Nvidia secured an agreement with Modal to introduce its Vera central processors into Modal's AI-agent computing infrastructure.
Modal previously relied on Intel and AMD CPUs for these workloads and now plans to introduce Vera alongside its existing processors.
According to Modal's early internal benchmarks, Vera delivered up to twice the task throughput of comparable x86 systems under heavy concurrent workloads. The company has not announced a completed production migration or a confirmed general-availability date. Reuters
Sources: Modal's October 19 engineering announcement · Reuters — October 19
Why it matters strategically
AI infrastructure competition is increasingly extending beyond GPU acceleration.
AI agents require CPUs to execute code, coordinate tools, manage containers, and perform other computational tasks between model interactions.
These activities can create bottlenecks even when GPU capacity is available.
Modal's reported performance advantage concerns throughput under concurrent workloads.
It does not establish that Vera is twice as fast for every computational task.
The distinction matters because infrastructure operators must optimize for actual workload characteristics.
