The Next AI Breakthrough Isn't Chips — It's What They're Made Of (with Shalini Sharma, Ph.D.)

Is chip materials science, not just GPUs, becoming the next AI bottleneck? In this video: what large quantitative models (LQMs) are and how they differ from LLMs, why the shift to 3D chips makes new materials critical, and how SandboxAQ plans to use its $500M CHIPS award. Shalini Sharma, Ph.D., who leads semiconductor materials innovation at SandboxAQ and previously worked at Applied Materials, joins us to explain how AI and quantum-based simulation can screen millions of candidate chemistries before anything reaches a lab. We cover SandboxAQ's business model and four CHIPS focus areas: PFAS-free chemistries, catalysts and precursors, fab backup batteries, and rare-earth-free magnets. We also discuss what faster materials R&D could mean for suppliers like Applied Materials, Entegris, Qnity, and Shin-Etsu, and for onshoring the chip supply chain. Chapters 00:00 Intro: Shalini Sharma and SandboxAQ 01:15 LQMs vs. LLMs explained 02:20 Why materials are the hidden AI bottleneck 03:55 From 2D scaling to 3D chips and the Angstrom era 06:30 CFETs, compound semis, and science as a search problem 09:15 Do LQMs have a scaling law? 10:30 GPUs, ASICs, and where quantum computing fits 12:10 Applied Materials vs. SandboxAQ culture 16:45 Favorite science meme 17:20 The $500M CHIPS award 20:00 Supply chain risk: 90% of key magnet materials made abroad 20:55 How SandboxAQ makes money 22:45 Three go-to-market paths, including LQMs on Claude 24:15 Customers and the four CHIPS verticals 30:20 PFAS in data center cooling fluids 31:50 What this means for materials suppliers 33:45 A new wave of industrial startups? 36:40 Applied Materials' AI tools for process recipes 40:05 Wild card: best spot in Palo Alto 41:45 Wrap-up — LINKS & RESOURCES — Join Semi Insider — $120/year Semi Insider is Chip Stock Investor's paid research service: all research as it happens, the CSI 100 + quarterly outlook, dashboard tools (Compare Companies, Reverse DCF, supply-chain mapping), weekly live Q&A, portfo