Every projection of AI growth eventually collides with a physical fact: data centers need enormous, uninterrupted, around-the-clock power, and the grid was not built for a load that doubles in a few years. The chip shortage of the early 2020s is giving way to a power shortage—and it is reshaping how the largest technology companies think about energy.
A load the grid did not plan for
The International Energy Agency projects that electricity demand from data centers will roughly double by 2030, to about 945 terawatt-hours—close to 3% of global consumption and growing about four times faster than overall electricity demand. In the US, data centers are expected to drive nearly half of all electricity demand growth to the end of the decade; the country will soon use more power for data centers than for making aluminium, steel, cement, and chemicals combined.
The problem is not only total energy but where and when. AI clusters want gigawatts in specific locations, delivered continuously, on timelines measured in months—while new transmission and generation take years. That mismatch is the bottleneck.
Hyperscalers become power buyers
Unwilling to wait for utilities, the largest AI operators have started to secure generation directly. One signed a twenty-year contract to restart a shuttered reactor at Three Mile Island. Another contracted for nearly two gigawatts from an existing nuclear plant to feed adjacent data centers. A third signed the first corporate deal to buy power from small modular reactors; a fourth lined up more than six gigawatts of nuclear across several developers.
The common thread is a preference for firm, carbon-free, always-on power—exactly what nuclear provides and what intermittent renewables, on their own, do not. AI has done more to revive nuclear economics in two years than a decade of policy did.
Bridges, not silver bullets
Nuclear is not a near-term fix. Restarts and existing plants can add capacity this decade, but small modular reactors will not arrive at scale until around 2030. In the interim the load is met by a messy mix: natural gas turbines (with long equipment backlogs), grid upgrades, batteries, behind-the-meter generation, and demand-response deals that let data centers curtail during peaks.
Each of these is its own investable value chain—turbines, transformers, switchgear, transmission, storage—and each is now sold out well into the future. The scarce commodity has quietly shifted from silicon to interconnection.
What to watch
For anyone allocating capital or planning capacity, the useful questions are physical, not financial: where is firm power actually available, how long is the local interconnection queue, and who controls the transformers and turbines on the critical path.
The AI story is increasingly an energy story. The winners will be decided as much by megawatts as by models.
Sources
Disclaimer. This report is produced by Dendra Capital for informational purposes only. It reflects views as of the date of publication, draws on sources believed to be reliable but not guaranteed, and is subject to change without notice. It does not constitute investment research, or financial, legal, or tax advice, nor an offer or solicitation to buy or sell any security. Past performance is not indicative of future results.