Why Perplexity Is Choosing Nvidia's Vera CPU Over Traditional Intel And AMD Chips

Unlike human users, AI agents work continuously without pauses between tasks. Kupp said Nvidia's Vera processor completed AI coding workloads around 1.5 times faster than conventional CPUs.

"Vera really stood out to us as just like a dead-on fit for a lot of the core workloads that we have," Kupp said, as per Reuters.

Perplexity did not disclose how many CPUs it intends to purchase.

What is Nvidia's Vera CPU?

Nvidia describes Vera as a CPU designed specifically for AI factories, where processors handle tasks surrounding AI models, including code execution, data processing, tool calling, memory management and result verification.

According to the company, Vera is powered by Nvidia's custom Olympus CPU core, which delivers 50% higher instructions per cycle than the previous Grace architecture. It said Vera also provides up to 1.2TB/s memory bandwidth and 3.4TB/s core-to-core bandwidth, enabling all 88 cores to maintain high performance without creating bottlenecks.

Nvidia claims Vera delivers 1.8 times the sustained per-core performance of conventional x86 processors during AI agent workloads.

It also announced that its CPU roadmap will continue with the next-generation Rosa processor based on the Rigel core.

Why are Intel and AMD Under Pressure?

Intel and AMD have dominated the server CPU market for decades, supplying processors for enterprise servers, cloud infrastructure and personal computers.

However, many of these processors were designed before the emergence of AI agents, whose computing patterns differ significantly from traditional software.

Nvidia's strategy extends beyond launching another processor. Instead, it is integrating Vera directly into its AI infrastructure, pairing its own CPUs with its GPUs inside complete server systems, as per a report by Forbes.

According to that report, this changes the economics of AI servers by reducing customers' reliance on standalone Intel Xeon or AMD EPYC processors. Rather than buying CPUs separately, customers increasingly purchase Nvidia-designed systems where the CPU is simply one component within a tightly integrated platform.

Why do AI Agents Need Different Chips?

Nvidia said AI agents operate in continuous reasoning loops rather than responding to occasional user requests.

Each step in an agent's workflow — whether executing code, processing data or verifying results — depends on the previous one finishing quickly. Increasing the number of CPU cores alone does not accelerate these sequential tasks, the company said.

Instead, the company reasoned that higher single-threaded performance, predictable latency and sustained memory bandwidth are more important than simply adding more cores.

This, Nvidia says, allows AI factories to keep expensive GPUs fully utilised instead of waiting for CPUs to complete supporting tasks.

Who Else is Adopting Nvidia CPUs?

Perplexity joins a growing list of AI companies that have committed to Nvidia's CPUs.

Nvidia has previously disclosed that OpenAI, Anthropic and Oracle also plan to deploy Vera processors, as per Reuters.

According to Forbes, OpenAI, Anthropic and SpaceX were among the earliest companies to receive Vera systems, reflecting growing interest in integrated AI infrastructure among companies building large-scale AI models.

Is Nvidia Becoming the Next Intel?

Nvidia is rapidly expanding beyond its traditional AI GPU business, but industry observers believe its strategy is broader than simply replacing Intel or AMD in the CPU market.

The company's objective is to control more of the AI computing stack by offering tightly integrated systems that combine its CPUs, GPUs, networking and software, as per Forbes.

Instead of selling standalone processors, Nvidia is increasingly positioning itself as a full AI infrastructure provider.

At the same time, Nvidia's relationship with Intel reflects both competition and collaboration. The company had announced a $5 billion investment in Intel in September 2025, acquiring roughly a 4% stake in the US chipmaker.

As part of the partnership, Intel is expected to build customised x86 system-on-chips that integrate Nvidia RTX GPU chiplets, while the two companies will collaborate on future computing platforms, as per a report by Fortune.

This means Nvidia is not abandoning the x86 ecosystem altogether. Rather, it is developing its own Arm-based CPUs for AI-focused workloads while also working with Intel on customised processors for specific use cases.

Although Intel and AMD continue to hold strong positions in the server CPU market, Nvidia's push into AI-native processors suggests the competitive landscape is shifting. As AI agents become more common, the battle may increasingly centre on who controls the entire computing platform rather than just the CPU socket.