Modal to Deploy Nvidia Vera CPUs for AI Agents

On Oct. 19, cloud compute startup Modal, which rents GPU capacity to AI developers, announced plans to adopt Nvidia's new Vera CPUs to host AI agents — long-running AI systems that execute tasks on behalf of humans. Previously, Modal ran such workloads only on Intel and AMD chips.

The company said early tests showed the Nvidia chip doubled throughput for AI agent tasks, prompting it to integrate Nvidia for the first time, which required deep changes to its software stack. Modal said it will run a hybrid cluster, deploying Nvidia chips alongside existing Intel and AMD hardware and scheduling tasks to the best-suited processors. CEO Erik Bernhardsson stressed the company is not abandoning current suppliers.

Modal completed a $750 million funding round led by Accel last week, at a $15.75 billion valuation. The company said it expects revenue of about $1 billion in 2027, with usage up threefold and year-over-year revenue growth of more than 15x. Nvidia has said OpenAI, Anthropic, Oracle and Perplexity also plan to use the chip.