SandboxAQ has signed a definitive agreement with the U.S. Department of Commerce's CHIPS Research and Development Office for a $500 million award to attack one of the most stubborn weaknesses in American chipmaking: foreign control of the materials and chemistries that semiconductor factories depend on. The deal, announced on June 17, 2026, ranks among the largest federal bets to date on physics-based AI for scientific discovery.
What the award funds
The money is directed at four programmatic areas where overseas supply has long suppressed domestic production: PFAS-free process chemicals, catalysts for ultra-pure gas generation and exhaust abatement, rare-earth-free permanent magnets, and battery chemistries that avoid lithium and other foreign chokepoints. SandboxAQ will virtually screen millions of candidate molecules, validate the most promising with lab partners, then hand the strongest results to American manufacturers for scale-up. In exchange, the Commerce Department will receive a minority, non-voting equity stake in the company.
Physics-trained models, not language models
At the center of the program is ReAQT, SandboxAQ's AI simulation platform, and its Large Quantitative Models (LQMs) - AI systems trained on the laws of physics, chemistry and biology rather than on human language. ReAQT generates its own high-fidelity training data through Density Functional Theory, molecular dynamics and reaction modeling, allowing the models to predict the behavior of materials that have never been synthesized. "Securing America's semiconductor future means controlling the materials that drive this vital sector," said SandboxAQ CEO Jack Hidary.
A national-security supply-chain play
China controls more than 90% of global neodymium magnet production, and those magnets sit inside virtually every chip-printing machine, vacuum pump and wafer-positioning actuator. By compressing materials-discovery timelines from years of lab trial-and-error into AI-driven campaigns, Commerce hopes to blunt that dependency. The award lands amid a wave of large U.S. semiconductor and compute commitments, including Google's order of 3 million Intel-made TPUs and NVIDIA's multiyear memory pact with SK Hynix.
For SandboxAQ, the award validates a commercial thesis it has pursued since spinning out of Alphabet in 2022 and follows a year of heavy private investment across the robotics and AI hardware stack, such as THEKER's record European robotics round. The company says the funding will deepen ReAQT's capabilities and accelerate the move from simulation to scaled domestic manufacturing.
Reporting based on coverage from SandboxAQ, the U.S. Department of Commerce/NIST, Reuters and PR Newswire.