Infinity Startup Raises $15M to Build Universal AI Chip Software and Challenge Big Tech Dominance

A promising new AI company called Infinity has secured $15 million in seed funding to develop software that helps different chips run advanced AI inference models smoothly.

The round valued the startup at a $100 million post-money valuation and drew interest from Touring Capital along with researchers from leading AI labs, including OpenAI and Anthropic.

Challenging Nvidia's Dominance in AI Infrastructure

Nvidia has long led the field thanks to its CUDA software platform, which makes it easier for developers to build AI applications that run efficiently on Nvidia chips.

Infinity wants to create a similar easy-to-use system that works across many chip types, including those from rival hardware makers.

This approach could give smaller teams more hardware options and help spread AI development beyond one dominant supplier.

Founder Jeremy Nixon, a former Google Brain researcher, now uses AI agents to automatically create and test the complex code needed for efficient model inference.

The company's Ignition tool writes low-level instructions for chips, then measures and improves performance in a smart optimization loop.

How Efficient AI Inference Could Transform Everyday Technology

One early customer is chipmaker d-Matrix, which benefits from faster AI inference on its hardware.

The business model shares in the savings from better speed and lower costs rather than charging upfront fees.

With only 26 team members, the startup focuses on self-improving systems that automatically adapt to new chip designs.

Such technology could eventually allow advanced AI to run more efficiently on phones and edge devices, enabling more capable AI applications for everyday users.

If adopted widely, the technology could also improve AI efficiency while reducing computing costs and energy use in large data centers as AI continues to grow in scale and importance.