doubleAI

AI Infrastructure · Software Optimization · GPU Compute

Launch

doubleAI launches Maxwell, an AI that rewrites code while it runs

The system reportedly cut one early customer's compute bill by 67%, reframing GPU spending as a software problem rather than a hardware one.

• doubleAI unveiled Maxwell, which it calls the first 'Code Execution Model,' an AI that lives inside running software and rewrites machine-level instructions on the fly without altering source code.
• Founded in 2023 by outgoing Mobileye CEO Amnon Shashua with co-founders Shai Shalev-Shwartz and Gal Beniamini, the Tel Aviv startup spent years in stealth before this launch.
• Maxwell formally and mathematically proves that each optimization preserves a program's original behavior, working across both CPUs and GPUs much like a just-in-time compiler but with higher-level reasoning power.
• On real-world workloads, doubleAI says Maxwell made data pipelines more than 38 times faster and AI inference kernels over 4.5 times faster, while cutting one early customer's compute costs by 67%.
• Early customer LTX told the Wall Street Journal that Maxwell doubled its model-inference speed, an outside signal for claims doubleAI has not yet validated across its full customer base.
• doubleAI has raised $200 million in a Series A co-led by Lightspeed Venture Partners and Bessemer Venture Partners, with Nvidia and Dell Technologies Capital among backers, in a round reportedly valuing the company above $1 billion.
• By putting AI to work for the hardware rather than the programmer, doubleAI is betting that software efficiency, not just more chips, can become a sellable alternative to buying additional compute.