Energy and infrastructure are major bottlenecks to advancing AI capabilities, with energy constraints playing a critical role alongside hardware and other infrastructure.

Beyond energy, bottlenecks in AI buildout also include interconnects, optical components, and memory, which collectively slow progress.

At Davos, OpenAI's CEO warned that AI’s energy demands require breakthroughs to reach superintelligent levels.

Projections from the IEA suggest data centers powered by AI could consume about 945 terawatt-hours of electricity worldwide by 2030, roughly 3% of total global consumption.

Achieving sustained AI growth may depend on breakthroughs in energy technology, including fusion energy, to scale the necessary infrastructure.

Current energy infrastructure is insufficient to meet rising AI demand, marking a major bottleneck in scaling AI capabilities.