Amnon Shashua, founder of Mobileye and one of the most prominent entrepreneurs in Israeli tech, is unveiling a new venture. doubleAI, a Tel Aviv company he co-founded, has emerged from stealth with $220 million in total funding, according to a Wall Street Journal report on October 19. Of that, $200 million is a Series A co-led by Lightspeed and Bessemer. Nvidia and Dell are also among the investors.
What Maxwell does
The company's first product is called Maxwell, which doubleAI describes as a "code execution model". Most AI developer tools help write code. Maxwell is meant to work at a different stage: while the software is already running. According to the company, it rewrites the machine code to fit the hardware and the data it actually encounters, without changing the source code the developers wrote.
The company's claims are ambitious. According to doubleAI, one early customer cut its compute costs by 67 percent, and on workloads using pandas, the popular Python data analysis library, it measured a 38.4x speedup. These figures were published by the company itself and have not yet been independently verified.
Pay only for the savings
doubleAI's business model differs from a standard software subscription: the company plans to charge a share of the savings a customer achieves. Shashua illustrated it with a compute bill falling from $10 million to $6 million, with doubleAI taking part of the $4 million saved. He did not specify the fee rate. For the customer the offer lowers the risk: no savings, no payment. For the company, revenue grows as customers grow.
According to reports, about half of the money raised went on hardware, buying around 1,000 Nvidia B300 chips for the company's own computing needs. It is a telling detail about the economics of AI companies in 2026: even a young startup needs the computing power of a mid-sized company just to get started. The company is about two years old and has around 40 employees.
Why now
The timing is no accident. The big AI labs, and companies running models at scale, are spending unprecedented sums on infrastructure. Every percentage point saved on compute is worth a lot of money. Shashua himself said he expects labs such as OpenAI and Anthropic to benefit most. If Maxwell delivers on its promises, doubleAI could find itself in a very comfortable position in a market where everyone is looking for ways to spend less on power and chips.
The story adds to a strong year for Israeli tech, in which mega-rounds in cybersecurity and AI are leading the funding tables. The big question about doubleAI is the same one facing any promising technology: will the results the company has shown hold up across many customers, varied workloads and external testing?
