Nvidia Diversifies into Central Processing Units
Initially achieving market dominance due to the massive global demand for its Graphics Processing Units (GPUs)—the primary chips utilized in artificial intelligence deployment—Nvidia has expanded its focus. The chipmaker is now introducing its own central processing units (CPUs) for AI, branded as Vera. The company released detailed technical specifications, benchmarks, and architectural data to help potential customers thoroughly assess the new hardware.
This move positions Nvidia in a new competitive arena within AI servers, challenging established leaders like AMD and Intel. Previously, Nvidia’s core strength was in GPUs, which were fundamental for creating and deploying AI. However, by shipping its own CPUs, cloud service providers now have an alternative to purchasing hardware from Advanced Micro Devices or Intel.
Representatives from Nvidia confirmed that Vera chips had already been delivered to major clients, including OpenAI, Anthropic, and SpaceX, in June.
Strategic Goals and Market Competition
Nvidia has historically been recognized as a key determinant of the information technology industry’s trajectory. While traditionally viewed as a challenger in the CPU sector, the company’s entry represents a significant strategic push to vertically integrate its systems. The goal is not merely to sell chips, but to sell complete, high-powered computing racks.
This strategy aims to help engineers maximize the performance of Nvidia’s GPUs, thereby maintaining the company’s systems as the preferred choice for leading AI research labs, particularly as competition from custom chips and rival CPUs intensifies.
Historically, the CPU was the most critical and expensive component in a server before the AI boom. Early AI servers, such as those available when ChatGPT launched in 2022, typically paired up to eight GPUs with a single CPU, signaling a major shift in market reliance toward Nvidia’s GPUs.
However, the rise of “agentic AI”—systems capable of running autonomously with minimal human intervention—has brought renewed attention to the CPU’s role, which is necessary for managing and feeding data to these agents. Financial market observers have noted this shift, with CPU incumbents AMD and Intel showing strong performance in 2026, rising 128% and 149% respectively, surpassing Nvidia’s growth of 8%.
Technical Details of the Vera Chip
During a presentation, Nvidia’s vice president of hyperscale, Ian Buck, noted that agentic AI has made CPUs “much more integral,” specifically citing the importance of “how fast a CPU can answer one question.”
Nvidia stated that Vera is the first server CPU designed entirely from its core, rather than using a standard, off-the-shelf design provided by Arm, which reduces necessary engineering work. The chip was engineered specifically to address bottlenecks inherent in AI agents. Nvidia claims that the Vera chip provides 50% better performance for AI agents compared to x86 chips, the type of silicon manufactured by Intel and AMD.
The company emphasized that while previous CPU generations from Intel and AMD emphasized core count, the Vera chip and its Olympus core are optimized for single-core speed. Hannah Coutand, a Vera product marketer, explained that the focus on per-core speed, high memory bandwidth, and low latency allows agents to return to the GPUs quickly, thus keeping these expensive and valuable assets highly utilized.
The Vera chip is designed to be sold independently or paired with Nvidia’s GPUs in a system called Vera Rubin. It is also available in a liquid-cooled rack configuration containing 256 Vera chips, or in a setup featuring two Vera chips within a single server.
Technically, the CPU is power-intensive, drawing between 250 watts and 450 watts of power. It also incorporates a large amount of low-power memory, similar to that found in smartphones and laptops, supporting up to 1.5 terabytes of memory per chip.
Industry Forecasts and Market Adoption
Market analysts have suggested that Nvidia may have created a category of CPUs that Intel and AMD do not currently have a direct countermeasure for. Karl Freund, founder of Cambrian AI Research, stated that Nvidia’s new CPU is intended for demanding AI tasks, not traditional server functions like website hosting. However, obtaining adoption from major cloud providers remains a challenge.
While Nvidia declined to provide pricing details, Wolfe Research estimated in May that the average selling price for a Vera chip would be around $5,000, projecting that Nvidia would ship approximately 1.3 million units this year. Nvidia reportedly anticipates that the entire server CPU market could eventually reach $200 billion, a significant increase from a Bernstein estimate of $37 billion for the mature server CPU market in 2025.
Regarding market share, Gartner analyst Kevin Knox noted that AMD reportedly commands about 33% of the server CPU market, while Intel holds 66.8%. Knox added that while AMD is gaining market share and maintains strong ties with hyperscale cloud providers, “AMD’s done a great job building their ecosystem around their chips.”
Despite the competitive landscape, Nvidia remains aggressive in its expansion, with reports indicating that OpenAI plans to deploy Vera chips in large quantities starting in the current quarter.
