cd /news/artificial-intelligence/the-us-is-paying-an-ai-startup-500m-… · home topics artificial-intelligence article
[ARTICLE · art-31270] src=thenextweb.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

The US is paying an AI startup $500M to break China’s grip on chip materials

The US Department of Commerce awarded AI startup SandboxAQ $500 million under the CHIPS Act to develop new semiconductor materials, reducing reliance on Chinese supply chains. The government takes an equity stake and royalties, acting like a venture investor. SandboxAQ uses AI trained on physics and chemistry to discover materials faster, though no products exist yet.

read2 min views1 publishedJun 17, 2026

The US government has just become a shareholder in an AI startup, in exchange for a bet that artificial intelligence can break China’s hold on the materials inside every chip factory.

The Department of Commerce has awarded SandboxAQ $500mn under the CHIPS Act to develop new chemicals and metals for domestic semiconductor manufacturing, Reuters reported. In return, Commerce takes a minority, non-voting equity stake in the company, and royalties if the work pays off.

SandboxAQ, backed by Nvidia and valued at $5.75bn last year, will target four materials where US chipmaking leans on foreign supply: “forever chemical” PFAS substitutes, catalysts, magnets that avoid Chinese rare earths, and batteries that avoid imported lithium.

Government as venture investor #

The structure is the story. This is not a simple grant. By taking equity and a cut of future royalties, Washington is behaving like a venture fund, the same model it used in a recent $2bn quantum-computing award.

SandboxAQ’s chief executive Jack Hidary declined to say how big the government’s stake is, only that it carries no voting rights or board seat. It sits alongside backers including Google, Eric Schmidt and Ray Dalio.

An AI that learns physics, not language #

What SandboxAQ sells is not a chatbot. Its “Large Quantitative Models” are trained on physics and chemistry rather than human text, and are meant to screen millions of candidate materials in software before anything is made in a lab.

The company claims this compresses discovery from years to weeks. Its catalyst models, built on 13.5 million calculations run with Nvidia, are pitched as roughly 20,000 times faster than conventional methods, though those are SandboxAQ’s own figures.

The catch #

This is a research bet, not a product. No new magnet or PFAS-free chemical exists yet, and the award funds the search, not a guaranteed result.

There is an irony worth noting too: the same administration funding PFAS replacements rolled back some PFAS drinking-water deadlines last year. But the strategic logic is hard to miss, with China controlling more than 90 per cent of the rare-earth magnets that the machines making America’s most advanced chips depend on, supply-chain sovereignty has become the point.

Get the TNW newsletter #

Get the most important tech news in your inbox each week.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @sandboxaq 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/the-us-is-paying-an-…] indexed:0 read:2min 2026-06-17 ·