# Jack Hidary's SandboxAQ wins $500 million CHIPS award for AI chipmaking materials > Source: > Published: 2026-06-17 12:23:57+00:00 [Jack Hidary](https://www.sandboxaq.com/company/leadership/jack-hidary?ref=runtimewire)'s [SandboxAQ](https://www.sandboxaq.com/?ref=runtimewire) signed a definitive agreement with the U.S. Department of Commerce for a $500 million CHIPS Research and Development award to use AI to discover semiconductor materials, a deal [Reuters](https://news.google.com/rss/articles/CBMiwwFBVV95cUxNdk4xYlJEQzhuS1ZNa3dQRHhibDlrS3VuV0pqTEdZNGRSVE43SjdIaHBaVjZlMnEzaWY5UmhNamNkWlF3T3NoR1pPY1RFcHVzU0xUQ2I2SzhyTHJTWUZJMl9abG5KclAzd0JXRHJ2Z2laV2drcHdKbzRQNlZ6RmhESG45QzRFRUg5ekhtYjhYRVVFMzZKQldxYXozTmtBb3NDRXVlaWZNU1VEQTVITkxxcEdCd1ZIOWd4SkNVQ3YyTFdQSjA?oc=5&ref=runtimewire) framed as a win for a NVIDIA-backed company. The [Commerce announcement](https://www.nist.gov/news-events/news/2026/06/department-commerce-announces-definitive-agreement-sandboxaq-500-million?ref=runtimewire) says the award sits under the CHIPS and Science Act and is aimed at AI-driven materials discovery for the semiconductor supply chain. For Hidary, the award is a public-sector test of the thesis he has been selling since SandboxAQ came out of Alphabet: that AI built for physics, chemistry, and quantitative reasoning can move from research labs into the hard inputs of industry. SandboxAQ's own leadership page says Hidary formed SandboxAQ during six years at Google leading teams in AI and quantum technologies; it also says he studied neuroscience at Columbia and later held a Stanley Fellowship in Clinical Neuroscience at NIH, where he worked on functional brain imaging and artificial neural networks. That background matters because this is not a conventional software contract. Commerce is not paying SandboxAQ to deploy a dashboard or a workflow automation tool. The department is backing a private AI company to help find substitutes and upgrades for materials that chip fabs depend on but do not always control. The key financial detail is buried below the headline number: in connection with the award, Commerce says it will receive a minority, non-controlling equity stake in SandboxAQ. That makes the deal closer to a strategic industrial-policy investment than a plain grant. It also gives taxpayers a claim, however small and non-controlling, on any upside if SandboxAQ's materials platform becomes more valuable because of the federal program. ### The four materials bottlenecks Commerce listed four priority areas for the award: PFAS-free process chemicals, catalysts for semiconductor fabrication, rare earth-free magnets, and advanced battery chemistries for backup power at semiconductor facilities. The PFAS work is about finding process chemicals for heat-transfer, lubricant, insulating coating, and surface-treatment applications that can meet chip-fab performance requirements without the environmental and bioaccumulation risks attached to so-called forever chemicals. The catalyst work targets high-purity catalysts used in upstream precursor generation and exhaust-gas abatement for fabs. The magnet work is aimed at rare earth-free magnetic materials for semiconductor manufacturing equipment, where Commerce says China controls more than 90% of global production of neodymium-based permanent magnets. The battery work is focused on new solid-state or hybrid energy storage systems for the uninterrupted power fabs require. That list shows why the award is being routed through CHIPS R&D rather than treated as a generic AI procurement. Washington has spent years trying to rebuild domestic chip capacity by subsidizing fabs. This award attacks a layer beneath fab construction: the specialty chemicals, magnets, catalysts, and power systems that determine whether a fab can actually operate without exposure to foreign-controlled supply chains. ### What SandboxAQ is selling SandboxAQ calls its core approach Large Quantitative Models, or LQMs. On its [materials discovery page](https://www.sandboxaq.com/solutions/material-discovery?ref=runtimewire), SandboxAQ says those models combine physics-based simulation with machine learning to accelerate chemical and materials discovery. In a [technical post](https://www.sandboxaq.com/post/sandboxaq-helping-build-americas-semiconductor-supply-chain?ref=runtimewire) tied to the award, SandboxAQ says its platform generates physics-grounded data, trains proprietary AI and physics models on that data, and then combines the models in automated design, make, test, learn loops. The most specific performance claim in the company's semiconductor materials post concerns catalysts. SandboxAQ says its AQCat workflows, built on 13.5 million high-fidelity chemistry calculations developed with NVIDIA, can screen catalyst candidates at near-quantum-chemistry accuracy 20,000 times faster than conventional methods. The company ties that directly to precursor chemistry for advanced chips, saying the work could help reduce development timelines from months to weeks and support manufacturing toward sub-2nm nodes. Those claims remain SandboxAQ's claims until the new materials are validated, qualified, and manufactured at scale. That is the hard part of the deal. A model can narrow a search space. It does not, by itself, make a PFAS replacement acceptable to a fab tool vendor, prove a rare earth-free magnet can survive industrial operating conditions, or qualify a battery chemistry for backup power at a facility where minutes of disruption can create yield loss. Commerce appears to understand that gap. The announcement says SandboxAQ will work with American manufacturing partners to move the strongest results into domestic manufacturing and commercialization. The award, in other words, is not just for discovery. It is for turning AI-selected candidates into something a supply chain can actually use. ### Why NVIDIA's backing matters Reuters called SandboxAQ NVIDIA-backed, and the financing history supports that description. In April 2025, SandboxAQ [said](https://www.sandboxaq.com/press/sandboxaq-closes-450m-series-e-round-with-expanded-investor-base?ref=runtimewire) NVIDIA, Google, BNP Paribas, Ray Dalio, and Horizon Kinetics were added to its Series E round, which raised more than $450 million. SandboxAQ also said at the time it had raised more than $950 million since spinning out from Alphabet in 2022. A separate Reuters report published through Investing.com said the April 2025 extension brought SandboxAQ's Series E to $450 million, valued SandboxAQ at $5.75 billion, and took total funding to $950 million. That report also said SandboxAQ employed about 200 people at the time and quoted Hidary saying strategic investors were drawn by the company's delivery for customers. NVIDIA's presence is strategically useful for SandboxAQ for two reasons. First, materials simulation is compute-hungry, and SandboxAQ's own LQM page says it is collaborating with NVIDIA across biopharma, chemicals, advanced materials, financial services, cybersecurity, navigation, and medical imaging. Second, the CHIPS award pushes SandboxAQ into the semiconductor manufacturing stack, where NVIDIA has every reason to see faster materials discovery as part of the broader AI infrastructure buildout. The relationship should not be overstated. The public materials reviewed do not show NVIDIA leading the CHIPS award, owning a specific share of SandboxAQ, or acting as a named manufacturing partner under the Commerce agreement. NVIDIA is a strategic investor and collaborator; Commerce is the party signing the $500 million definitive agreement. ### The real test The award gives SandboxAQ a larger role than most AI infrastructure startups get: it is being asked to help solve the physical bottlenecks behind compute. That is a stronger assignment than selling model access, and a more unforgiving one. Semiconductor materials must clear qualification, cost, reliability, environmental, and supply-chain tests before they matter commercially. The deal also reflects a shift in how Washington is using startup-backed technology. CHIPS policy began with fabs and capacity. This agreement treats AI materials discovery as a national capability, with the government taking a small equity position to align public funding with private upside. What Commerce has not disclosed is the payment schedule, milestone structure, or how much of the $500 million is immediately obligated versus tied to performance. That distinction matters. A definitive agreement is a serious commitment, but in materials commercialization the expensive proof comes after the model produces a candidate: validation, qualification, partner manufacturing, and adoption by customers whose tolerance for failure is low. That is the burden Hidary has taken on. SandboxAQ now has federal capital, strategic investors, and a clearly defined set of industrial targets. The next question is whether its LQMs can move beyond narrowing possibilities and produce materials the semiconductor industry will trust inside fabs.