Tech news in 3 minutes
Infinity raises $15M for inference from Touring Capital, OpenAI, Anthropic
AI infrastructure startup Infinity announced a $15 million funding round at a $100 million valuation on Monday from investors including Touring Capital, Principal VC, and researchers from OpenAI and Anthropic, with the core goal of building a CUDA-alternative software stack that enables AI chips from any vendor to efficiently run AI models. The company addresses a key reason behind Nvidia’s dominance: its CUDA software, which allows GPUs to function as general-purpose processors and is the foundation of frameworks like PyTorch and TensorFlow. Most application-level startups lack the resources to write low-level kernel code to port models to non-Nvidia chips. Infinity’s universal inference library aims to automate that process for any chip type—SRAM, GPUs, phone chips, and systolic arrays. Founded last year by former Google Brain researcher Jeremy Nixon (creator of the AGI House hacker community), Infinity is part of a wave of startups chipping away at Nvidia’s market lead. Its AI research agent, Ignition, writes, tests, and optimizes low-level inference code for alternative chips, self-optimizing across different architectures and claiming to deliver CUDA-level performance. A case study showed the agent reduces development from months or years to hours or days. Human engineers still provide high-level direction. Customers include AI chip maker D-Matrix, and Infinity is in talks with other major chip and cloud companies. Instead of upfront license fees, the startup charges a cut of performance gains and cost savings measured in tokens per second. Infinity currently has 26 employees across design, operations, and engineering.