Infinity Raises $15M to Build a Universal CUDA Alternative for AI Inference

Infinity, an AI infrastructure startup founded by former Google Brain researcher Jeremy Nixon, has raised $15 million at a $100 million valuation. The round included Touring Capital, Principal VC, and individual researchers from OpenAI and Anthropic. The company is tackling one of the most entrenched moats in AI hardware: Nvidia’s CUDA software stack.

Nvidia’s dominance isn’t just about raw chip performance — CUDA’s tight integration with PyTorch and TensorFlow means most AI applications effectively require Nvidia hardware. Infinity wants to break that lock-in with an AI agent called Ignition, which automatically writes, tests, and optimizes low-level kernel code for any chip architecture, including SRAM, GPUs, phone chips, and systolic arrays.

According to Nixon, the agent can reduce what normally takes months or years of manual kernel engineering to hours or days. Humans still provide high-level direction, but the system handles the tedious grunt work and continuously self-optimizes. Infinity doesn’t charge upfront license fees; instead, it takes a cut of performance gains measured in tokens per second.

The startup already counts D-Matrix, a would-be Nvidia challenger, as a customer and is in talks with other chip and cloud companies. With 26 employees, Infinity is part of a growing wave of startups chipping away at Nvidia’s software advantage, one kernel at a time.