NVIDIA Hand-Delivers First Vera CPUs to Anthropic, OpenAI, and SpaceXAI as Agentic AI Drives a New CPU Moment
NVIDIA VP Ian Buck personally delivered the first production Vera CPU systems to Anthropic, OpenAI, SpaceXAI, and Oracle in California, marking the commercial arrival of NVIDIA's first purpose-built agentic AI processor.
Overview
NVIDIA Vice President of Hyperscale and High-Performance Computing Ian Buck made personal deliveries of the first Vera CPU systems to Anthropic, OpenAI, SpaceXAI, and Oracle Cloud Infrastructure across four California locations on May 17 and 20, 2026, according to the NVIDIA Blog. The hand-deliveries mark the commercial arrival of Vera — NVIDIA’s first custom CPU built specifically for agentic AI workloads and, as previously reported, the companion processor to the Rubin GPU in NVIDIA’s next-generation data center platform.
The Deliveries
Buck brought the first systems to Anthropic’s San Francisco office, then to OpenAI’s Mission Bay headquarters, and then to SpaceXAI’s Palo Alto facility on Friday, May 17, according to the NVIDIA Blog. A fourth delivery followed the following Monday at Oracle Cloud Infrastructure’s Santa Clara campus.
James Bradbury, Anthropic’s head of compute, responded to the delivery with a statement focused on compute scaling: “Scaling compute is an important accelerant for the growth of models. We’re excited to see Vera emerge as a promising part of the ecosystem when solving for agentic workloads,” according to the NVIDIA Blog.
Oracle Cloud Infrastructure’s product management lead Karan Batta signaled the largest known deployment commitment: “OCI plans to deploy hundreds of thousands of NVIDIA Vera CPUs beginning in 2026 because agentic AI demands sustained performance at massive scale,” as reported by the NVIDIA Blog.
What Vera Is
Vera is built around 88 custom NVIDIA-designed Olympus cores — the die contains 91 cores, with three held in reserve for redundancy, according to Tom’s Hardware. Each active core can run two tasks concurrently through NVIDIA Spatial Multithreading. Memory bandwidth reaches up to 1.2 TB/s via LPDDR5X, and the chip connects to NVIDIA GPUs at 1.8 TB/s of coherent bandwidth over NVLink-C2C — a figure NVIDIA describes as seven times the bandwidth of PCIe Gen 6 — according to the NVIDIA Newsroom.
A standard Vera CPU rack holds 256 liquid-cooled units and is rated for more than 22,500 concurrent CPU environments running independently at full performance, per the NVIDIA Newsroom. NVIDIA claims the processor delivers twice the energy efficiency and is 50% faster than traditional CPUs.
Vera runs a single product configuration. “We only are going to build one Vera CPU [SKU],” Ian Buck told Tom’s Hardware at GTC in March, adding that Jensen Huang foresees the product becoming “a multi-billion dollar business.” NVIDIA has also stated it has no intention to compete broadly against AMD EPYC or Intel Xeon in general-purpose server compute, focusing instead on agentic AI orchestration.
The Agentic AI Argument
NVIDIA’s case for a dedicated CPU rests on a shift in how frontier AI models are being used. Jensen Huang, NVIDIA’s Founder and CEO, framed it this way at the March 16 GTC launch: “Vera is arriving at a turning point for AI. As intelligence becomes agentic — capable of reasoning and acting — the importance of the systems orchestrating that work is elevated. The CPU is no longer simply supporting the model; it’s driving it. With breakthrough performance and energy efficiency, Vera unlocks AI systems that think faster and scale further,” according to the NVIDIA Newsroom.
Buck expanded on the operational context in his delivery statement: “Agentic AI is creating a new CPU moment in the AI factory — as models move from answering to acting, Vera is purpose-built to keep that work moving at scale,” as reported by the NVIDIA Blog.
External testing supports some of the performance framing. Alex Gallego, Founder and CEO of Redpanda, reported that the company tested Vera on Apache Kafka-compatible workloads and “saw dramatically better performance than other systems we’ve benchmarked, delivering up to 5.5x lower latency,” adding that “Vera represents a new direction in CPU architecture, with more memory and less overhead per core,” according to the NVIDIA Newsroom. John Cazes, Director of HPC at the Texas Advanced Computing Center (TACC), said in the same release that testing across six scientific applications showed “impressive early results” and that “Vera’s per-core performance and memory bandwidth represent a giant step forward for scientific computing.”
Cursor’s cofounder and CEO Michael Truell also cited productivity gains: “We’re excited to use NVIDIA Vera CPUs to improve overall throughput and efficiency so we can deliver faster, more responsive coding agent experiences for our customers,” per the NVIDIA Newsroom.
What We Don’t Know
The delivery event signals the beginning of Vera’s commercial availability, but key questions remain open. Volume production shipments to the broader partner ecosystem are not scheduled until the second half of 2026. The actual workload configurations deployed at Anthropic, OpenAI, and SpaceXAI have not been disclosed, and no independent benchmarks of production Vera systems have yet been published. Pricing has not been made public. Whether Vera’s performance advantages prove durable across diverse agentic workload profiles, rather than the curated benchmarks in NVIDIA’s press materials, will not be known until third-party data emerges.
OCI’s stated intent to deploy “hundreds of thousands” of units is the largest concrete adoption signal, but represents a forward-looking commitment rather than completed deployments.