News
NVIDIA said Nebius, Nscale and other partners are building on Vera CPUs for agentic AI infrastructure and data-center workloads.
NVIDIA said European technology partners are building with its Vera CPU to support the next wave of AI infrastructure, as the company pushes the chip for agentic computing workloads that depend on fast CPU-side processing.
Evidence and context
In a blog post published Oct. 20, 2026, NVIDIA said Nebius and Nscale are among the global technology partners looking to Vera CPUs to help build infrastructure for agentic AI, where systems retrieve information, run software tools, execute code and evaluate results. NVIDIA said those steps rely on fast CPU performance and that Vera is designed to keep response times steady even when thousands of tasks are running at once.
The post said AI cloud providers Nebius, Nscale and Verda are building with Vera to support the computing needs of businesses and developers, while Aleria, described by NVIDIA as a sovereign AI infrastructure and platform provider, is part of a broader ecosystem focused on customer-controlled data and operations. NVIDIA also said server manufacturers Bull, Dell Technologies, HPE, Lenovo and Supermicro are building Vera-based systems, giving customers more deployment options for AI data centers and high-performance computing.
NVIDIA said data-processing companies including ClickHouse, Cloudera, Duck Labs, Elastic, Polars, Redpanda, SAP and Starburst observed strong results on Vera in early testing for analytical, extract-transform-and-load and semantic search workloads. The company attributed those results to Vera’s per-core performance, LPDDR5X memory bandwidth and NVIDIA Scalable Coherency Fabric, but did not publish comparative figures in the Oct. 20 post.
The blog also highlighted adoption by software and developer-platform firms. NVIDIA said Atlassian is evaluating Vera for data processing that builds semantic search indexes for AI-powered products, Inherent has tested Vera for environments where AI agents work alongside engineers on scientific research, and Nous Research is using it to authenticate agent identities and validate actions before execution. NVIDIA said Baseten-Blaxel, Daytona, E2B, Modal, Prime Intellect and Vercel are building cloud-based sandboxes on Vera to support code execution, tool use and test runs for AI agents.
The update builds on earlier NVIDIA disclosures about Vera. NVIDIA said on Aug. 27, 2026, that Vera was shipping, and background material supplied with this report says the chip was launched on March 16, 2026 for agentic AI and reinforcement-learning workloads, including code execution, compilation, sandboxing, orchestration and data processing. NVIDIA has said Vera supports standalone servers and can serve as the host CPU for Rubin GPU systems. The company’s published specifications include 88 custom Olympus Armv9.2 cores, 176 threads using NVIDIA Spatial Multithreading, up to 1.5 TB of LPDDR5X memory, up to 1.2 TB/s memory bandwidth and 3.4 TB/s bisection bandwidth through its second-generation Scalable Coherency Fabric.
Market data supplied with this report showed NVIDIA shares at $216.47 as of 2026-10-21T10:43:56Z, down 0.35% on the day, with a market value of $5,256.97 billion. That price data is separate from the infrastructure announcements and does not indicate how the Vera program is being received by customers.
Be the first to comment