Infinity, an AI infrastructure startup building software that lets AI models run on chips other than Nvidia’s, raised $15M at a $100M valuation from Touring Capital, Principal VC, and researchers at OpenAI and Anthropic. Nvidia’s dominance rests partly on CUDA, the software layer that PyTorch and TensorFlow were built on, and most application startups lack the resources to write their own kernels for rival hardware. Infinity’s research agent Ignition writes that low-level code automatically, then tests, benchmarks, and rewrites it to squeeze more speed out of whatever architecture it lands on. Chipmaker D-Matrix is already a customer, and founder Jeremy Nixon, a former Google Brain researcher, told TechCrunch the company charges no license fee and instead takes a cut of the performance gains it delivers.
