doubleAI launches Maxwell code execution model in early access

doubleAI describes Code Execution Models, or CEMs, as a new AI category that places intelligence between code and silicon.[1] Its first CEM, Maxwell, is available for early access.[1] The company says a CEM acts on behalf of the hardware, while Maxwell continuously analyzes a running program, the hardware it uses and the data it processes.[1][2] doubleAI says the source code stays exactly as it was, and that Maxwell formally and mathematically verifies each optimization to preserve the program’s behavior.[1]

doubleAI says CEMs can get more out of CPU or GPU hardware.[1] The company reports that Maxwell made data pipelines more than 38 times faster on real-world workloads and that AI inference kernels were over 4.5 times faster.[1] Maxwell’s product site lists a separate 4.25-times-faster figure for AI inference kernels.[3] CTech reported that the performance figures had not been independently verified and that the company had not provided enough test detail to assess how broadly they could be replicated.[2] Professionals evaluating the claims should distinguish reported results from independently validated performance.

Maxwell’s product site says a 4K generation was nearly twice as fast: a clip took 134 seconds on two B200s, down from 250 seconds.[3] Separately, the Wall Street Journal reported that doubleAI said it cut one early customer’s compute costs by 67%.[4] CTech reported that the company plans to charge customers based on the computing-cost savings it delivers.[2] These examples point to a cost-and-performance proposition, but the evidence does not establish how results vary across workloads or what savings a prospective customer should expect.

The Wall Street Journal reported that doubleAI had raised $220 million in total.[4] Its launch announcement described a $200 million Series A co-led by Lightspeed Ventures and Bessemer Venture Partners.[1] Techmeme reported that the company used about half of its funding to buy 1,000 Nvidia B300 chips.[5] The launch announcement also named Dell Technologies Capital, Greenfield, NVIDIA, Pitango and BRM among its investors.[1] CTech reported that doubleAI employed approximately 40 people.[2]