Speculative load, not real demand, inflates the AI power forecast.
Interconnection queues are stuffed with duplicative, unfinanced projects — Ohio's take-or-pay collateral pruned a 30,000 MW request pipeline to a fraction of signed commitments, Texas froze its entire queue to audit a 367 GW forecast, and PJM's large-load forecast halved in three months — so grid plans and capacity procurements sized off raw forecasts are systematically oversized.
The same conclusion keeps arriving from across the workspace's research — 2 topics independently instantiate this theme. Filter the evidence by where it came from:
Take-or-pay minimums and collateral requirements pruned the 30,000 MW speculative request queue down to 17,861 MW of binding, financially backed contracts — speculative load deflating on contact with financial requirements.
A state now freezes its entire interconnection queue to audit the load forecast itself, because the 367 GW pipeline is stuffed with unverified, speculative demand rather than contracted load.
Even the backstop procurement's own target rests on a load forecast that halved within three months as speculative projects were culled, with no mechanism to shrink alongside it.
Interconnection requests at 10x estimated actual use, ERCOT facing 194 GW against double its capacity, and only 20 of 100+ proposed Texas facilities applying for permits show the load pipeline inflating forecasts is largely speculative.