Modal said Oct. 19, 2026, that it plans to support Nvidia’s Vera processors for its AI-agent sandboxes, adding another hardware option as cloud operators look for ways to handle more simultaneous computing tasks.
The planned move would extend Modal’s infrastructure beyond its existing x86 systems. Vera uses an Arm-compatible architecture, requiring changes to Modal’s infrastructure and software-image building system, according to the company’s engineering announcement.
Company tests focus on crowded servers
Modal reported up to twice the task throughput of an equivalent Intel Granite Rapids system in its early testing under load. Those are company-generated results, not an independent performance assessment or a promise that every workload will run twice as fast.
Its benchmark used 50 Terminal-Bench tasks with network calls removed. The tests examined work such as compilation and file operations, rather than the entire performance of an AI agent. Modal said results on a quiet server were less uniformly favorable than tests with competing tasks.
A mixed fleet, not a replacement
Chief Executive Erik Bernhardsson told Reuters that Modal would retain Intel and AMD processors alongside Nvidia’s chips and assign work to hardware suited to each job. He also said newer competing processors would be tested when available.
The announcement did not establish a public rollout date. For customers, the distinction is important: a hardware plan and an internal benchmark do not yet establish service availability, pricing or performance for their own applications.