Normal Computing has taped out CN101, the world’s first thermodynamic ASIC for AI/HPC data centers. Leveraging thermodynamics, it outperforms traditional chips in efficiency.
Unlike conventional models, these chips operate in quantum/probabilistic realms, using noise to solve problems. Starting semi-randomly, they reach balanced states as solutions—unsuitable for daily tasks but excelling in AI image work.
CN101 handles linear algebra and matrix operations efficiently, with a dedicated sampling system for probabilistic tasks, boosting energy efficiency by up to 1000x for AI training.
Normal plans integrating CPUs, GPUs, and its thermodynamic ASICs in servers, with new CN products in 2026 and 2028. As silicon hits limits, such alternatives could lead chip breakthroughs.
Focused on generative AI optimization, Normal builds infrastructure to solve critical enterprise/government issues.
ICgoodFind: CN101 opens new AI computing paths, potentially reshaping data centers.
