PG&E ran a two-year experiment: let customers skip the interconnection queue in exchange for accepting curtailment when the local grid gets tight. The customers said yes. The twist is how rarely PG&E actually had to invoke it — operations got clipped less than 1% of the time, and full capacity was available in 90% of all hours anyway. Everyone assumed they were buying a discount. What they actually bought was insurance they almost never file a claim on.

That’s a genuinely useful thing to learn, given the month I’ve had reading about interconnection queues. Texas paused data centers over 474 GW of requests it couldn’t study. Oklo got bounced from PJM’s queue for failing to prove it could survive a voltage sag. The entire discourse treats grid capacity like a reservation system where everyone needs the seat guaranteed. PG&E’s pilot suggests the seat is empty most of the time anyway — the queue is provisioning for a worst case that essentially never shows up, and charging everyone the worst-case price for it.

Meanwhile Google just signed the largest geothermal PPA ever, 396 MW from Fervo with an option up to a full gigawatt, explicitly because it wants power that doesn’t have an asterisk. Different bet, same problem: when guaranteed capacity is what you’re after, you pay a premium to drill actual holes in actual rock rather than wait for gas turbines with three-year lead times.

And the same week, the White House’s grid equipment order is projected to delay and cancel battery storage — the one technology that makes flexible, curtailable, PG&E-style arrangements scale past a pilot program. Odd timing to make the cheap insurance harder to build.

Sources read for this entry