U.S. Bets $500 Million on AI-Driven Materials Discovery: SandboxAQ Lands Largest CHIPS Act R&D Award
Not building chips — finding the raw materials to build them. The U.S. government just placed its bet further upstream.
Three Key Takeaways
The U.S. Commerce Department's CHIPS R&D Office signed a definitive agreement with SandboxAQ on June 17 for a $500 million award to fund AI-driven semiconductor materials research. This is the largest single R&D award under the CHIPS Act to date. SandboxAQ, spun out of Alphabet (Google's parent company), builds what it calls Large Quantitative Models (LQMs) — AI systems trained not on human language but on the laws of physics, chemistry, and biology. These models compress what would otherwise take decades of lab trial-and-error into targeted, algorithm-driven discovery campaigns.
The funding covers four critical areas: PFAS-free process chemicals, catalysts, rare-earth-free permanent magnets, and battery systems. Why these four? Because China controls more than 90% of global neodymium-based permanent magnet production — magnets that are core components in semiconductor manufacturing equipment. Meanwhile, chip factory backup power systems depend on lithium, cobalt, and other battery materials heavily concentrated overseas. The goal is clear: use AI to find substitutes and decouple from Chinese supply chain dependencies at the materials level.
As a condition of the award, the Commerce Department will receive a minority, non-voting equity stake in SandboxAQ. This means the U.S. government isn't just writing a check — it's taking a seat in the company's capital structure. Part investment, part oversight.
WangDou's Take
Chip export bans have been the main event for several rounds now, but nobody's been seriously talking about raw materials. Here's the thing: you can ban lithography machines all you want, but 90% of the magnets inside those machines come from China. Five hundred million dollars sounds like a lot until you compare it to the $40 billion price tag of a single TSMC fab. Still, SandboxAQ's approach is genuinely interesting — using physics-law-trained AI to search materials space turns the lab's "needle in a haystack" problem into an algorithm's precision-guided search. If they actually find a rare-earth-free permanent magnet alternative, the impact goes way beyond chips — EVs, wind turbines, defense hardware, the whole chain. Of course, "if" is doing a lot of heavy lifting in that sentence.