Federal Investment Targets Semiconductor Supply Chain Vulnerabilities

The U.S. Department of Commerce’s CHIPS Research and Development Office has signed a definitive agreement to award $500 million to SandboxAQ. This funding aims to address acute supply chain vulnerabilities in domestic semiconductor manufacturing by developing alternative chemical formulations and advanced materials to bypass foreign chokepoints.

This award represents a shift in how the federal government structures R&D support. In a unique arrangement, the Department of Commerce will receive a minority, non-voting equity stake in the NVIDIA-backed enterprise, alongside future royalty payments from formulas licensed to industrial partners for mass production. This structure moves the government from a pure grantor to a stakeholder in the commercialization of high-assurance technology.

The technical core of this initiative relies on Large Quantitative Models (LQMs) and the ReAQT simulation platform. While standard models train on text, LQMs train directly on the fundamental laws of physics, chemistry, and biology. By using high-fidelity quantum chemistry simulations—including Density Functional Theory (DFT) and Molecular Dynamics—the platform generates its own training sets. This allows for the screening of millions of untried chemical candidates to find commercially viable alternatives, compressing materials discovery timelines from decades into weeks.

The immediate focus is the mitigation of regulatory and supply chain risks in chip fabrication. SandboxAQ will work to replace Per- and Polyfluoroalkyl substances (PFAS), or “forever chemicals,” which are used in lithography lines as lubricants, insulating coatings, and heat-transfer fluids. The project includes designing drop-in replacement fluids and developing techniques to break down legacy PFAS waste.

The firm is utilizing its AQCat workflows, which are built on 13.5 million high-fidelity simulations, to drive this process. This physics-first methodology enables automated Design-Make-Test-Learn (DMTL) loops, reducing the reliance on traditional trial-and-error laboratory experimentation.

The success of this program depends on whether these simulated molecular maps can translate into stable, mass-producible industrial chemicals.

Watch for the first industrial partners to announce licensing agreements for these new chemical formulations.