NVIDIA Delivers First Vera CPUs to OpenAI and Other AI Labs, Reveals Taiwanese Supply-Chain Partners

Agentic AI depends on more than GPUs. Models still rely heavily on CPUs to call tools, run programs and manage long contexts. NVIDIA is therefore using its in-house Vera processor to challenge Intel and AMD in the roughly $200 billion data-center CPU market, while driving upgrades across Taiwan's server supply chain.

NVIDIA launched Vera on March 16, 2026, claiming it was 50% faster per core than conventional rack-scale CPUs and offered twice the energy efficiency. The first systems were delivered to Anthropic, OpenAI and xAI on May 15, followed by Oracle Cloud Infrastructure on May 18. Taiwanese partners include ASUS, Compal, Foxconn, GIGABYTE, Pegatron, QCT, Wistron and Wiwynn.