Key facts
- The AI infrastructure race is now bottlenecked by electricity availability, not chip supply.
- Demand for AI compute requires power equivalent to small cities, straining existing grids.
- NVIDIA's valuation soared due to its control over the AI chip supply, demonstrating the power of controlling a bottleneck.
- Companies are securing long-term power deals, including restarting nuclear plants, to meet AI energy demands.
- Bitzero Holdings has secured over 1 gigawatt of low-cost power capacity in strategic locations.
- Bitzero's power costs in Norway are significantly lower than the U.S. average.
The race to build the infrastructure for artificial intelligence has shifted from a focus on chip manufacturing to the availability of electricity, a critical bottleneck that could limit the expansion of AI capabilities.
Two and a half years ago, NVIDIA was valued at $300 billion as a gaming chip company. Today, it is one of the most valuable companies in history, exceeding $4 trillion, driven by the explosive demand for AI compute. This demand, particularly after the launch of ChatGPT in November 2022, overwhelmed the supply of high-end GPUs, giving NVIDIA significant pricing power.
However, the chip shortage is largely resolved with increased foundry capacity. The new constraint is electricity. Training large language models requires power on the scale of small cities, and industry forecasts predict a surge in data center power demand of up to 165% by 2030. This demand far exceeds the grid's capacity, which was designed for slower, predictable growth. Many planned data centers may never be built due to these grid limitations, with over 70% of U.S. grid interconnection requests being withdrawn.
Unlike the chip shortage, which was an 18-24 month manufacturing problem, the power shortage is a decade-long infrastructure challenge with no quick solutions. Building new nuclear plants takes 10-15 years, and new transmission lines take 8-12 years. Hyperscalers are already making long-term commitments, with Microsoft restarting the Three Mile Island nuclear plant, Amazon acquiring a campus near a nuclear station, Google pursuing small modular reactors, and Meta seeking gigawatts of new nuclear capacity.
