NVIDIA pushes beyond GPUs to CPUs

1/ Nvidia’s latest message to investors is that the company no longer wants to be understood as a GPU vendor with attached software. It is selling a full AI system: CPU, GPU, networking, memory architecture, racks and deployment model. NVIDIA’s May 20 earnings release also recast reporting around two platforms — Data Center and Edge Computing — a sign of how it wants the business framed. (investor.nvidia.com) 2/ The line that set off the latest round of coverage came from CFO Colette Kress on the fiscal Q1 2027 call. The Register reported she said Nvidia has “visibility to nearly $20 billion in total CPU revenue this year,” which she said would set it up to become “the world’s leading CPU supplier.” Reuters-style caution is warranted here: that quote is reported by The Register, not stated in the press release itself. (theregister.com) 3/ Why that matters: Nvidia already had CPUs. Grace has been in the market as the Arm-based host processor paired with GPU systems. The new claim is broader. It suggests Nvidia sees CPU revenue becoming large enough to stand on its own, not just as a supporting component inside GPU-heavy AI boxes. (theregister.com) 4/ The roadmap behind that claim is now fairly clear. Blackwell is the current revenue engine. Vera Rubin is the next major platform. Nvidia said in March that the Vera Rubin platform combines the Vera CPU, Rubin GPU, NVLink 6, ConnectX-9, BlueField-4 and Spectrum-6 into one rack-scale system for AI factories. (nvidianews.nvidia.com) 5/ CNBC reported in February that Vera Rubin is expected to ship in the second half of 2026. Nvidia’s March product announcement then said the platform was in full production, framing it as infrastructure for pretraining, post-training, test-time scaling and agentic inference. Those two points together explain why investors are treating late 2026 as the next important handoff after Blackwell. (cnbc.com) 6/ This is not just a chip cadence story. It is a bill-of-materials story. The more of the system Nvidia supplies itself — CPU, GPU, interconnect, NIC, DPU, Ethernet, rack design — the more of the revenue pool stays inside Nvidia rather than being split across Intel, AMD, Broadcom, Marvell or server OEM layers. That inference follows directly from Nvidia’s own rack-scale product descriptions. (nvidianews.nvidia.com) 7/ Nvidia’s reporting changes reinforce that read. In the May 20 release, the company said Data Center will now be split into Hyperscale and ACIE, covering AI Clouds, Industrial and Enterprise, while Edge Computing will include PCs, game consoles, workstations, AI-RAN, robotics and automotive. That is a wider map than “sell accelerators to hyperscalers.” (investor.nvidia.com) 8/ Jensen Huang’s wording on the quarter was similarly expansive. He said Nvidia is “the only platform that runs in every cloud, powers every frontier and open source model, and scales everywhere AI is produced — from hyperscale data centers to the edge.” That is platform language, not component language. (investor.nvidia.com) 9/ Vera Rubin also shows how Nvidia is trying to make the CPU matter inside its own stack. Nvidia’s March announcement put the Vera CPU alongside the Rubin GPU as part of a single “giant supercomputer,” rather than describing the CPU as a commodity host chip. In other words, the CPU is being positioned as a designed-in part of the system architecture. (nvidianews.nvidia.com) 10/ The near-term sequence now looks like this: Blackwell drives current sales; Vera Rubin is the late-2026 transition; Rubin-era systems are the vehicle for Nvidia to deepen control over more of the platform. The exact revenue split between standalone CPUs and CPUs bundled into Rubin systems has not been fully disclosed in Nvidia’s official materials, so any sharper breakdown should be treated as reported analysis rather than confirmed company guidance. (investor.nvidia.com) 11/ The financial backdrop is why the market is paying attention. Nvidia reported first-quarter fiscal 2027 revenue of $81.6 billion, up 85% from a year earlier, with data center revenue of $75.2 billion, up 92%. A company with that scale can use a roadmap transition to pull customers toward its preferred system design, not just its next chip. (investor.nvidia.com) 12/ The practical takeaway: Nvidia is trying to move the conversation from “which GPU wins?” to “which company defines the whole AI machine?” The next public checkpoints are Vera Rubin shipments in the second half of 2026 and whatever Nvidia discloses in upcoming quarterly filings and calls about CPU revenue, platform mix and the Blackwell-to-Rubin transition. (cnbc.com)