Infinity

Infinity is a San Francisco-based AI infrastructure startup building the software layer that makes any AI chip ready to run inference. Founded in 2025 by Jeremy Nixon (previously a researcher at Google Brain and creator of the AGI House hacker community), the company develops a generative optimization engine that automatically writes and improves the low-level compute kernels that determine how efficiently a chip runs AI models. Its autonomous AI research agent, Ignition, produces the kernel code needed to run inference on Nvidia-alternative chips, aiming to build a CUDA-alternative, universal inference library that works across hardware types including GPUs, SRAM-based accelerators, phone chips and systolic arrays. Infinity is part of a wave of startups chipping away at Nvidia's data-center dominance by making rival chips easier to deploy for production-grade inference. Customers include AI chip maker d-Matrix, and the company is in talks with other major chip and cloud companies.