DoubleAI launches Maxwell, claiming a 67% cut in one early customer's compute spending

Quartz, citing The Wall Street Journal, reports that Nvidia-backed DoubleAI launched Maxwell, which the startup says lowered one early customer's compute spending by 67%. Quartz also reports that LTX found Maxwell made inference for its own model twice as fast. DoubleAI plans a broader release before the end of October 2026 and intends to charge customers a share of their compute savings; CEO Amnon Shashua did not specify the percentage.