AI BriefingNVIDIA Delivers First Vera CPUs for Agentic AI to Major AI Labs
Key point
NVIDIA has delivered its first Vera CPUs for agentic AI to major AI labs and OCI.
Details
Vera is a standalone CPU designed by NVIDIA for agentic AI. After being introduced as a next-generation business at GTC San Jose in March, customer deployment has now begun for the first time.
Ian Buck personally delivered the systems to Anthropic, OpenAI, SpaceXAI, and Oracle Cloud Infrastructure (OCI). NVIDIA explained that agents require heavy CPU-centric workloads not just for generating answers but also for tool call, orchestration, long-context retrieval, and sandboxed execution.
Key specifications are as follows.
- 88 custom NVIDIA-designed Olympus cores
- 1.2 TB/s memory bandwidth
- 50% improvement in per-core performance
OCI plans to deploy hundreds of thousands of units of Vera CPUs starting in 2026, making it the first cloud provider to adopt Vera at scale. Vera is part of NVIDIA's co-design strategy alongside Rubin GPU, BlueField 4 DPU, Spectrum-X, and MGX. In the Vera Rubin NVL72, it is connected to Rubin GPUs via NVLink-C2C, supporting a shared memory architecture and high GPU utilization, which NVIDIA describes as delivering 2x energy efficiency compared to existing infrastructure.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Stanford and NVIDIA Release 'Contrastive Language Model' CLM with 9x Faster Inference
Anthropic Pacing to $100B ARR; Rumors of OpenAI Solving Navier-Stokes Spark ControversyOpenAI and Anthropic's RSI Strategy and the Uncertainty of AI Profitability