NVIDIA's Vera CPU opens a new front against AMD and Intel in the AI server
NVIDIA detailed its Vera data-center CPU — 250 to 450 watts, up to 1.5 terabytes of memory per chip — pushing NVIDIA into the server-CPU market long held by AMD and Intel. Paired with Rubin GPUs into one supercomputer, Vera is NVIDIA's move to own the whole rack, not just the accelerator.
Building its own CPU is NVIDIA following the bottleneck. At frontier scale the constraint is moving data between chips, and a CPU co-designed with the GPU and the interconnect lets NVIDIA optimise the whole rack as one machine rather than bolting its accelerator onto someone else's processor. Vera is the piece that turns a GPU vendor into a full-system vendor.
The 1.5-terabyte memory figure is the tell about the target workload. Massive per-chip memory is aimed at the inference and long-context serving that now dominates AI compute, where feeding the accelerators matters as much as the accelerators themselves. Vera is designed for the workload the market actually runs, not the training benchmark that sells headlines.
For AMD and Intel it is a direct threat to a market they assumed was theirs. A vertically integrated NVIDIA rack — CPU, GPU, networking co-designed — is a harder thing to compete against piecemeal, and it extends NVIDIA's lock on the frontier from the accelerator to the entire compute node.
CNBC — Nvidia details its next-generation Vera CPU for AI, setting up challenge to AMD and Intel → · iTiger — Nvidia announces Rubin AI chips for 2026, trillion-dollar data center boom →