Nvidia's Vera CPUs now available on Oracle Cloud, outside GPU racks
Standalone Vera availability suggests customers want the CPU for work that does not need a Rubin GPU, though production use is still unproven.
Nvidia and Oracle have made Vera Rubin NVL72 racks and standalone Vera CPUs generally available on Oracle Cloud Infrastructure Superclusters, a move that puts Nvidia's CPU into a hyperscaler's general-purpose fleet rather than only alongside its GPUs. Tom's Hardware's own benchmarks of a Vera server, which were not run on Oracle's infrastructure, show its memory bandwidth more than doubling a dual Intel Xeon 6980P system in the Stream triad test, and it also posts strong Python and SQLite results. Oracle has said it intends to offer the HGX NVL8 platform, pairing eight Rubin GPUs with a dual 64-core Xeon 6 host and 4TB of DDR5, and the source frames the broader question as whether customers will adopt Vera in production beyond agentic AI loops.
Why it matters
For cloud users, Oracle's offer means Nvidia's Arm-based Vera CPU can now be rented on its own rather than only inside GPU systems, which widens the choice of general-purpose compute on a major hyperscaler. The benchmark results in the source come from hardware outside Oracle's cloud, so they indicate Vera's memory bandwidth advantage over a dual Xeon 6980P server rather than the performance customers will see on Oracle. The open question is whether customers will adopt Vera for production workloads beyond agentic AI, and the facts do not yet answer that.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.