Nanoscale Thermal Prediction for Real-World Chip Design

April 1, 2026

Defense Advaned Research Projects Agency (DARPA)
Arlington, VA
www.darpa.mil

As microelectronics push far below the 10-nanometer scale, heat has become one of the most significant barriers to next-generation chip performance. Packing billions, or even trillions, of transistors into fingernail-sized chips concentrates heat in ever-smaller structures, where it becomes trapped and can ultimately cause device failure.

Chipmakers need tools that can predict these thermally driven failures before investing years and hundreds of millions of dollars in fabrication. But existing commercial modeling tools have not been able to fully capture nanoscale heat flow.

DARPA’s Microsystems Technology Office launched the Thermal Modeling of Nanoscale Transistors (Thermonat) effort to close this gap. Thermonat aimed to combine the accuracy of atom-level physics with the speed required for industry design timelines, targeting predictions within 1°C of ground truth and reducing computation time by more than 1,000x.

“The Thermonat teams pushed the boundaries of what is possible in chip-scale thermal prediction,” said Dr. Yogendra Joshi, Thermonat’s Program Manager at DARPA. “By connecting fundamental physics with design-ready tools, they created capabilities that can accelerate innovation for both national-security applications and the broader semiconductor industry.”

The Thermonat performers didn’t stop once they delivered on these ambitious technical goals; they’ve gone on to pursue a range of commercial pathways for their work. One research team formed a startup, AtomTCAD Inc., to bring high-accuracy thermal modeling tools directly to semiconductor designers. The newly formed company has received funding from the Colorado Office of Economic Development & International Trade.

IBM, another Thermonat performer, has already integrated key elements of its atoms-to-circuits modeling approach into internal product design kit (PDK) processes. With this adoption, IBM has seen that designers can evaluate technology options earlier in development.

DeepSim, a startup working with Stanford University under Thermonat, has gained early commercial traction. The company was selected for the Y Combinator program, raised seed funding, and is in active discussions with several semiconductor companies to commercialize its DARPA-enabled modeling capabilities.

Visit Here