AI Power Crisis: Delphi Digital Report Shows Energy as the Real Bottleneck
Delphi Digital and accelxr released a major report showing every AI power forecast has been too low, with data centers potentially consuming 15-17% of US electricity by 2030, and electricity — not chips — is now the binding constraint on AI growth.
What we know
The report, amplified by an account on X and multiple RTs from another account on X, covers generation (solar, wind, nuclear), transmission bottlenecks, storage (China, integrators, long-duration), and investment plays. Key stat: data centers used under 5% of US electricity in 2024; Berkeley Lab now projects up to 15% by 2030, EPRI up to 17%. Delphi's thesis: power takes years to build, so anyone who can deliver electricity faster than the grid has structural pricing power. This intersects with another account on X's report of OpenAI CEO Sam Altman saying: 'We always just want more compute' and that OpenAI is building robots to construct data centers.
By the Global Eye 24 desk, with AI assistance. Updated as the story develops. Spot an error? Write to corrections@globaleye24.com.