NVIDIA Vera CPU Challenges AMD and Intel with $20B Server CPU Forecast

Nvidia's Q2 earnings beat sent a clear signal to builders: the company is no longer just a GPU company for AI. CFO Colette Kress guided 70% annual revenue growth in fiscal 2028, well above the 44% the Street expected, and revealed a new stand-alone CPU called Vera designed specifically for running AI agents. Nvidia projects $20 billion in CPU revenue for fiscal 2027, with plans to more than double that in fiscal 2028. That's a direct challenge to AMD and Intel in the server CPU market, and it changes how AI builders should think about infrastructure planning.

Vera CPU targets agent orchestration, not inference

The Vera CPU isn't meant to replace GPUs for heavy inference. Instead, it handles the orchestration layer: coordinating tool calls, managing state between model steps, and running agents continuously. Nvidia claims Vera can complete agentic tasks 1.8x faster than industry-standard CPUs and delivers five times the bandwidth per watt. For builders deploying multi-agent systems or long-running automated workflows, those numbers matter because CPU overhead becomes a bottleneck as agent count scales.

CEO Jensen Huang said on the earnings call that "most AI is now agentic" and predicted companies will eventually run hundreds of thousands of agents continuously. That vision shifts the balance: GPUs still handle model inference, but CPUs become critical for orchestrating agents at scale. If Vera delivers on its claimed performance, Nvidia now offers a tightly integrated CPU-GPU stack for agent workflows, which could reduce the appeal of mixing AMD or Intel CPUs with Nvidia GPUs.

The revenue numbers that worry AMD and Intel

Intel reported $6.3 billion in data-center and AI revenue in Q2, while AMD reported roughly $6.7 billion. Neither company breaks out CPU revenue individually, but Nvidia's $20 billion fiscal 2027 CPU projection already rivals their total data-center segments. And Kress expects that figure to more than double in fiscal 2028. That's a massive swing, especially since Nvidia only recently introduced Vera as a stand-alone product.

To put it in perspective, Bank of America analyst Vivek Arya raised the total addressable market for server CPUs to $210 billion by 2030, driven by AI agents. The market is growing fast, but Nvidia is capturing a disproportionate share of that growth. For AMD and Intel, the silver lining is that agentic AI adoption is still early, and there may be enough demand for all three to grow. But Nvidia's ability to bundle CPUs with its dominant GPU ecosystem gives it a built-in distribution advantage that AMD and Intel lack.

Caveats and what to watch

These claims are Nvidia's own projections and benchmarks. Vera's real-world performance at scale, especially in heterogeneous deployments mixing other vendors' hardware, isn't independently verified yet. The $20 billion CPU revenue forecast is a company guidance, not a guarantee. And the shift from prompted AI to agentic AI, while real, may take years to fully materialize in enterprise environments. Builders should watch for third-party benchmarks, customer adoption reports, and how AMD and Intel respond with their own AI-optimized CPU roadmaps.

For now, the important takeaway is that AI infrastructure planning should account for CPU capacity for agent orchestration, not just GPU compute. If Vera delivers, Nvidia becomes a one-stop shop for both, which simplifies procurement but reduces optionality. If it doesn't, AMD and Intel still have time to defend their turf.