The rapid expansion of artificial intelligence is creating a major shift in how the United States produces and distributes electricity. What was once a relatively stable power system is now under pressure as data centres, driven by AI, demand unprecedented levels of energy.
Electricity demand in the US had been largely flat for years, but since 2020 it has started rising again, largely due to the growth of data centres. This trend is accelerating fast. Power demand from these facilities is expected to increase sharply in the short term and could nearly triple by 2030. AI infrastructure is fundamentally different from traditional computing, requiring far more energy at a much faster scale.
Utilities are being forced to react quickly. Major energy providers are increasing long term investment plans by tens of billions of dollars to expand capacity and upgrade infrastructure. In some regions, expected peak demand is arriving years earlier than forecast. The grid, originally designed for steady and predictable growth, is struggling to keep up with this sudden surge.
As a result, the energy mix itself is beginning to change. Natural gas is seeing renewed growth because it can deliver reliable, constant power at scale. At the same time, nuclear energy is experiencing a revival, with tech companies investing in reactors and signing long term supply agreements. The urgency for stable, always-on electricity is reshaping priorities across the sector.
One of the most significant developments is that large technology companies are no longer relying solely on public utilities. Instead, they are funding their own power infrastructure. This includes building dedicated plants, transmission lines, and even entire parallel systems sometimes referred to as a “shadow grid.” These private energy networks are designed to guarantee reliability and avoid delays tied to public grid limitations.
This shift raises broader concerns. There are increasing worries about grid stability, potential power shortages, and rising electricity costs for consumers. Some forecasts suggest supply gaps could emerge within the next decade, with risks of outages if infrastructure does not keep pace. At the same time, competition for equipment and resources is intensifying, pushing costs higher.
There are also environmental implications. While renewables remain part of the long term strategy, many new projects rely heavily on natural gas due to its reliability. This could slow progress on emissions reduction in the near term.
To manage demand, new ideas are emerging. One approach is to make data centres more flexible, allowing them to reduce power usage during periods of grid stress. However, this is difficult in practice, as downtime for AI systems can be extremely costly.
Overall, the AI boom is no longer just a technology story. It is becoming an energy story. The scale and speed of demand are forcing a complete rethink of how electricity is generated, distributed, and financed. The outcome will likely define not only the future of AI, but also the structure of the energy system itself.