Vertiv's 10x argues the AI infrastructure trade runs through the air conditioning

The mechanism is physical and unglamorous. Putting Nvidia GPUs and similarly dense gear into a room requires cooling of a kind data centers did not previously need, and Vertiv's direct-to-chip liquid systems are doing more of that work than any competing approach, according to Fortune ^[15]. That is a supplier position rather than a technology bet. The customer's design decision creates the requirement, and the requirement has to be bought from someone.

Compound the two annualized figures out over the same six-and-a-half-year window and the distance between the accelerator and the air conditioner is narrower than the framing suggests: Nvidia becomes roughly 17x, Vertiv roughly 14x, so the chipmaker returned about 1.2 times what its customers' cooling vendor did ^[12]. An allocator who never formed a view on which accelerator would win, and bought the thing every accelerator needs instead, kept most of the compounding without needing to be right about silicon.

The two headline numbers do not quite reconcile, and it is worth saying so. The market-cap move implies about 42% a year compounded, some eight points below the annualized return quoted beside it ^[13]. Gaps of that shape usually come out of share count, dividends, or a start date that does not sit exactly where the market-cap base does, and the word "under" in "under $11 billion" is carrying weight ^[4]. The direction is not in question. The precision is looser than a figure quoted to one decimal place implies.

What is transferable here is the filter rather than the outcome. Cote told Fortune he held weekly calls with a Goldman Sachs group and that they looked at more than 1,000 companies before the bank surfaced Vertiv ^[2]. His criteria predate any interest in AI: a dominant or potentially dominant position, an industry already profitable or on its way there, room to grow internally and by acquisition, scope for wider margins, and the chance to lead a new technology in the sector ^[8]. He had already run that screen through more than 100 acquisitions at Honeywell ^[7]. "It was not like I was an AI savant," he says, "but I knew that if I positioned myself in the middle of the data industry there was a good chance something good would happen" ^[9].

Fortune's read is that Cote noticed a large trend and made himself indispensable to it rather than chasing a hot technology ^[11]. The screen that found an air-conditioning company in 2019 is now published, and the supply of dominant-position industrials sitting next to a data center is finite.