Bezos-backed CuspAI raises $2.6B to use AI for chip materials

CuspAI addresses a real and growing constraint in the AI hardware supply chain. As demand for AI computing infrastructure accelerates, chipmakers face a pinch: critical materials like gallium, iridium, and tantalum are either geopolitically concentrated (China controls 99% of refined gallium) or becoming costlier and harder to source. Traditional materials research is slow and expensive—a process that can consume years and millions of dollars. By using AI to virtually model alternatives and develop new syntheses, CuspAI compresses that timeline to months and slashes costs, turning a protracted technical problem into something addressable in real time.

The opportunity is substantial enough to draw marquee players. Nvidia and AMD, whose products sit at the center of AI infrastructure buildouts, are naturally interested in securing next-generation materials. Meta, which operates enormous data centers, has the same incentive. Bezos' involvement signals confidence that the problem is durable: supply constraints are not a temporary blip but a structural challenge that will persist through the decade. Forecasts from Mordor Intelligence and Precedence Research project the generative AI material science industry will grow at more than 24% annually at least through 2030, indicating that investors and analysts expect sustained demand for this type of work.

CuspAI is not the only bet Bezos is placing on AI-enabled manufacturing. In November, Bezos Expeditions also invested in Prometheus, an AI startup focused on design and manufacturing of complex physical products, including potentially microchips. The two investments reinforce each other: Prometheus could use materials discovered by CuspAI to build better hardware. Together, they sketch out a vision in which AI becomes a tool for collapsing both the cost and timeline of physical innovation—a significant shift for industries that have historically moved slowly.