Grid access · resource ledger
Are AI data centers clogging the grid interconnection queue?
No. The median generation project completed in 2023 waited five years in interconnection queues, and about 2,600 gigawatts of proposed generation, over 95% zero-carbon, was still waiting; only 14% of studied capacity was ultimately built (LBNL, 2024). That backlog predates the AI buildout. New data centers instead trigger separate large-load studies. Grade: Scapegoated.
| Value | Measures | Source | Grade |
|---|---|---|---|
| 5 years | median request-to-operation time for generation projects completed in 2023 | Lawrence Berkeley National Laboratory (2024) | Scapegoated |
| ≈2,600 GW | of proposed generation capacity waiting in U.S. queues, over 95% zero-carbon | Lawrence Berkeley National Laboratory (2024) · Federal Energy Regulatory Commission (2026) | Background |
| 14% | of queued generation capacity in the studied cohorts ultimately built | Lawrence Berkeley National Laboratory (2024) | Background |
How does AI demand affect this market?
Generation queues study projects that supply the grid. Large-load processes study customers drawing power from it. They are separate processes with shared transmission impacts.
Who pays, and who benefits?
Delayed generators, large loads and ultimately customers pay for slow studies and network upgrades. Faster, better-managed queues reduce delays across grid projects.
What limits supply?
Study throughput, speculative projects, transmission planning, cost allocation and incomplete tariff rules for flexible or co-located loads.
What the evidence cannot show
The five-year generation queue predates the AI buildout. New data centers add separate large-load studies and transmission work.
What would change the grade
Comparable queue data showing AI-related large-load studies displace generation projects or increase completion times changes the grade.
Sources