The case for AI's bottleneck shifting from models to infrastructure

In a discussion about a16z's new Machine Age Fund, the panel argues that AI's constraints increasingly lie beneath the models—in chips, memory, networking, power, cooling and data centers. The speakers point to surging capital spending by major cloud providers, critical components booked years ahead and growing compute demands from reasoning and agents. They also argue that capital and compute can increasingly tackle problems once constrained by engineering, opening opportunities for new infrastructure companies.