Infinity, a new AI infrastructure company, has raised $15 million from investors including Touring Capital and researchers from OpenAI and Anthropic to address Nvidia's software dominance by creating a CUDA-like software stack that works across various AI hardware platforms.
- Infinity creates a universal inference software stack for non-Nvidia AI chips
- AI agent automates kernel code writing, tuning, and debugging to boost hardware efficiency
- Startup funded by Touring Capital and AI researchers from OpenAI and Anthropic
Market signal
Infinity's $15 million funding round highlights growing investor interest in software solutions that enable broader adoption of diverse AI chips beyond Nvidia's dominant CUDA ecosystem. The startup addresses a critical bottleneck in AI deployment: the need for low-level, optimized code that can run AI models efficiently on different hardware architectures without requiring expensive and time-consuming manual engineering.
This funding supports Infinity’s development of a universal inference kernel library combined with an AI-driven software agent that autonomously generates and refines the underlying code for a range of chips including GPUs, SRAM, and specialized AI accelerators. Backing from influential investors and AI researchers signals confidence in software as a strategic lever to chip away at Nvidia’s market share and expand AI infrastructure options.
Operator impact
For AI chip operators and cloud service providers, Infinity offers a way to increase the performance and efficiency of alternative AI hardware by reducing dependence on Nvidia’s CUDA. By leveraging Infinity’s self-optimizing software agent, operators can streamline low-level code development, significantly reducing time and costs associated with optimizing AI workloads on diverse chips.
Infinity’s pay-for-performance licensing aligns incentives around actual improvements in throughput and cost efficiency, measured as token processing rates. This means operators can avoid upfront licensing fees and instead benefit directly from the software’s ability to increase hardware utilization and lower operating expenses, supporting more competitive AI infrastructure offerings.
What to watch next
Monitoring Infinity’s progress in securing partnerships with major chip makers and cloud providers will be key, as broad adoption of its software stack is critical for challenging Nvidia’s entrenched ecosystem. The startup’s ability to integrate with popular AI frameworks and maintain compatibility across proprietary chip designs also remains an important technical milestone to watch.
Emerging competitors attempting similar cross-chip software solutions may influence the trajectory of AI infrastructure diversification. Additionally, advances in Infinity’s AI research agent and its automation of kernel generation could set new standards for how AI inference software adapts to new hardware, accelerating innovation cycles in AI chip deployment strategies.