Vera CPU gains ecosystem traction
NVIDIA's Vera CPU from GTC is getting ecosystem callbacks — Starburst published optimizations for Vera to improve federated data access for AI. That early tooling signals Vera is already being positioned as part of agentic AI reference architectures. (vocal.media) (hpcwire.com)
Starburst’s Model Context Protocol (MCP) Server exposes a standardized Agent data gateway that lets external agents such as Claude and ChatGPT inspect catalogs, sample rows, and retrieve governed business context for SQL grounding. (starburst.io) Early Starburst benchmark language in the company release reports “materially faster query performance” and “significantly higher CPU efficiency” versus comparable x86 configurations, and ties optimizations to Vera availability later in 2026. (starburst.io) NVIDIA describes Vera as an Arm-based design with custom “Olympus” cores reported at 88 cores / 176 threads and advertises the CPU as delivering twice the energy efficiency and 50% faster results than traditional rack-scale CPUs. (techpowerup.com) NVIDIA’s reference Vera rack integrates 256 liquid‑cooled Vera CPUs capable of sustaining more than 22,500 concurrent, independently running CPU environments, and pairs Vera with Rubin GPUs over NVLink‑C2C with 1.8 TB/s of coherent bandwidth. (nvidianews.nvidia.com) NVIDIA named early collaborators for Vera deployments including Alibaba, ByteDance, Meta, Oracle Cloud Infrastructure, CoreWeave, Lambda, Nebius and Nscale, and listed manufacturing/system partners such as Dell Technologies, HPE, Lenovo and Supermicro. (nvidianews.nvidia.com) Starburst’s public materials frame the Vera integration as enabling “lakeside” federated queries and fine‑grained governance to serve RAG, inference, and agent workflows without ETL-driven data duplication. (starburst.io)
