AI infrastructure startup Infinity announced in July 2026 that it has raised $15 million at a $100 million valuation. The round drew investment from Touring Capital, Principal VC, and individual researchers from OpenAI and Anthropic.
The company is building software that allows AI models to run on a broad range of chip types — including GPUs, SRAM, phone chips, and Systolic Arrays — without relying on Nvidia’s CUDA software stack. CUDA has become a dominant force in AI development because major frameworks like PyTorch and TensorFlow are built on top of it, meaning applications default to running on Nvidia hardware. Most smaller AI developers lack the resources to write the low-level kernel code required to support alternative chips, which is the gap Infinity is targeting.
Infinity’s core product is an AI research agent called Ignition, which writes, tests, debugs, and rewrites the low-level kernel code needed for inference on non-Nvidia chips. The system is self-optimizing and adapts to different chip architectures. According to the company, the agent can compress what would otherwise take months or years of human engineering work into hours or days. Humans remain involved, providing high-level direction while the agent handles lower-level tasks.
The company was founded by Jeremy Nixon, a former Google Brain researcher and creator of the hacker network AGI House. Nixon told TechCrunch his interest in “automated invention” — the idea that AI can function as a meta technology — motivated the company’s direction. He had previously developed a machine learning algorithm called Omega that generated and evaluated new algorithms automatically, and extended that thinking to chip-level software generation.
Current customers include AI chip maker D-Matrix, and Infinity says it is in talks with other chip and cloud companies. Rather than charging upfront licensing fees, the company takes a share of performance gains and cost savings, measured in tokens per second. Infinity currently employs 26 people across engineering, design, and operations.
Source: TechCrunch