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 on a mission to make Nvidia's stranglehold on the AI chip market a little less strangle-y by building software that lets AI models run on any chip, not just Nvidia's.

Nvidia's dominance isn't just about its powerful GPUs; it's also about CUDA, the software layer that turns graphics processors into general-purpose computing workhorses. Major AI frameworks like PyTorch and TensorFlow are built on CUDA, meaning apps written in Python automatically default to Nvidia chips. Most startups lack the resources to write their own low-level kernel software to port apps to alternative chips. Infinity aims to be the CUDA alternative, creating a universal inference library that works across SRAM, GPUs, phone chips, and Systolic Arrays. It's part of a wave of startups chipping away at Nvidia's market share, one product at a time.

Founded last year by Jeremy Nixon, a former Google Brain researcher and creator of the hacker network AGI House, Infinity was born from Nixon's obsession with "automated invention." He previously invented a machine learning algorithm called Omega, which automatically generated and evaluated new algorithms in a feedback loop. That success led him to hardware, where he believed automated systems could generate the low-level code needed to run chips more effectively.

Infinity's AI research agent, Ignition, writes low-level code for AI inference on non-Nvidia chips. It tests, debugs, measures performance, and rewrites code as needed, continuously learning and adapting to different chip architectures. The result, Nixon claims, is a CUDA-level software stack. Customers include D-Matrix, an AI chip maker and Nvidia challenger, and Infinity is in talks with other big chip and cloud companies.

Humans still provide high-level direction, but Ignition handles the grunt work. In one case study, the agent reduced what could have been a years-long process to hours or days. Infinity doesn't charge an upfront license fee; instead, it takes a cut of performance gains and cost savings, measured in tokens per second. The company currently has 26 employees spanning design, operations, and engineering.