Normal Computing Raises $50M to Rethink AI Efficiency with Thermodynamic Computing

AI has a cost problem nobody wants to headline. Not the glossy kind you pitch on stage, the real one buried in power bills, cooling limits, and silicon that takes longer to design than it does to become obsolete. Normal Computing stepped into that tension and decided efficiency was not a feature, it was the whole game.

Normal Computing just pulled in $50M, led by Samsung Catalyst Fund, with Galvanize, Brevan Howard Macro Venture Fund, and ArcTern Ventures stepping in alongside Celesta Capital, Drive Capital, Eric Schmidt’s First Spark Ventures, and Micron Ventures. That brings the total to more than $85M. Not pocket change, but more importantly, not dumb money either. This is a table full of people who understand that silicon is the new leverage.

Credit to Faris Sbahi, CEO, for steering this with a clear thesis. If AI is hungry, then feeding it the same way forever is a losing game. Normal Computing is building an AI-native EDA platform designed to accelerate chip design, while developing thermodynamic computing hardware aimed at squeezing more intelligence out of every watt. Translation for anyone not living in the weeds: faster chips, cheaper to run, and built with physics in mind instead of brute force optimism.

The company is already working with more than 50% of the top 10 semiconductor companies by revenue. That is not a vanity metric. That is distribution, validation, and pressure all rolled into one. When the incumbents start listening, you are no longer early, you are inevitable.

The business takeaway here is simple, but not easy. They did not raise this round on vibes. They anchored themselves at the intersection of 2 unavoidable problems: rising chip design complexity and the very real energy ceiling facing AI infrastructure. When you position yourself where pain is guaranteed, capital tends to find you.

Samsung Catalyst Fund leading this round is its own signal. When a strategic like Samsung leans in, it is not just about returns. It is about proximity to what could define the next generation of compute. Everyone else at the table followed that logic.

And the subtle pitch to the market is baked right in. If you are designing advanced silicon and still treating energy as an afterthought, you are already behind. Normal Computing is betting that “normal” compute economics will not survive the next decade, and they are building accordingly.