AI infrastructure startup Infinity has raised $15 million at a $100 million valuation, with backing from Touring Capital, Principal VC, and researchers from OpenAI and Anthropic. The company is developing software designed to let AI chips run models more efficiently, aiming to break Nvidia's grip on the market, which stems largely from its proprietary CUDA software allowing GPUs to run general-purpose AI workloads. Infinity is one of a growing number of startups attempting to erode Nvidia's dominance by building alternative tools that work across different chip types.

Founded last year by ex-Google Brain researcher Jeremy Nixon, Infinity uses an AI research agent called Ignition to automatically write, test and refine the low-level "kernel" code needed to run AI inference on non-Nvidia chips, including SRAM, phone chips and Systolic Arrays. The system is self-optimising and adapts to varying chip architectures, with humans providing high-level oversight while the agent handles repetitive work; Nixon says it can cut processes that once took months down to hours or days. Rather than charging licence fees, Infinity takes a cut of resulting performance and cost improvements. The company, which counts chipmaker D-Matrix among its customers and is in talks with other major chip and cloud firms, currently employs 26 people.