Lamb Labs is building Model Processing Units (MPUs), custom AI chips that hardcode LLM weights directly into silicon. Rather than adapting general-purpose processors to run AI models, we're rethinking how LLMs are implemented in hardware to make inference dramatically faster and more energy-efficient. The company was founded by Niki Kotecha; Thomas Lanning.
It is part of YC Summer 2026. It is based in San Francisco, CA, USA.