Infinity lands $15M to make AI chips easier to use

Infinity, an AI infrastructure startup trying to loosen Nvidia’s grip on the market, said Monday it has raised $15 million at a $100 million valuation. Backers include Touring Capital, Principal VC, and researchers from companies including OpenAI and Anthropic.

The company is building software designed to make it easier for AI chips to run AI models. A major part of Nvidia’s dominance comes not just from its hardware, but from CUDA — Compute Unified Device Architecture — the software layer that lets GPUs act as general-purpose processors. Major AI frameworks including PyTorch and TensorFlow sit on top of CUDA, which means developers can write applications in languages like Python and have them run by default on Nvidia chips.

Infinity wants to offer an alternative at the kernel level, where most application startups lack the expertise or resources to write their own low-level chip software. Its goal is a universal inference library that can work across different hardware types, including SRAM, GPUs, phone chips, and systolic arrays.

Recommended reading

World Cup 2026 Is Stress-Testing Supply Chains

The startup was launched last year by Jeremy Nixon, a former Google Brain researcher and creator of the hacker community AGI House. Nixon told TechCrunch he started the company out of an obsession with “automated invention” — the idea that “AI systems can actually be a meta technology.” He said he previously created a machine learning algorithm called Omega, which generated new machine learning algorithms and evaluated them in a feedback loop.

Infinity’s AI research agent, Ignition, is built to write the low-level inference code needed for Nvidia alternatives. According to Nixon, it tests, debugs, benchmarks, and rewrites code automatically to improve hardware performance, while adapting to different chip architectures, including proprietary ones. The company says the result is a CUDA-level software stack.

Infinity says customers include AI chip company D-Matrix, and Nixon said it is also in talks with other major chip and cloud companies. The company keeps humans in the loop for high-level guidance, but says the agent can shrink work that might otherwise take months or years down to hours or days.

Instead of charging an upfront license fee, Infinity takes a share of performance gains and cost savings, measured in tokens per second. The company currently has 26 employees across design, operations, and engineering.