Nvidia plans to expand from GPU dominance into the CPU market, forecasting nearly $20 billion in total CPU revenue this year. Nvidia previously introduced an Arm datacenter CPU, Grace, and largely integrated it into GPU systems for AI datacenters and supercomputers. In February, Nvidia said Meta began deploying standalone Grace CPU Superchips for multiple workloads, including AI agents. At GTC, Nvidia added a standalone Vera CPU system featuring 88 custom Olympus Arm cores with simultaneous multi-threading and confidential computing. Vera supports up to 1.5 TB of LPDDR5x SOCAMM memory with up to 1.2 TB/s bandwidth. Nvidia claims performance, power efficiency, and rack density advantages versus x86 alternatives, with reference designs using up to two Vera CPUs per board and NVLink interconnects. Nvidia pairs Vera CPUs with Rubin GPUs in rack-scale AI platforms, and major hyperscalers and system builders plan to deploy the chips.
"“We have visibility to nearly $20 billion in total CPU revenue this year, setting us up to become the world's leading CPU supplier,” Nvidia CFO Colette Kress said during the company's Q1 2027 earnings call on Wednesday."
"“Vera will deliver up to 1.5x faster performance per core, 2x performance per watt, and 4x density per rack compared to x86-based alternatives,” Kress claimed."
"Nvidia's reference designs pack up to two Vera CPUs onto a single board and via high-speed NVLink interconnects. Nvidia's Vera is also paired in a 2:1 ratio of Rubin GPUs to CPUs in its most powerful rack-scale AI compute platforms."
"At its GTC conference in March, Nvidia officially expanded its CPU line up to include a standalone Vera CPU system. Each chip features 88 custom Olympus Arm cores with support for simultaneous multi-threading (SMT) - that's Hyperthreading in Intel speak - along with confidential computing capabilities."
Read at theregister
Unable to calculate read time
Collection
[
|
...
]