NVIDIA used its GTC Taipei keynote at Computex 2026 to unveil Vera, the first CPU it has designed specifically for agentic AI workloads. The chip is now in full production and will reach customers this fall through standalone Vera servers, the NVIDIA Vera Rubin AI platform and the Vera BlueField-4 STX AI storage processor. NVIDIA says Vera is 1.8x faster than x86 CPUs on task completion across agentic AI, reinforcement learning and data processing — the workloads that move agents from answering questions to running tools, writing code and evaluating results.
Olympus Cores, LPDDR5X Memory, And A Different Math For The Data Center
Vera ships with 88 of NVIDIA's own "Olympus" cores — a departure from the off-the-shelf Arm Neoverse cores that powered Grace — plus Spatial Multithreading and an LPDDR5X memory subsystem delivering up to 1.2 TB/s of bandwidth. NVLink-C2C provides up to 1.8 TB/s of coherent bandwidth between Vera and Rubin GPUs in tightly coupled accelerated systems, and the chip extends NVIDIA Confidential Computing at rack scale to protect agentic workloads end to end. "AI agents will be the largest users of computing," said NVIDIA founder and CEO Jensen Huang. "Vera is the first CPU designed for that future — built to run agentic AI at hyperscale with extraordinary performance, efficiency and programmability."
Customer List: NYSE, Anthropic, OpenAI, OCI, ByteDance
NVIDIA disclosed an unusually broad first-customer list. NYSE is deploying Vera in partnership with Redpanda and HPE to scale capacity on a system that processes more than 1.1 trillion messages a day. Anthropic — under head of compute James Bradbury — is evaluating Vera "to scale CPU-intensive agentic workloads," and OpenAI and SpaceXAI are on the same first-wave list. Oracle Cloud Infrastructure committed to OCI Superclusters built on Vera; Mahesh Thiagarajan, EVP of OCI, framed the deployments as supporting "high-throughput reasoning and data processing workloads across next-generation AI environments." Hyperscalers ByteDance, CoreWeave, Lambda, Nebius, Nscale and Akamai are also on the list, alongside system builders Dell Technologies, HPE, Lenovo and Supermicro that will sell standalone Vera CPU servers — NVIDIA's first standard CPU server option beyond x86.
An AI-Factory Pitch, Not A Chip Pitch
The Vera message is explicitly economic: AI-factory economics are shifting from "cores per dollar" to "tokens per dollar," and the CPU is what keeps expensive accelerators fed. Vera is positioned to handle the Python runtimes, sandboxed code execution, orchestration logic and analytics pipelines that sit on the critical path of every modern agent, so GPUs spend more time inferring and less time waiting. Phoronix benchmarks cited by NVIDIA show Vera leading on code compilation, Python, Java and database processing — exactly the workloads that determine agent throughput in production.
Where This Sits In NVIDIA's Computex Salvo
Vera is one of several major NVIDIA launches at Computex. We covered the RTX Spark superchip for Windows AI PCs, the Cosmos 3 open physical-AI foundation model and the Vera Rubin platform's TSMC supply-chain ramp. Together with Vera, those releases give NVIDIA a chip-to-rack story spanning consumer PCs, hyperscale agent infrastructure and physical AI for robotics — and a competitive answer to AMD's MI400 and Intel's SambaNova-powered rackscale push.
Reporting based on coverage from NVIDIA Newsroom, HPCwire, CNBC and the Computex 2026 keynote.