Modal says Nvidia’s Vera CPU doubles throughput for its AI sandboxes

Modal plans to run its Sandboxes product on Nvidia’s Arm-based Vera CPU, marking its first move away from x86 chips. In Modal’s own tests, Vera handled up to twice as many tasks as an Intel server when packed with containers; the results have not been independently verified.

Analysis

The investable signal is architectural optionality, not yet evidence of a broad CPU share shift. If dense, short-lived agent workloads achieve materially better throughput per server on Arm, operators could defer x86 purchases and weaken Intel’s pricing leverage in a workload category with potentially high growth. But Modal’s in-house result is not independently verified; migration friction, software compatibility, and total cost of ownership could erase the apparent throughput advantage. The scope is one workload and one operator, not a read-through to Intel’s installed base or consolidated results.

For NVIDIA, a successful CPU deployment could strengthen its ability to sell a fuller AI-infrastructure stack and make its platform more relevant beyond GPU-bound compute. The counterpoint: CPU adoption without associated GPU or networking demand may have limited economic value, and wider Arm adoption could also benefit other Arm ecosystem participants rather than accrue uniquely to NVIDIA.

Near term, expect limited fundamental impact absent customer commitments or disclosed deployment scale. Over 1–3 months, watch independent benchmarks, production adoption, and workload economics. Over 6–18 months, repeated evidence of lower cost per sandbox could affect server procurement and x86 refresh decisions. Thesis weakens if production throughput or total cost fails to match the vendor test, or if adoption remains isolated.

Key Decisions for Investors

• No immediate directional trade: the evidence is a single vendor’s unverified test, with no disclosed deployment scale or financial contribution.
• Treat this as a watch item for INTC rather than a standalone short catalyst. Reassess if multiple operators report Arm migration at scale or Intel identifies measurable pressure on server demand, pricing, or guidance.
• NVDA is a conditional beneficiary if Vera deployments pull through demand for its broader AI infrastructure; do not assume CPU adoption alone produces meaningful incremental earnings.
• Falsification checks: independent workload benchmarks, production utilization and cost per sandbox, software-porting friction, and any company guidance or customer evidence tying Arm adoption to x86 procurement changes.