Modal Adds Nvidia Vera CPU to Host AI Agents, Claims 2x Throughput

Modal said it will add Nvidia's Vera central processor to the mix of CPUs it uses to host AI agents, after early tests showed Vera running up to twice the throughput of comparable x86 hardware under heavy concurrent load. Until now the infrastructure startup had run agent work only on Intel and AMD silicon, Reuters reported, and it plans to keep both vendors in rotation while routing jobs to whichever chip runs them best.

Modal CEO Erik Bernhardsson said the company tested Vera against Intel and AMD parts already in production. Running a mixed setup, he told the news agency, means Modal can "always deploy whatever hardware performs best." On future Nvidia parts he was similarly non-committal: "We'll test them when they're available," he said, framing Vera as an additional lane rather than a replacement.

Modal raised $750 million last week at a $15.75 billion valuation in a round led by Accel, saying usage has tripled and revenue has grown more than 15 times year-over-year, enough to put it on track to record $1 billion in sales over the next year. Vera is Nvidia's new Arm-compatible CPU, built for the Vera Rubin GPU platform but also offered as a standalone CPU server. Nvidia has told investors it expects at least $20 billion from Vera this fiscal year in a market it values at roughly $200 billion.

The 2x throughput figure is Modal's own early-test number, not an independent benchmark. The news arrives in a week already thick with alt-silicon moves on our AI infrastructure desk, from Tenstorrent's $1.5B Series E to Amazon's TorchNeuron release.