Infinity Eyes Nvidia's Moat With $15M Raise

AI infrastructure startup Infinity has raised $15 million at a $100 million valuation, TechCrunch reports. The funding round included participation from Principal VC, Touring Capital, and researchers from Anthropic and OpenAI.

Founded last year by former Google Brain researcher and AGI House creator Jeremy Nixon, Infinity is targeting the technical bottleneck that has cemented Nvidia's market dominance: CUDA. While Nvidia's high-performance chips are a primary draw, TechCrunch notes that the CUDA software stack allows developers to run apps on Nvidia GPUs without needing the resources to write their own low-level kernels.

Infinity is developing a universal inference library designed to work across various hardware, including GPUs, phone chips, Systolic Arrays, and SRAM. The company utilizes an AI research agent called Ignition to automate the writing, debugging, and optimization of this low-level code. The system is self-optimizing, Nixon says, and in one case study the startup found it reduced what could have been a years- or months-long process to hours or days.

Infinity currently employs 26 people across engineering, operations, and design. The company does not charge upfront licensing fees; instead, it earns a percentage of cost savings and performance improvements, which are measured in tokens per second. Among its current customers is AI chip maker D-Matrix, and Nixon told TechCrunch the company is currently in discussions with other cloud and chip firms.