AI infrastructure startup Infinity has secured $15 million in new funding at a $100 million valuation, with support from Touring Capital, Principal VC, and researchers connected to OpenAI and Anthropic.
The company is developing software designed to help AI chips run models more efficiently across different hardware environments. Its goal is to create a flexible inference layer that can work beyond a single chip ecosystem, offering an alternative approach to the software stack that has long supported Nvidia's dominance.
Infinity's platform focuses on low-level kernel code, the technical layer that helps hardware perform AI inference tasks. By automating this process, the startup aims to make it easier to adapt models to GPUs, phone chips, SRAM-based systems, and systolic arrays without requiring teams to manually rewrite code for each architecture.
Founded by Jeremy Nixon, formerly of Google Brain, Infinity is also building an AI research agent called Ignition. The system is designed to write, test, debug, and refine code on its own, while humans provide strategic direction. According to the company, this approach can reduce optimization work from months or weeks to hours or days.
The startup says customers already include AI chip maker D-Matrix, and it is in discussions with other chip and cloud companies. Infinity currently has 26 employees across engineering, design, and operations.
As AI hardware continues to diversify, software that can unify performance across chips may become a key layer in the next generation of computing.
