NVIDIA AI Updates: May 28, 2026

1. NVIDIA Reframes AI Factories as Token Factories

NVIDIA. A new post on the company blog frames data centers as token factories where performance-per-watt and cost-per-token are the operational metrics that matter as agentic AI scales. The argument is mostly a marketing setup for upcoming Vera Rubin and Blackwell Ultra positioning, but the unit economics framing — token throughput as the production metric — is a useful lens for capacity planning that’s worth borrowing regardless of vendor. Source

2. Vera CPU Benchmarks Land on Phoronix

NVIDIA. Phoronix published the first independent Vera CPU benchmarks and NVIDIA’s blog highlights the results: significant memory-bandwidth gains on agentic workloads versus x86 incumbents. As always with vendor-amplified third-party benchmarks, the specific configs (memory channels, NUMA topology) matter, but Vera’s bandwidth headroom does appear to land where NVIDIA promised when it announced the CPU last year. Source