The US government is turning a CHIPS Act award into something closer to an industrial bet. On June 17, the Commerce Department signed a definitive agreement to provide SandboxAQ with $500 million for AI-driven materials discovery for semiconductor manufacturing.
The unusual part is the structure. Washington is not only paying for research; it is taking a non-voting minority stake in SandboxAQ and expects a share of future licensing revenue if the work produces useful patents.
Four bottlenecks are in scope
The program targets PFAS substitutes used in cooling, lubrication, coatings and surface treatment; high-purity semiconductor catalysts; rare-earth-free magnets; and battery materials for backup power that reduce dependence on lithium and cobalt supply chains dominated by China.
SandboxAQ, spun out of Alphabet in 2022 and backed by Nvidia, says its AI approach can screen molecules in software before years of lab testing. For the Commerce Department, the point is speed: use AI to narrow the search for domestic alternatives to materials that sit inside vulnerable parts of the chip supply chain.
The bet is clear. If the models identify manufacturable substitutes, several pressure points in US semiconductor production loosen. If they do not, taxpayers have helped fund a risky materials experiment with equity upside but no guaranteed industrial result.
Sources: NIST, Reuters via Business Recorder, HPCwire, CocoLoop; checked the CHIPS R&D award, equity structure, royalty terms, target material categories and SandboxAQ background.