doubleAI launches Maxwell to optimize CPU and GPU execution
The AMW Read
Maxwell introduces a distinct execution-optimization offering with savings-based pricing, but undisclosed benchmark conditions and unverified gains limit evidence of broader impact.
doubleAI launches Maxwell to optimize CPU and GPU execution
Amnon Shashua's doubleAI has introduced Maxwell, its first Code Execution Model, for early access. The company says Maxwell analyzes running programs, hardware and data, then optimizes machine instructions without changing the original source code. It supports CPUs and GPUs, with each optimization formally verified to preserve program behavior, according to doubleAI. The company has raised a $200 million Series A co-led by Lightspeed Venture Partners and Bessemer Venture Partners, with participation from NVIDIA and Dell Technologies Capital, among others.
Maxwell places doubleAI in the AI infrastructure execution layer: its proposed value comes from making existing software run more efficiently. That separates its economic proposition from coding assistants focused on writing and editing software. Charging against delivered computing savings could tie revenue directly to customer outcomes, but it also makes measurement central to the business. doubleAI reports speedups exceeding 38 times for data pipelines and 4.5 times for AI inference kernels; these claims have not been independently verified, and the announcement lacks enough test, hardware and baseline detail to establish how broadly they apply.
For builders evaluating early access, the concrete test is whether Maxwell delivers measurable savings on their own workloads while preserving behavior. Performance comparisons need equivalent hardware, credible baselines and a clear accounting of optimization overhead. Investors should distinguish the reported workload gains from evidence of repeatable customer economics: a savings-based pricing model depends on reliably establishing which reductions Maxwell actually delivers.
