDIA compute.

This matters because it shows NVIDIA acting less like a parts supplier and more like a company shaping the market it sells into. It also raises questions about concentration and risk, which we cover below.

Sovereign AI: selling to countries

A second pillar is national demand. NVIDIA says sovereign AI revenue, delivered mainly through regional neoclouds, grew 35 percent from the previous quarter and more than tripled year over year. Management noted that the growth of sovereign AI and specialized clouds now balances the contribution of traditional hyperscalers.

Governments want their own AI capacity for security, economic, and cultural reasons, and they often can’t build custom chips of their own. That makes a complete, ready-to-deploy platform especially attractive, and it’s another reason NVIDIA sells systems instead of chips. The infrastructure behind this demand also feeds the power and supply issues we discussed in why AI data centers need much more power.

Physical AI: robots and simulation

NVIDIA is also pushing into machines that move in the real world. In the Q2 announcements, Amazon agreed to adopt NVIDIA’s full physical AI stack, including Omniverse, Cosmos, Isaac, and Jetson, to power its fleet of warehouse robots. The same partnership expands NVIDIA’s work with AWS, including deploying an additional 2 million GPUs.

Robotics is still small next to data center sales, but it extends the platform logic: simulation tools to design and train robots, edge computers to run them, and models to give them skills. It’s the same strategy applied to a new market.

Locking in supply

A less visible shift is how NVIDIA manages its own supply chain. Futurum reports that procurement commitments jumped to $279 billion from $119 billion the previous quarter, mainly tied to memory. NVIDIA says supply stays a bottleneck through fiscal 2028, and Futurum notes that rising memory costs create a near-term margin tradeoff, with gross margin expected to bottom at 71 to 72 percent. As we explained in why semiconductor supply chains matter for AI and why TSMC is so important to the global AI industry, reserving manufacturing, packaging, and memory capacity is now part of the competitive advantage.

Why it matters for rivals

This platform strategy is NVIDIA’s response to the trend we covered in how cloud companies are becoming AI hardware companies. Amazon, Google, and Microsoft now sell their own silicon, so NVIDIA competes by offering something harder to replace: a full stack, a financing network, and relationships with customers that don’t build chips at all. Even so, it works with rivals when it makes sense, as in the AWS expansion.

The risks

The strategy carries real risk. The customers NVIDIA depends on are spending enormous sums: CNBC noted that several top customers are cash-flow negative, a theme we explored in why Big Tech is spending billions on AI infrastructure. When a supplier also invests in and backstops those customers, its fortunes become more tied to theirs, and a downturn could hit both sides at once. The call summary notes NVIDIA’s exposure is expected to decline as OpenAI makes its lease payments, but that depends on the buyer performing.

Competition is another challenge. The labs NVIDIA funds are also designing or buying other chips, and the cloud providers are scaling their own. And because NVIDIA says demand exceeds supply, its growth guidance of about 70 percent for fiscal 2028 is limited by what it can manufacture, not by what customers want.

The bigger picture

NVIDIA has become more than a GPU company because the GPU alone no longer captures how AI infrastructure gets built. It now supplies the chips, the system around them, the software developers rely on, and increasingly the capital that lets customers buy them, while extending into sovereign AI and robotics. That makes it the closest thing the industry has to an infrastructure platform, and also a company whose success is tied to the health of the entire AI build-out. Follow our AI hardware coverage for the next steps in that story.