NVIDIA and Oracle Expand NVIDIA Vera Rubin to Multi-Gigawatt Scale and Bring NVIDIA Vera CPU to Oracle Cloud Infrastructure - NVIDIA Newsroom

Strategic read: NVIDIA is no longer a chip vendor—it is a compute-sovereignty platform. Vera CPU embedding into OCI severs the x86 moat AMD and Intel have defended for two decades. Once GPU, CPU, and interconnect close the loop under one vendor, the best-of-breed procurement model is dead. Multi-gigawatt is the real tell. AI training campuses now draw power comparable to a mid-sized city. The binding constraint has shifted from silicon to electrons: grid interconnection, transformer capacity, and liquid-cooling infrastructure are the new bottlenecks. Oracle pushing OCI into multi-GW territory means NVIDIA's revenue model is decoupling from unit chip sales and re-anchoring to energy throughput—selling megawatts, not dies. Upstream, this locks China/Taiwan 3nm/2nm wafer capacity and HBM4 supply through 2027. Downstream, it forces AMD's MI350 roadmap into a two-front war: raw FLOPS and full-stack software coherence, where CUDA's fifteen-year ecosystem moat remains nearly impenetrable. Twelve-to-twenty-four-month tail: (1) micro-nuclear reactors co-located with AI campuses become standard; (2) Intel's foundry business faces existential pressure as Arm-based Vera CPU structurally undercuts Xeon in inference; (3) EU and Asian data-center power-consumption legislation will reshape where multi-GW facilities get sited. Bottom line: the real competitor isn't AMD. It's the grid.