U.S. power consumption hit its second straight annual record in 2025 — and the EIA now projects it will climb further in both 2026 and 2027, driven by AI data centers and electrification. Two consecutive records, with a third and fourth on the way. This is not a blip. This is the curve bending upward.
The market is already pricing in the constraint. In Texas, developers are bypassing the grid interconnection queue entirely — co-located solar-plus-storage projects can reach data center customers in as little as 18 months, versus years-long waits for grid-connected alternatives. When the queue becomes the bottleneck, capital routes around it. That's not a workaround — that's the market telling you where the infrastructure gap is.
The federal response is taking shape too. The DOE has identified 16 potential sites on department lands for co-located data centers and new energy generation, with a target of commencing operations by end of 2027. Fast-track permitting, in-place energy infrastructure, proximity to national lab research capacity. The logic is sound: stop fighting the queue, build where the power already is.
And the buildout is going global. Convalt Energy agreed to spend $6.2 billion on a 1,200-megawatt hydro project and AI data center in Lesotho — a US firm planting electricity-maximalist infrastructure in southern Africa. The demand signal is so strong it's pulling capital to places the old energy map barely registered.
This is what the abundance thesis looks like in motion: record consumption, capital routing around bottlenecks, federal land unlocked for co-location, and a 1.2 GW hydro project in Lesotho. The curve is bending. The only question worth asking now is whether the grid can be built fast enough to stay ahead of it.