NVIDIA announced Vera at GTC Taipei and said the new CPU class is now in full production. Designed specifically for agentic AI workloads, Vera targets the critical CPU tasks that sit on the modern AI stack’s critical path-from orchestration to sandboxed code execution — while promising improved energy efficiency in data centers. Cloud and hyperscale operators, AI service providers and enterprises running agent pipelines are the most immediately affected as they evaluate new server and cloud options.
NVIDIA claims Vera completes tasks 1.8x faster than comparable x86 CPUs and explicitly targets workloads such as agentic AI, reinforcement learning and data processing. The company says the Vera design builds on its existing Grace family of CPUs, which have nearly 2.5 million shipments to date, and is intended to accelerate the software layers that drive agent throughput and responsiveness. Vera will appear across NVIDIA’s own hardware lines: standalone Vera servers, Vera Rubin systems and Vera BlueField‑4 STX AI storage platforms. Those integrated products are positioned alongside a broader OEM and systems ecosystem so customers can access Vera in multiple form factors and deployment models.
A range of cloud, hyperscale and AI organizations are already planning to adopt or evaluate Vera. Named customers and partners include Anthropic, OpenAI, SpaceXAI, ByteDance, CoreWeave, Lambda, Nebius, Nscale and Oracle Cloud Infrastructure. The New York Stock Exchange is testing Vera in collaboration with Redpanda and HPE, highlighting interest from both financial and enterprise IT environments. Major system vendors and Taiwan builders are preparing Vera‑based offerings at scale. Dell Technologies, HPE, Lenovo and Supermicro are listed among the system vendors, while ASUS, Foxconn and Quanta Cloud Technology are named as Taiwan manufacturers expected to build Vera systems into their portfolios.
Independent benchmarking cited by NVIDIA-from Phoronix — reported that Vera delivered the fastest overall performance across the agentic workloads included in the tests, specifically code compilation, Python and Java execution and database processing. NVIDIA positions faster CPU completion of these tasks as a direct lever to increase agent throughput and interactivity in production pipelines. Company executives framed Vera as infrastructure for the next phase of AI. Jensen Huang said, “AI agents will be the largest users of computing,” and NVIDIA positioned Vera as purpose‑built for running agentic AI at hyperscale. Oracle’s Mahesh Thiagarajan described Vera‑powered OCI Supercluster deployments as supporting high‑throughput reasoning and data processing for next‑generation AI environments.
Sources
Replies (0)
No replies in this topic yet.
