DoubleAI unveils Maxwell AI model to slash GPU and CPU compute costs
DoubleAI introduced Maxwell, a code-execution AI model that promises to reduce hardware expenses, and will bill clients based on the savings achieved.
DoubleAI, backed by Nvidia and other investors, announced Maxwell, a code execution model that operates inside live programs to improve performance and lower compute spend. Co-founder Amnon Shashua said the key breakthrough was a formal, mathematical verification of code changes. Early adopters such as LTX reported that Maxwell doubled inference speed for their own AI model, translating into measurable GPU cost reductions.
The company will invoice customers a share of the savings, though the exact percentage was not disclosed. Shashua expects major AI developers with high training and inference bills, like OpenAI and Anthropic, to benefit most, potentially saving hundreds of millions. DoubleAI has raised $220 million and allocated half of that to purchase 1,000 Nvidia B300 chips for its own infrastructure.
Why it matters
Maxwell could dramatically lower the operating costs of AI workloads, affecting the economics of large-scale machine-learning projects.
In this story
