Key Highlights

  • Modal plans to run Nvidia’s Vera processor for AI agents after relying only on Intel and AMD chips.
  • Early tests showed the Nvidia chip doubled the work AI agents complete versus Intel and AMD chips on the market.
  • Modal raised USD 750 million last week at a USD 15.75 billion valuation and keeps a mixed chip fleet.
  • Nvidia has said it expects at least USD 20 billion of Vera sales this fiscal year.

A Crack in a Long-Held Market

Nvidia (NASDAQ:NVDA) has won central processor chip business from another fast-growing AI startup that until now relied entirely on chips from Intel (NASDAQ:INTC) and AMD (NASDAQ:AMD). It is an early sign that Nvidia’s push into a market long controlled by those two rivals has traction.

Modal, a cloud-computing startup that rents out computing power to AI developers, said it plans to start using Nvidia’s new Vera central processor for the first time for hosting agents, the AI systems that carry out long-running tasks for human users. Modal has so far run such work only on Intel and AMD chips.

Test Results and a Mixed Fleet

Modal said early tests showed the Nvidia chip doubled the amount of work AI agents get done. The result prompted the company to support Nvidia’s chip design for the first time, which requires some deep software changes on Modal’s side.

Chief Executive Erik Bernhardsson stressed that Modal is not abandoning its existing suppliers. The company runs the Nvidia chips alongside its current Intel and AMD offerings in a mixed fleet, placing each job on whatever hardware runs it best.

Bernhardsson said the tests compared Vera with the Intel and AMD chips already on the market, and that both rivals are developing newer models aimed at narrowing the gap with Nvidia. “We’ll test them when they’re available,” he said, adding that a mixed setup means the company “can always deploy whatever hardware performs best.”

The Funding Behind the Move

Modal last week raised USD 750 million in a funding round led by Accel at a USD 15.75 billion valuation. The company puts its sales at USD 1 billion over the next year. It said usage of its services has tripled and revenue has grown more than 15 times year over year.

Nvidia’s Processor Push

Nvidia has said it expects to generate at least USD 20 billion from Vera chip sales this fiscal year, its opening bid for a market it values at roughly USD 200 billion. It has previously said OpenAI, Anthropic and Oracle (NYSE:ORCL) also plan to use the chips, along with AI startups such as Perplexity.

Vera entered full production in spring 2026, with first deliveries to OpenAI, Anthropic and SpaceX (NASDAQ:SPCX) in June. Nvidia disclosed the technical details on July 21 and 22, and a dedicated Vera rack sits beside its NVL72 system as a separate machine for agent workloads. It replaces the earlier Grace chip as Nvidia’s flagship central processor and is the first server processor Nvidia developed largely in-house. Nvidia says it delivers up to 50 percent better performance on AI agent workloads than x86 processors from Intel and AMD, and more than twice the performance of Grace on agentic workloads. Those claims lack independent verification.

Where the Incumbents Stand

Intel holds nearly two-thirds of the server processor market and AMD roughly one-third, with AMD gaining ground. Nvidia sells Vera as a standalone chip, in dual-processor servers, in liquid-cooled racks of 256 processors that offer over 20,000 cores for agents, and as part of Vera Rubin systems alongside Rubin graphics processors.

Nvidia says the design emphasizes single-core performance over core count, built to remove bottlenecks that slow AI agents. Meta Platforms (NASDAQ:META) and CoreWeave (NASDAQ:CRWV) are named as early customers, and Oracle is the only major cloud provider listed among launch partners. OpenAI plans to begin deploying Vera at scale this quarter, though Nvidia says adoption is still early.