Nvidia says Vera CPUs expand addressable market to $200B, counting China

Nvidia’s latest message is that Vera is not a side project to support GPUs. It is part of a broader attempt to sell whole AI systems — CPU, GPU, networking, interconnect and software together. Jensen Huang said on May 23 that Nvidia’s new $200 billion CPU market estimate includes China, extending the company’s long-term market framing even as near-term China revenue remains constrained. (cnbc.com) That matters because Nvidia has spent the past three years being valued primarily as the dominant seller of AI accelerators. (cnbc.com) Vera changes the pitch. Nvidia said in March that the Vera CPU was built for “agentic AI and reinforcement learning,” and described it as a processor designed to work inside rack-scale AI factories rather than as a general-purpose server CPU sold on its own. (techtimes.com) ### Why is Nvidia talking about CPUs now, after years of a GPU-led boom? Jensen Huang tied the CPU push directly to a bigger systems opportunity. Asked in Taipei whether the $200 billion CPU forecast included China, he said it did, according to CNBC’s report of his comments on Saturday, May 23. (cnbc.com) Nvidia’s own product disclosures show why the company is emphasizing that figure. The Vera Rubin platform combines a Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9 SuperNIC, BlueField-4 data processing unit and Spectrum-X Ethernet switch in one system design. In other words, Nvidia is presenting the CPU as one component of a tightly integrated platform, not as a standalone x86 replacement story. (nvidianews.nvidia.com) ### What does Vera actually add to Nvidia’s addressable market? Nvidia said in its March launch material that Vera is its “first processor purpose-built” for the current AI buildout and said it delivers twice the efficiency and 50% faster performance than “traditional rack-scale CPUs.” The company linked that claim to workloads such as reasoning, tool use, code execution and validation. (cnbc.com) The financial framing is also explicit. A transcript summary of Nvidia’s May 20 earnings call said management pointed to roughly $20 billion in standalone CPU revenue for the year and a $200 billion market opportunity opened by the Vera platform. That suggests Nvidia wants investors to think about CPUs as incremental revenue on top of GPUs, not as a separate adjacent business. (techtimes.com) ### What did Colette Kress say that kept the demand story alive? Colette Kress’s most cited signal was pricing. 24/7 Wall St. reported that Nvidia’s CFO said H100 rental prices had risen 20% so far in 2026 and older A100 rental prices had climbed 15%, a sign that older AI chips were not following the usual depreciation pattern. (nvidianews.nvidia.com) Nvidia’s official results showed the scale behind that backdrop. The company reported first-quarter fiscal 2027 revenue of $81.6 billion on May 20, up 85% from a year earlier, with data center revenue of $75.2 billion, up 92%. Nvidia also guided second-quarter revenue to about $91 billion. (fool.com) ### Why is Taiwan suddenly central to this story? Taiwan is central because Vera Rubin is a manufacturing-heavy platform, not just a chip announcement. TechTimes reported that Huang traveled to Taiwan ahead of Computex 2026 to meet TSMC Chairman C.C. Wei as Nvidia sought production commitments for Vera Rubin. (247wallst.com) The same report said each NVL72 Vera Rubin configuration links 36 Vera CPUs and 72 Rubin GPUs, contains nearly two million parts and relies on about 150 Taiwanese supply-chain partners. Huang said the overlap between Grace Blackwell GB300 output and Vera Rubin ramp-up would make for “a very busy second half” for Taiwan’s supply chain. (investor.nvidia.com) ### So is this still a GPU story, or something broader? Nvidia’s own materials point to a broader stack. Its GTC 2026 releases bundled Vera Rubin with networking, storage, software and AI-factory reference designs, while Huang’s public comments in Taipei tied the CPU opportunity to a systems market that includes China over the long run. (techtimes.com) The next concrete checkpoint is Huang’s Computex and GTC Taipei appearances and any further disclosures from Nvidia and TSMC on Rubin production capacity, packaging and delivery timing. Nvidia has already said Vera Rubin is in full production, and investors will now be watching whether that manufacturing ramp matches the demand signals Nvidia described on May 20 and May 23. (nvidianews.nvidia.com 1) (nvidianews.nvidia.com 2) (techtimes.com)