THE QUEUE PROBLEM
Every new generator or large load must wait in an interconnection queue while the grid operator studies whether the network can handle it. In the US, the queue has ballooned past 2,600 GW of pending projects — more than twice the entire installed capacity of the country.
WHY DATA CENTERS BREAK THE MODEL
Traditional industrial loads grow at 1–3% per year. Hyperscale AI campuses request 500 MW to 2 GW at a single site — the equivalent of a mid-size city — and want to energize within two years. The queue was never designed for either the size or the speed.
FIRM VS FLEXIBLE LOAD
A firm load expects power 24/7 at full nameplate. A flexible load agrees to curtail during system stress — peak hours, contingencies, heat waves. Studies show that allowing even 0.25% of annual curtailment unlocks roughly 100 GW of interconnection headroom on the existing US grid without new transmission.
WHY TRANSMISSION IS THE BOTTLENECK
A new high-voltage line in the US takes 10–15 years from proposal to energization — siting, environmental review, multi-state permitting, and right-of-way acquisition each consume years. Generators and loads can be built in 2–3; the wires between them cannot.
THE DEMAND RESPONSE PRECEDENT
Aluminum smelters, chlor-alkali plants, and bitcoin miners have priced curtailable contracts for decades. ERCOT's industrial demand response program in Texas routinely sheds 1–2 GW within minutes during system emergencies. Treating AI training as interruptible extends a model utilities already understand.
WHO BEARS THE RISK
Under a firm interconnection, the utility (and ultimately ratepayers) funds the transmission upgrade. Under a flexible interconnection, the data center accepts the curtailment risk in exchange for skipping the upgrade queue. The policy question is whether flexibility should be voluntary, incentivized, or mandatory for queue entry.