The AI buildout has a power problem
The first piece in this series argued the AI buildout is a financing story: $700-730 billion in 2026 hyperscaler capex, up 70-80% from last year, and money is not the constraint. This one is about the constraint that actually is. The industry is spending $700 billion a year to build data centers faster than anyone can power them. The demand is not theoretical. Lawrence Berkeley National Lab's…
The AI buildout faces a significant power constraint. The industry is investing heavily in data centers, with $700 billion annually spent to expand them at a rate faster than electricity generation can keep up. In the US, data center electricity use could reach 649 TWh by 2030, accounting for 11.8% of all electricity consumption.
Major tech companies are actively contracting for new power generation capacity. Microsoft, Amazon, and Oracle have signed long-term power purchase agreements (PPAs) with nuclear plants, including restarting Three Mile Island Unit 1 and purchasing electricity from Wisconsin's Point Beach plant. These contracts amount to more than 10 GW of new nuclear capacity.
However, most of these deals are options rather than actual power. Only a few projects are under construction, with most SMR (small modular reactor) deals still in development or at the licensing stage. Notable SMR projects include Oklo's Aurora unit, targeting commercial operation in 2028, NuScale's certified design but no firm build, and GE Vernova's BWRX-300 under construction in Ontario, with first unit slated for 2029.
Despite these efforts, the timeline for new nuclear power delivery remains uncertain, with most new electrons expected to arrive between 2029 and 2033, assuming no delays.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.