The CHIPS Act grant tasks an Alphabet spinoff with finding alternatives to chemicals and rare-earth elements that keep US chipmaking dependent on foreign supply chains
The US Department of Commerce just handed SandboxAQ, a company that spun out of Google’s parent Alphabet in 2022, a $500 million research and development grant to find new materials for making semiconductors. The funding comes from the CHIPS Act, which Congress passed in 2022 to rebuild America’s chipmaking infrastructure.
Here’s the thing: this isn’t about building another chip factory. It’s about solving a quieter, arguably more important problem. The chemicals and raw materials that go into semiconductor manufacturing are themselves a chokepoint, and many of them flow through supply chains that run directly through China.
What SandboxAQ actually does #
SandboxAQ isn’t a typical chipmaker or even a typical AI company. The Palo Alto-based firm specializes in what it calls physics-based AI, combining machine learning with the laws of physics to solve problems that pure software approaches struggle with.
Its flagship tool for this project is a platform called AQCat, which screens potential catalyst candidates for chemical and materials processes. The speed advantage is not subtle. AQCat can evaluate candidates approximately 20,000 times faster than traditional lab methods.
The company already has a track record with the US government, holding contracts in cybersecurity and quantum sensing. But this grant represents a massive expansion of its mandate into semiconductor supply chain resilience.
SandboxAQ is valued at $5.75 billion following a Series E funding round in April 2025 that pushed its total fundraising past $1 billion. Nvidia is among its backers.
Why PFAS and rare earths matter #
The grant has two primary focus areas: developing alternatives to PFAS chemicals and reducing dependence on rare-earth elements in chip manufacturing.
PFAS, sometimes called “forever chemicals,” are used extensively in semiconductor fabrication because of their heat resistance and chemical stability. They’re also an environmental nightmare, persisting in soil and water essentially forever. Regulatory pressure to phase them out is mounting globally, but the chip industry doesn’t yet have viable replacements for many applications.
Rare-earth elements present a different kind of problem. They’re not actually rare in geological terms, but the mining and refining infrastructure is overwhelmingly concentrated in China, which controls a dominant share of global processing capacity. When geopolitical tensions spike, as they have repeatedly in recent years, that concentration becomes a strategic vulnerability for every country that makes or uses semiconductors.
The bigger CHIPS Act picture #
This grant fits into a broader pattern of CHIPS Act spending that has increasingly moved beyond just building fabs. The original selling point of the legislation was straightforward: bring chip manufacturing back to American soil. And billions have flowed to companies like Intel, TSMC, and Samsung for exactly that purpose.
But a factory is only as useful as its inputs. If you build a state-of-the-art semiconductor fab in Arizona and still need to import critical chemicals from a geopolitical rival, you haven’t fully solved the supply chain problem. You’ve just moved the bottleneck.
For context, SandboxAQ’s roots in Alphabet give it access to deep technical talent, and the AQCat platform’s 20,000x speed advantage over traditional screening methods suggests the technology is already functional, not theoretical. The $500 million is meant to scale and direct that capability toward specific semiconductor applications.
What this means for investors #
The most immediate signal here is about government spending priorities. Half a billion dollars going to an AI-driven materials discovery company, rather than another chip fabrication facility, tells you where policymakers see the next wave of semiconductor investment heading.
The risk, as with any materials R&D program, is timeline uncertainty. Discovering a promising catalyst candidate in simulation and scaling it to industrial production are two very different challenges. The 20,000x screening speed is impressive, but it accelerates only the first step of a long pipeline from lab to fab.
Investors should also note that SandboxAQ’s $5.75 billion valuation makes it one of the more richly valued private companies in the AI-for-science space.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our