Archive — US — techcrunch
Inference startup Infinity raises $15M from Touring Capital, OpenAI and Athropicresearchers
Infinity, an AI infrastructure startup, raised $15M from Touring Capital, OpenAI, and others to develop a universal software for AI chips, aiming to rival Nvidia's dominance through a CUDA-alternative kernel system. Founded by ex-Google Brain researcher Jeremy Nixon, the startup’s self-optimizing AI agent writes and improves low-level code for efficient chip operation, making AI model deployment more accessible across different architectures.
Key points
- Infinity raised $15M at a $100M valuation to build software for AI chips.
- Investors include Touring Capital, OpenAI, and others.
- Infinity's software aims to create CUDA-alternative kernels for various chips.
- The company believes automated systems can generate necessary code for better chip performance.
- Its AI agent, Ignition, optimizes low-level code autonomously.
- Customers include AI chip maker D-Matrix; discussions with other big firms ongoing.
- Infinity's business model relies on performance gains and efficiency rather than upfront licensing fees.
Why it matters
- Current trend of AI startups developing solutions to minimize reliance on Nvidia.
- The rise of companies focusing on alternative infrastructure for AI deployments aligns with the growing emphasis on chip diversity.
- Machine learning and automated coding trends are prevalent among startup ecosystems, reflecting a shift towards self-optimizing systems.
Infinity's innovative approach to developing universal AI chip software positions it uniquely against established giants, potentially transforming AI infrastructure as it addresses growing needs for flexibility and efficiency in chip utilization.