Edition 26, 22 September 2026. Prices, returns and drawdowns are at the 18 September close unless a date is given.
Last week I wrote that Bloom Energy sells a fast pass around the grid. Yesterday it joined the S&P 500. In the month before those two dates, the stocks I’ve spent three editions on had one of their worst months since the AI trade began, and the companies behind them reported the best order quarters in their histories.
I have been writing since March about the pressures on the grid forced by the data center buildout. That narrative suggests a strong market for all the companies contracted to make that buildout happen.
That is the subject this week. Not Bloom, though Bloom is in it.
The gap between the companies’ results in July and what their shares did in August and September, and what an investor is supposed to do with a gap like that.
I’ll give you the numbers first, then the explanation, then what my process did with it. The explanation is less flattering to the narrative than I’d like. The process did better than the narrative.
The gap
Take the four Edition 24 names and the 22 companies that supply them.
The 22 are the listed companies one step behind the four: the firms that make their components, build their plants, and deliver their fuel.
I picked them by reading their June-quarter filings for data center and power demand, and the table in the paid section says which of the four each one sits behind.
Twenty-six names. In July, every one that disclosed orders reported a healthy increase. No disclosed book-to-bill in the group fell below one.
GE Vernova’s Electrification business booked 1.7 times what it shipped.
Powell booked 3.0 times.
Comfort Systems added $1.6 billion of backlog in a single quarter.
Over the three months to Friday’s close, 23 of the 26 fell. The median drawdown from the 52-week high is 23%. Five names are more than 40% below their highs, and two more were until last week. The S&P 500 over the same window is up 2% and 2% from its own high.
The full table of all 26, with what each reported against what each did, is in the paid section. Here’s what it shows: the names whose prices held aren’t the names with the best order books. The name that rose most last week, though, did rise on an order, after I’d written the first draft.
The market read the July numbers and, in several cases, rewarded them on the day. EMCOR rose 19% on its print. Quanta rose 17%. Vertiv rose 25% in the five days after its release. Then the whole group was sold in August and again last Monday, and the July numbers didn’t matter.
So the question isn’t whether the order books are real. They are, and there are two soft spots in them that I’ll show you. The question is: why is there a disconnect between the order book and the stock price?
Narrative from above, narrative from below
Every stock in this trade has a story built from two directions.
From above comes the grid. Load is arriving faster than the system can connect it. The interconnection queue is measured in years. The turbine makers are sold out into 2028 and beyond. And in the towns where the campuses are supposed to be built, the opposition is organized. This is the macro context.
Edition 25 was mostly that story. That’s why the equipment is scarce and why the route around the grid exists.
From below come the filings. Orders, backlog, revenue, book-to-bill. The June quarter was the strongest set of those numbers the group has ever had.
The two stories meet in one place: the share price. And this quarter, the venue was like the East Wing of the White House – under construction.
I scored each of the 26 names for its exposure to the constraint from above, its order momentum from below, and what the shares did.
Until last Wednesday, no name on the list saw both narratives translate into price. One did, and I’ll show what it took. The names whose prices held, Williams and Kinder Morgan, held because they trade as pipelines, and pipelines went up.
Bloom’s price held for a month because an index committee put it in the S&P 500. Everything else with a constraint mechanism and accelerating orders fell between 9% and 58%, except Eaton, the largest of them, which rose 1%.
Which leaves one explanation standing. I explore that below the paywall.
