Most “AI investing” talk is basically a chip conversation wearing a new hat.

Which is fine. Chips are real. Software is real. But what matters is the unglamorous middle layer that nobody brags about at a dinner table: the building, the cooling, the power distribution, the controls. All the stuff that turns a warehouse into something that can run hot servers 24/7 without melting down.

Comfort Systems USA (NYSE: FIX) sits right there.

Quick time anchor: this is from the quarter ended March 31, 2026.

Revenue: $2.87B (up 56.5% year over year). Diluted EPS: $10.51 (vs $4.75 a year ago). Operating cash flow: $388.8M (vs an $88.0M outflow in Q1 2025). And backlog: $12.45B as of March 31, 2026 (vs $11.94B at Dec 31, 2025 and $6.89B at March 31, 2025).

That backlog number is the one I keep coming back to. It is not perfect, but it is hard to fake demand when customers are reserving labor and capacity ahead of time.

And yes, management also called out 51% organic growth for the quarter. That detail matters more than it sounds like it should.

Here’s where I’m at with FIX.

The “AI buildout” is a real thing, but Wall Street loves to turn real things into over-owned trades. So I try to separate two questions that get mashed together:

• Is the end market strong?
• Is the stock priced like it can never have a bad quarter?

End market: hard to argue with it right now. Data center work has been a meaningful driver for the company, and backlog is still moving higher.

Stock: I’m skeptical by default when a contractor becomes a “must own.” Contractors are still constrained by people, scheduling, project timing, and a million little execution risks that do not show up in a one-line growth chart.

Slight tangent, but it matters: you can tell who has never managed a real-world build when they talk like growth is just a dial you turn. It is not. It is permits, equipment lead times, subs, weather, inspections, the customer changing their mind midstream. The fact FIX is posting numbers like this anyway is, in a weird way, the bullish case.

So what would make me change my mind, either direction?

The part people skip is that you do not need a macro call. You just need a few company-level signals.

• Backlog: does it keep building, or does it flatten out for a couple quarters?
• Cash generation: does operating cash flow stay strong over a full year, even with normal quarter-to-quarter swings?
• Labor: do they keep staffed up without giving away margins to do it?

If backlog starts slipping while everyone is still talking like data centers are unstoppable, that is usually when the stock stops being fun.

If backlog holds and cash flow stays solid, the stock can stay expensive longer than you want it to. That is just how these market phases work.

One more small thought.

People keep asking, “What is the next AI winner?”

Sometimes it is not “next.” Sometimes it is already there, doing work that sounds boring, and the only clue is a backlog line item that went from $6.89B to $12.45B in twelve months.