AI infrastructure startup Modal says it plans to start using Nvidia's Vera CPUs for the first time for hosting agents, alongside its current Intel and AMD chips
Nvidia … These are the main chips used to run AI workloads, and as a result, Nvidia continues to see surging growth.Jared Blikre /Yahoo Finance:Micron, Intel lead chip recovery, adding nearly $250B as most of Wall Street falls: Chart of the DayDavid Moadel /24/7 Wall St.:Intel and AMD Rise 4% as JPMorgan Upgrades INTC After Steep Pullback; NVIDIA Edges Higher
Context & Ripple Effects
Nvidia’s Vera CPUs had already been positioned with early large users, and initial Vera Rubin benchmarks put Nvidia’s CPU design ahead of Intel and AMD x86_64 chips. Modal’s plan adds an AI-infrastructure operator to that adoption arc while retaining Intel and AMD in its fleet.
The move lands as CPUs have become a contested layer of AI infrastructure, not merely a host for accelerators. It tests whether Nvidia can extend its AI platform deeper into workloads such as agent hosting rather than rely solely on GPU demand.
First-order effects
- Modal will add Vera CPUs for agent hosting alongside its existing Intel and AMD infrastructure, giving it a third CPU option for that service.
- Nvidia gains a named infrastructure customer for Vera CPUs, broadening the product’s adoption beyond the early users cited in its June rollout.
Second-order effects
- Intel and AMD must compete to retain Modal workloads on CPU performance and operational fit as Nvidia enters the same infrastructure accounts with its own CPU architecture.
- AI-capacity operators evaluating agent-hosting hardware gain another non-x86 option, making mixed CPU fleets a more practical procurement model.
Third-order effects
- If more operators adopt Vera while preserving x86 capacity, AI infrastructure is likely to become more heterogeneous, with CPU choice increasingly tied to AI workload design rather than legacy server standards.
- Nvidia’s ability to pair CPUs with its wider AI hardware stack would intensify pressure on Intel and AMD to differentiate at the platform level, not only on standalone processor specifications.
The trend: AI infrastructure is shifting toward heterogeneous CPU fleets as operators match agent and inference workloads to architectures beyond traditional x86 servers.