NVIDIA targets AI data centers with 88-core Vera CPU as demand surges

- Early clients include OpenAI, Anthropic, Oracle, ByteDance, SpaceXAI, CoreWeave - Vera claims 1.8x speed boost over x86 CPUs in AI workloads On June 1, 2026, CoinDesk reported that NVIDIA had made its most decisive move yet into the AI data center market by unveiling the Vera CPU. The 88-core, Olympus-powered chip is NVIDIA’s first in-house CPU purpose-built for AI agent workloads, and this positions it as a direct challenger to Intel and AMD’s dominance in the segment. According to CoinDesk on June 1, 2026, Vera is tailored for demanding tasks such as code compilation and database operations, and it provides an alternative to legacy x86 CPUs while also moving beyond NVIDIA’s previous ARM-based attempts like Grace. CoinDesk further reported on June 1, 2026, that Vera’s launch has already drawn in leading AI research labs and top cloud providers, and early clients include OpenAI, Anthropic, Oracle, ByteDance, SpaceXAI, and CoreWeave. In addition, major server OEMs such as Dell, HPE, Lenovo, and Supermicro have committed to shipping Vera-powered systems, and this marks the first time a non-x86 CPU will be broadly offered as a standard option across the industry. According to the same CoinDesk report, NVIDIA integrates its full-stack infrastructure into Vera, as it connects the CPU to GPUs via second-generation NVLink-C2C and to networking and storage through BlueField-4 technology. The 88-core chip also leverages high-efficiency LPDDR5X memory with 1.2 TB/s of bandwidth, and observers say this specification could increase pressure on low-power DRAM supply chains. CoinDesk reported that NVIDIA-sanctioned benchmarks tout Vera’s ability to surpass x86 CPUs, and the company claims up to 1.8x performance in select tasks and 10% faster Linux kernel compilation times compared to AMD’s flagship EPYC 9575F. However, CoinDesk also noted that reviewer Michael Larabel of Phoronix emphasized these results are based on NVIDIA-selected workloads, and he further pointed out that comprehensive, independent benchmarks—particularly for general-purpose and legacy x86 applications—are still lacking. According to CoinDesk on June 1, 2026, Vera systems are expected to ship to cloud partners starting in the third quarter of 2024. As a result, NVIDIA’s aggressive entry signals a turning point in the AI data center hardware ecosystem, as it challenges entrenched players and promises substantial performance advantages, while industry attention now focuses on independent validation and real-world performance as the next wave of AI infrastructure takes shape.