Key Takeaways:
Perplexity will run its agentic AI workloads on Nvidia's new Vera CPU, handing the chipmaker one of its first major design wins for its push beyond accelerators.
Perplexity plans to deploy Nvidia's Vera CPU for agentic AI workloads, the AI search startup said Wednesday, handing the chipmaker a marquee customer for its push into general-purpose processors that has already signed Anthropic and OpenAI.
"Vera's single-threaded performance is a dead-on fit for the agentic coding tasks that chain together dozens of model calls," Nate Kupp, vice-president of enterprise infrastructure at Perplexity, said.
Vera, the processor half of Nvidia's Vera Rubin platform, is a custom Arm design pairing dozens of Nvidia-built cores with fast, low-power memory and a high-bandwidth link to the company's accelerators. Perplexity said Vera ran its agentic coding tasks about 1.5 times faster than the traditional CPUs it had been using. The company now handles more than 400 million search queries a month, each running through an inference pipeline that already relies on Nvidia's H100 GPUs and its Triton and TensorRT software.
The deal validates Nvidia's bet that its CPU line can generate roughly $20 billion in sales this fiscal year, its opening bid for a general-purpose computing market the company sizes at about $200 billion. Nvidia shares fell 3.2% to $118.40 in Wednesday trading, extending a recent rut as broader market concerns about valuation and sector headwinds outweighed the positive catalyst.
Why Vera matters for agentic workloads
When an AI system chains together many model calls, tool uses and code runs to complete a task, the general-purpose processor coordinating all of it becomes a bottleneck as real as the accelerator doing the heavy mathematics. Nvidia's pitch is that a CPU designed by the same company making the GPUs can move data between the two with less friction than an off-the-shelf part from Intel or AMD.
Perplexity's choice also signals where AI search economics are heading. As rivals fold generative answers into everything, the cost of running those answers at scale has become a competitive problem. Faster CPUs let a company serve more queries for the same spend, or push into agentic features that string together many calls without the bill spiraling. For Perplexity, which has grown fast and raised heavily to fund that growth, holding down the cost per query is close to an existential question.
The competitive landscape
Nvidia's CPU push is partly defensive. Several of its largest customers, OpenAI among them, are designing their own AI chips, and a CPU line lets Nvidia sell more of every rack even where its accelerators face competition. The company has also been offering startups compute now and payment later to lock in demand.
The wider ecosystem around Vera Rubin is filling in quickly. On the memory side, Nvidia and SK Hynix have sealed a multi-year HBM4 deal to supply the platform, one of several supply agreements underpinning the roadmap. Vera itself is built on a custom Arm architecture, positioning Nvidia against Intel and AMD in a server CPU market those two companies have dominated for decades.
Neither Nvidia nor Perplexity has said when Vera ships in volume, and Perplexity has not detailed how much of its fleet will move across. The harder test comes later, when Vera has to win over enterprises that already know and trust Intel and AMD in a way they do not yet know Nvidia's CPUs.
Nvidia shares, trading at about 32 times forward earnings, have fallen roughly 8% over the past month as investors weigh the company's dominance in AI accelerators against rising competition and a potential slowdown in data center spending. Morgan Stanley's Joseph Moore maintained his $165 price target on the stock, calling the Vera CPU opportunity "a long-term revenue diversifier rather than a near-term earnings driver."
This article is for informational purposes only and does not constitute investment advice.
