Perplexity says it plans to use Nvidia’s new CPUs for AI agent workloads

Perplexity, the AI startup behind the Perplexity Search product, said Tuesday that it plans to use Nvidia’s new central processing units, or CPUs, for its work. The update positions Nvidia not only as a supplier of graphics processing units that power most large AI systems, but also as a contender for general compute inside AI applications, including “agent” style tasks.

The company’s statement, reported by Yahoo Finance, did not spell out which specific CPU model Perplexity expects to adopt, when the switch would occur, or whether the plan involves dedicated hardware access or a broader purchasing relationship. It also did not provide performance targets or cost disclosures tied to the change.

’s coverage framed the move as part of Perplexity’s broader push toward AI agents, in which software takes actions or orchestrates steps to complete user goals rather than only generating answers. Using Nvidia CPUs for those workloads suggests Perplexity sees value in running portions of its pipeline on the same vendor’s compute stack, potentially alongside existing GPU-based infrastructure.

For Nvidia, the Perplexity disclosure echoes a theme the company has pursued publicly over the past year: widening its market beyond accelerators by offering chips that can serve more roles in data centers and AI systems. Competing on CPUs matters because entrenched players, including Intel, have longstanding relationships in server CPUs, and data center buyers often standardize across hardware platforms.

The latest news also highlights how quickly AI startups are turning from “model-first” deployment choices toward full-stack compute decisions. As agentic systems mature, companies typically need more control over latency, scheduling, and workload partitioning across different kinds of compute, which can raise demand for CPU resources even in GPU-heavy environments.

Still, major details remain unclear. Neither the Yahoo Finance report nor the related market coverage described contract terms, pricing, or the specific “new” Nvidia CPU referenced. The timeline for ramping production use, the expected proportions of CPU versus GPU work, and any customization Perplexity may require were not included in the accessible reporting.

What to watch next is whether Perplexity’s plan becomes a broader announcement from other AI developers, and whether Nvidia’s forthcoming CPUs are named in additional customer announcements. Industry observers will also look for technical disclosures that indicate how these CPUs are used in agent workflows, such as whether they handle orchestration, pre/post-processing, or inference-side components alongside GPUs.