AI Infrastructure Startups Attract Record Investment as Capital Shifts Away from Models and Chips

Venture capital is increasingly flowing toward foundational AI infrastructure rather than frontier models and semiconductors, as investors recognize that supporting systems are critical bottlenecks. On Monday, two early-stage startups announced major funding rounds backed by prominent investment firms including Sequoia, a16z, and Khosla Ventures. Mariana Minerals, which focuses on critical minerals essential to semiconductor production, secured funding from these investors to address supply chain constraints. The trend reflects growing recognition that AI's explosive compute demands create cascading infrastructure needs: nuclear companies have attracted significant capital in recent months, including Antares, which raised $470 million, and Commonwealth Fusion Systems, which raised $1 billion, both developing small modular reactors. Similarly, companies targeting mineral extraction like Terra AI and EnergyX are gaining investor attention. The shift signals a market understanding that the companies supplying compute power, energy, networking, cooling, and industrial capacity may prove as valuable as those building AI models themselves. Historically stable sectors like nuclear energy and critical mineral mining offer investors a hedge: these resources retain utility and value regardless of AI's ultimate trajectory, making infrastructure plays potentially more durable than bets on specific model capabilities.