
On 17 June 2026, the U.S. Department of Commerce and SandboxAQ signed a definitive agreement for a $500 million CHIPS Act R&D award — confirmed by NIST's own announcement. Under the agreement, SandboxAQ will use physics-based AI to discover, validate, and commercialize new chemical formulations for domestic semiconductor manufacturing, across four programmatic areas: PFAS-free process chemicals, catalysts, rare-earth-free magnets, and battery systems.
The PFAS component is the most direct link to industrial chemistry. Per- and polyfluoroalkyl substances — the "forever chemicals" used throughout chip fabrication as heat-transfer fluids, lubricants, insulating coatings, and surface treatments — currently have no compliant substitutes at scale. The award funds AI-driven screening of candidate molecules aimed at that gap.
The deal's structure is also notable for a federal R&D award: rather than a conventional grant, the Department of Commerce takes a minority, non-voting equity stake in SandboxAQ, plus a share of future royalties.
Insights
ChemAbout connects chemical buyers and suppliers worldwide. Tell us what you need, or list what you offer.
Free, no sign-up — stays anonymous
Requires a free company account
An AI chip isn't just bigger than an ordinary chip — it passes through more lithography, etch, clean, and polish cycles, each consuming ultra-high-purity photoresist, CMP slurry, electronic-grade HF, sulfuric acid, and PGMEA at parts-per-trillion purity. A fact-based map of why this supply chain is short, hard to switch suppliers on, and now visibly tightening.

Between 2024 and 2026, some material directions advanced roughly on schedule — silicon carbide, carbon fiber — while others, like PFAS alternatives, green methanol and AI-designed catalysts, show a real gap between what's been announced and what's actually deployed. This piece separates genuine, dated progress from hype and unproven claims across five materials directions, drawing on English, Japanese and German primary sources.
Photoresist is a translator: it converts optical information into chemical information, then a physical pattern, then an electronic device. Every reduction in exposure wavelength - g-line to i-line to KrF to ArF to EUV - has forced this translator to rewrite itself, from the invention of chemical amplification at IBM in 1982 to today's photon-shot-noise limits at EUV. This deep dive traces the chemistry, the 2019 Japan-Korea export control episode, and why the real moat is process, qualification, data and time - not the formula.