Normal Computing

Normal Computing builds AI software for semiconductor engineering and develops thermodynamic computing hardware. Normal EDA helps chip verification teams turn design specifications into test plans, trace tests to their source requirements, and refine results through engineer feedback. It integrates with existing simulation environments, including VCS and XCelium. The platform uses autoformalization: combining language models with formal logic to translate specifications into mathematical representations. Alongside this software, its Carnot hardware program uses Normal EDA to develop silicon IP for AI workloads.

Founded in 2022 by CEO Faris Sbahi, Antonio Martinez and Yui-Hong “Matthias” Tan, Normal connects hardware verification research with alternative computing architectures. Its DRAMBench, developed with Fraunhofer IESE, evaluates AI translation of memory specifications into timed Petri nets that represent states, commands and timing constraints. Its CN101 prototype implements thermodynamic computing on standard digital CMOS: answers emerge from time-averaged stochastic dynamics. A 2026 research paper describes fabricated hardware and demonstrations involving image generation and scientific problems.

In March 2026, the company reported partnerships with more than half of the ten largest semiconductor companies by revenue. It also announced $50 million in strategic funding led by Samsung Catalyst Fund, bringing total funding to more than $85 million to advance semiconductor design software and physics-based computing hardware.