Nvidia Expands AI Strategy into Central Processing Units
Nvidia, already renowned for its Graphics Processing Unit (GPU)—the foundational chip for artificial intelligence—is now venturing into the central processing unit (CPU) market with its proprietary chip, named Vera. This move establishes a new competitive frontier in AI servers, directly challenging established industry leaders AMD and Intel. The company has released detailed specifications and architectural information to help potential customers fully evaluate the capabilities of the Vera CPU.
The chip giant has been the most valuable company largely due to the intense global demand for its GPUs. However, by beginning to ship its own CPUs, Nvidia is offering cloud service providers an alternative to products from Advanced Micro Devices and Intel. On Tuesday, Nvidia provided new details regarding its data center CPU, Vera. Furthermore, Nvidia representatives confirmed that the Vera chips were delivered to key clients, including OpenAI, Anthropic, and SpaceX, in June.
A Strategy of System Integration
Nvidia has historically been a company that dictates the direction of the information technology industry. While its entry into the CPU space positions it as a challenger against the deeply entrenched players, Intel and AMD, this strategy exemplifies Nvidia’s commitment to vertically integrating its systems. The goal is to sell entire computing racks—a complete system of power and technology—rather than merely selling individual chips. According to Nvidia, this approach will enable engineers to maximize the performance of their GPUs, helping the company maintain its position as the preferred choice for leading AI research laboratories amidst rising competition from custom chips and other players.
Historically, before the current AI surge, the CPU was the single most crucial and expensive component in a server. The initial wave of AI servers, available when ChatGPT was released in 2022, often paired up to eight GPUs with a single CPU, indicating a significant shift toward Nvidia’s GPU dominance. However, the recent rise of “agentic AI”—systems that can operate autonomously with minimal human oversight—has renewed focus on the CPU. These agents require the CPU to manage and feed data efficiently.
Performance Metrics and Industry Outlook
Financial markets have reacted strongly to this shift, with CPU incumbents AMD and Intel being among the best-performing chip stocks in 2026, having risen 128% and 149% respectively, compared to Nvidia’s rise of 8%. Speaking at a recent presentation, Nvidia’s vice president of hyperscale, Ian Buck, highlighted that agents have made CPUs “much more integral,” specifically noting “how fast a CPU can answer one question.”
Regarding market valuation, Nvidia projects the entire server CPU market could eventually reach $200 billion. This contrasts with a previous Bernstein estimate from earlier this year that valued the total mature server CPU market at approximately $37 billion in 2025. Furthermore, Wolfe Research forecasted in May that the average selling price for a Vera chip would be around $5,000, predicting Nvidia would ship about 1.3 million units of the chip this year. Nvidia declined to provide a comment regarding pricing.
Industry analysts have offered varying opinions on the competition. Kevin Knox, an analyst at Gartner, suggested that AMD is currently the leading contender in enterprise AI server CPUs. Knox noted that while Intel reportedly holds 66.8% of the server CPU market and AMD holds about 33%, AMD is successfully increasing its market share and maintains strong relationships with major cloud providers. Knox stated, “AMD’s done a great job building their ecosystem around their chips.”
Vera Chip Specifications and Capabilities
Nvidia announced that Vera is the first server CPU the company has designed entirely from its core, rather than relying on an existing design provided by Arm, which would reduce engineering workload. The chip was specifically engineered to mitigate bottlenecks encountered by AI agents. Nvidia stated that the Vera chip offers 50% better performance for AI agents compared to x86 chips, the architecture used by Intel and AMD. This performance gain is attributed to the fact that previous CPUs from Intel and AMD prioritized core count, while Nvidia’s chip, alongside the Olympus core, focuses on single-core speed.
In a presentation, Hannah Coutand, a Vera product marketer at Nvidia, explained that the chip’s design emphasizes high memory bandwidth, per-core speed, and low latency. This focus is intended to ensure that agents can return to the GPUs as rapidly as possible, thereby maximizing the utilization of the GPUs—which are recognized as a highly valuable and expensive asset in the AI sector. The chip is designed to be sold individually, but it will also be available in a liquid-cooled rack containing 256 Vera chips, as well as a configuration pairing two Vera chips within a single server, and paired with Nvidia’s GPUs in a system called Vera Rubin.
Technically, the CPU is designed to be power-intensive, operating between 250 watts and 450 watts. It also incorporates a substantial amount of low-power memory, similar to that found in laptops and mobile phones, supporting up to 1.5 terabytes of memory per chip.
Market Adoption and Challenges
Some experts suggest that Nvidia has created a unique category of CPUs that Intel and AMD currently lack a direct counter-product for. Karl Freund, founder of Cambrian AI Research, indicated that Nvidia’s new CPU is intended for highly intensive AI tasks, rather than traditional server functions like running websites. Despite this advanced technology, Coutand cautioned that the adoption of Vera is still in its “early innings.” While Nvidia listed Oracle as a partner, it did not name any major cloud service providers. However, the company indicated that OpenAI plans to deploy Vera chips in large quantities starting in the current quarter. Freund observed that “The CPU is something they’ve done to kind of unhook their customers from using Intel or AMD CPUs, and they covet that revenue.” He added, “What they’ve done is they decided to focus on a unique CPU that isn’t available in the market from anyone right now.”
