Inference Can Wait. The Grid Can't

The grid doesn’t have an energy problem; it has a peak-power problem. Chris Dunlap on why AI data centers face a choice between connecting now and flexing in the worst hours, or demanding firm power and waiting years—and which AI work can actually wait.

Chapters

1. 0:00 Is AI already paced by power?

2. 2:48 From F-18 backseater to the power grid

3. 4:15 Why is China out-building the US on power?

4. 6:10 How a Northern Illinois power crunch became the paper

5. 7:34 Why can’t the US connect power plants faster?

6. 9:56 What PJM is and why the queue is jammed

7. 12:50 Lawyers vs engineers: why regulators exist

8. 14:12 Energy vs power: why the peak matters

9. 15:45 If supply can’t catch up, can demand flex?

10. 19:16 A peak-power problem, not an energy problem

11. 21:47 Why flexible data centers still can’t jump the line

12. 23:25 Which AI work can wait? Urgency, surge and spot pricing

13. 28:35 Moving AI to wherever the power is

14. 32:04 What would make workload shifting possible?

15. 36:11 Who has to play ball: states, moratoriums, and your power bill

16. 41:22 Utilities, RTOs, and paying data centers to flex

17. 45:21 What to watch: will hyperscalers trade flexibility for speed?

18. 50:18 Can the US build fast enough to compete?

Timed transcript · Full show notes and sources

Sources

• Quantifying AI data center flexibility as a resource adequacy asset — Research Square (preprint)

• Quantifying AI data center flexibility as a resource adequacy asset (working paper) — The Institute for Climate and Sustainable Growth, University of Chicago

Disclosure

Adil Wali invests in startups. At 50:53 he mentions that startups he has invested in are moving workloads to Chinese open-weight models. Nothing in this episode is investment advice.