At its GTC 2026 developer conference, Nvidia revealed a counterintuitive approach to CPU manufacturing: scale a multi-billion-dollar business with a single product.
Despite demand beating expectations, Nvidia announced it will only produce one Vera CPU model, said Ian Buck, the company's VP and general manager of hyperscale and HPC business. The decision reflects a philosophy of focused engineering over market breadth.
The numbers highlight why this matters. Vera chips contain 91 cores, allowing Nvidia to retain three for redundancy and achieve decent yields with the 88-core part, while everything with fewer than 88 functional cores gets scrapped. By eliminating multiple SKUs, Nvidia avoids the costly process of binning, where chips get sorted by quality and sold at tiered price points.
The strategic rationale is clear. Nvidia's primary goal is using Vera processors for its NVL72 and VR300 rack-scale systems rather than building a fully-fledged CPU business. Buck noted that "the world is not going to be served by one SKU of CPU, and that is not our intention. We like a workload problem to go solve, to go swarm, and Nvidia is making one CPU to help in that agentic workload".
Jensen Huang, Nvidia's founder and CEO, acknowledged the shift: "We never thought we will be selling CPUs standalone, but we are selling a lot of CPUs standalone. This will for sure be a multi-billion dollar business for us".
Market interest has exceeded internal expectations. Customers collaborating with Nvidia to deploy Vera include Alibaba Cloud, ByteDance, Meta and Oracle Cloud Infrastructure. This evolution marks Nvidia's entry into direct CPU sales, positioning itself as a competitor to Intel and AMD in the traditional CPU market.
Vera delivers results with twice the efficiency and 50% faster than traditional rack-scale CPUs, according to Nvidia's performance claims. Unlike Nvidia's earlier Grace processors, which were primarily sold as companions to GPUs, Vera is positioned as a general-purpose data centre CPU with a strong focus on AI-centric workloads such as agentic frameworks, scripting-heavy pipelines, analytics, and code compilation.
The concentrated approach carries trade-offs. While reducing Nvidia's costs and enabling it to achieve strategic goals, this approach will limit market penetration. Organisations requiring different CPU performance tiers or alternative architectures have no Nvidia options.
The chips are now in full production and will be available to Nvidia's partners in the second half of this year. Manufacturing partners already adopting the Vera CPU include Dell Technologies, HPE, Lenovo and Supermicro, along with ASUS, Compal, Foxconn, GIGABYTE, Pegatron, Quanta Cloud Technology (QCT), Wistron and Wiwynn.
For investors and data centre operators, the bet reflects Nvidia's confidence in where computing demand is heading. By optimising Vera for the specific challenge of agentic AI, rather than competing across all server workloads, Nvidia is narrowing its addressable market while maximising economics in the segment it dominates. Nvidia's official announcement provided additional performance metrics for interested implementers.