SandboxAQ Receives $500M from CHIPS Act to Unite AI and Quantum Computing in New Materials

The United States just wagered half a billion dollars on an idea that, until recently, would have sounded like science fiction: using artificial intelligence and quantum computing to invent the materials that will power the next generation of chips. SandboxAQ, a spin-off from Alphabet, signed a definitive agreement in June worth $500 million with the Department of Commerce under the CHIPS Act. The money is not for building a factory. It is for discovering new chemistry.

What Happened

The CHIPS Act research and development program closed one of its largest individual commitments with a company that does not manufacture semiconductors, but rather molecules. SandboxAQ will use its ReAQT simulation platform and its Large Quantitative Models (LQMs)—models trained on physics, chemistry, and biology, not human language like the chatbots you know—to screen millions of candidates for new materials and formulations.

The target is specific and revealing. Four critical frontiers for the chip fabrication supply chain: replacing PFAS, the so-called forever chemicals (fluorinated compounds that do not degrade in the environment and face growing regulatory pressure); developing catalysts; creating rare-earth-free magnets; and advancing battery systems. In other words, attacking precisely the points where the US today depends on imports or substances that regulators want to ban.

There is a governance detail worth noting. In an uncommon structure for a federal R&D investment, the Department of Commerce receives a minority equity stake without voting rights in SandboxAQ, plus future royalties on formulas licensed to industrial partners. The American government is not simply funding research—it is positioning itself as a shareholder in the outcome.

Why This Matters in 2026

The larger context is a Cold War over technological sovereignty. Chips have stopped being a commodity to become a geopolitical asset, and the fragility lies not only in lithography factories—it lies in base chemistry. A banned PFAS or a rare earth under embargo can shut down an entire production line. Washington's bet is that AI shortens the path between the problem and the molecule that solves it.

And here is the conceptual turn that makes this news larger than it appears. Materials discovery has always been a game of trial and error at the bench, measured in decades. Think of how long it took the world to arrive at the lithium batteries we use today. What SandboxAQ proposes is compressing that cycle from decades to weeks, using simulation to eliminate dead ends before any physical experiment. AI is not the final product here. It is the instrument of discovery.

Analytical humility is warranted: none of this is guaranteed delivery. A model pointing to a promising candidate is different from a factory producing that compound at scale, with viable purity and cost. The bridge between simulation and the production line is long, and the $500 million value is precisely a risk bet on that crossing—not a check for a result that already exists.

Practical Implications

For companies, even those that will never touch a wafer, the signal is clear: the competitive frontier of AI has shifted from text to the physical. Models that understand physics and chemistry, not just words, open territory where advantage lies not in writing better, but in simulating better. Whoever masters the algorithm-chemistry-fabrication cycle gains time—and in R&D, time is compound money.

For the Brazilian market, there is a dual reading. The first is one of dependence: Brazil is a relevant supplier of raw materials, including rare earths, and a world that learns to replace them through design redesigns value chains that today favor us. The second is one of opportunity. We have strong universities in chemistry and materials, agribusiness with real formulation challenges, and sectors such as mining and energy hungry for optimization. The barrier to entry for using scientific AI has fallen—a national supercomputer is no longer necessary to run useful simulations.

The 10Dobro Angle

At 10Dobro, we read this movement as the convergence of the decade: AI ceasing to be a destination and becoming a discovery tool. The same principle we apply to automation and performance marketing—using the algorithm to compress the cycle between hypothesis and result—now operates at the level of matter itself. It is not the machine replacing the bench chemist. It is the machine multiplying what a good laboratory can test, at a ratio of roughly ten to one.

The operational lesson is transferable to any company. The question is no longer whether AI fits your process, but which expensive and slow stage of your business can be pre-screened by simulation before physical investment. Materials discovery is just the most visible case of a general pattern.

The takeaway is sharp: the future of chips will not be decided solely by who has the best factories, but by who finds first the chemistry no one has invented yet. And, for the first time, finding has become a software problem.

Sources: PRNewswire (official SandboxAQ/Department of Commerce announcement); The Register; The Quantum Insider; Quantum Computing Report.