Demand is outpacing supply, and the backlog keeps growing.
As artificial intelligence reshapes the demands placed on physical infrastructure, a quiet but essential player has emerged from the background of the data center world. Vertiv Holdings — maker of power systems, cooling units, and the thermal management equipment that keeps AI processors from destroying themselves — has seen its stock rise thirty-six percent this year, not because it builds the intelligence, but because it sustains the conditions under which intelligence can exist. In the long human story of technological transformation, it is often the suppliers of picks and shovels, not the miners, who endure — and Vertiv's swelling backlog suggests the market has begun to understand this distinction.
- AI processors generate heat at a scale that conventional data center cooling was never designed to handle, creating an urgent and unsolvable problem for every major technology company building at scale.
- Vertiv's liquid cooling systems — which pump coolant directly onto chips — have shifted from niche product to critical necessity, disrupting the assumption that software innovation alone drives AI progress.
- A $9.5 billion backlog growing at thirty percent year-over-year signals that demand is outrunning Vertiv's ability to manufacture, a rare and telling sign of structural rather than cyclical momentum.
- The company is navigating expansion across both AI training and inference phases, positioning itself to benefit as deployment infrastructure eventually dwarfs the current buildout in scale and geographic reach.
- At thirty-eight times forward earnings, the market is no longer pricing Vertiv as an industrial supplier but as a decade-long infrastructure anchor — a bet that the physical layer of AI is as valuable as the digital one.
When a hyperscaler installs a rack of the latest AI processors, it immediately confronts a problem no software can fix: the machines run extraordinarily hot and consume electricity at roughly ten times the rate of conventional servers. Air conditioning built for traditional data centers cannot cope. This is the opening Vertiv Holdings has stepped into — and why its stock has climbed thirty-six percent this year.
Vertiv makes the infrastructure that keeps data centers alive: uninterruptible power systems, power distribution units, and, most critically, liquid cooling systems that deliver coolant directly to the chips themselves. Without this equipment, packing enough computing power into a physical space to make AI training economically viable is simply impossible. Vertiv has become the leading supplier of these direct liquid cooling systems, offering hardware, specialized piping, and the ongoing maintenance the whole apparatus requires.
What gives Vertiv unusual durability is its independence from any single chip architecture or AI model. Every data center — whether training large models or running inference at scale — needs power management and thermal control. As AI shifts from training into widespread deployment, inference infrastructure alone could dwarf the current buildout, requiring a vast distributed network of facilities, all dependent on Vertiv's systems.
The financials reflect this structural position. Revenue grew twenty-nine percent over the past twelve months to reach $9.7 billion, with an eighteen percent operating margin. But the more telling figure is the backlog: $9.5 billion in unfulfilled orders, up thirty percent year-over-year, with a book-to-bill ratio of 1.4 — meaning new orders are arriving faster than the company can fulfill them.
The market has drawn its own conclusion. At roughly thirty-eight times forward earnings and a $61 billion market capitalization, Vertiv is being valued not as a cyclical industrial vendor but as essential, enduring infrastructure. The broader insight embedded in that valuation is simple: the AI opportunity does not live only in chips and software — it lives equally in the power, cooling, and reliability systems that determine whether any of it can be built and sustained at all.
When Google or Microsoft installs a rack of Nvidia's latest graphics processors to train an artificial intelligence model, they encounter a problem that no amount of software engineering can solve: the machines get unbearably hot, and they need staggering amounts of electricity to stay alive. A single AI rack can draw fifty to one hundred kilowatts of power—roughly ten times what a conventional server consumes. The air conditioning systems designed for traditional data centers simply cannot handle it. This is where Vertiv Holdings enters the picture, and it is why the company's stock has climbed roughly thirty-six percent this year while investors have begun to recognize what the company actually does: it is the plumbing and the nervous system of artificial intelligence infrastructure.
Vertiv manufactures the unglamorous but absolutely essential equipment that keeps data centers running. The company supplies uninterruptible power systems that prevent blackouts when the grid falters, power distribution units that safely channel electricity to densely packed equipment, and—most critically for the AI boom—liquid cooling systems that pump coolant directly onto the chips themselves. Traditional air cooling cannot dissipate the heat generated by thousands of processors working in parallel. Liquid cooling is not optional; it is the only way to pack enough computing power into a physical space to make AI training economically viable. Vertiv has positioned itself as the leading supplier of these direct liquid cooling systems, offering not just the hardware but the specialized piping, distribution units, and ongoing maintenance that keeps the entire apparatus functioning.
What makes Vertiv's position particularly durable is that its fortunes are not tied to any single artificial intelligence model or chip manufacturer. Whether the industry ultimately standardizes on one architecture or fragments across competing platforms, every data center—whether built for training massive models or running inference at scale—requires power management and thermal regulation. As AI moves from the training phase into widespread deployment, the demand for inference infrastructure will likely dwarf the current training buildout. Inference means running trained models millions of times per day to answer user queries, generate text, or make predictions. This will require a vast, geographically distributed network of data centers, all of them dependent on Vertiv's systems.
The company's financial trajectory reflects this structural advantage. Revenues have grown at an average annual rate of twenty-one percent over the past three years, and in the most recent twelve months they surged twenty-nine percent to reach nine point seven billion dollars. The operating margin sits at eighteen percent, the cash flow margin at sixteen percent, and the net margin at roughly eleven percent—solid but not extraordinary profitability. The real indicator of future growth, however, lies in the backlog. Vertiv is currently sitting on approximately nine point five billion dollars in unfulfilled orders, which has expanded thirty percent year-over-year. The company's book-to-bill ratio—the ratio of new orders to completed work—stands at roughly one point four times, meaning Vertiv is receiving significantly more new business than it can currently manufacture. Demand is outpacing supply.
This backlog provides exceptional visibility into revenue through 2026 and beyond. It also signals that the current surge in demand is not a temporary spike but the beginning of a sustained, multi-year expansion. The market has priced Vertiv accordingly. With a market capitalization of approximately sixty-one billion dollars, the stock trades at around thirty-eight times consensus earnings for fiscal 2025 and roughly thirty-one times earnings for 2026. These are not the valuations assigned to a cyclical industrial supplier that experiences periodic booms and busts. They are the valuations of a company that investors believe will be essential infrastructure for the next decade. The broader shift in how the market perceives artificial intelligence is embedded in this valuation: the opportunity is not merely in the chips themselves or the software that runs on them, but in the physical systems—the power, the cooling, the reliability—that determine whether AI can actually be built and sustained at scale. Vertiv is the company that makes sure the lights stay on and the machines do not melt.
Citas Notables
Vertiv is poised to benefit from AI growth, regardless of which model or chip ultimately sees more adoption.— Forbes analysis
La Conversación del Hearth Otra perspectiva de la historia
Why does Vertiv matter more than, say, the companies making the chips themselves?
Because chips are worthless if they overheat or lose power. Nvidia makes the processor, but Vertiv makes it possible to actually use the processor. The chip company gets the headlines; Vertiv gets the recurring revenue.
Is this a temporary advantage, or does Vertiv stay essential as AI matures?
It stays essential. Even after the training phase ends, inference—running models in production—will require vastly more data centers than training does. Every one of those facilities needs Vertiv's systems. The demand only grows.
The stock is up thirty-six percent this year. Is that already pricing in all the good news?
The valuation suggests the market believes this is a multi-year expansion, not a one-year spike. But the backlog—nine point five billion dollars—is real. That is money already committed. The question is whether demand stays this hot beyond 2026.
What could go wrong?
Manufacturing constraints. Vertiv is already receiving more orders than it can fulfill. If they cannot scale production fast enough, they lose market share. Or if AI deployment slows unexpectedly, the backlog shrinks. But right now, neither seems likely.
So you are saying Vertiv is a bet on AI infrastructure, not on any particular AI company or model.
Exactly. Vertiv wins if OpenAI wins, if Google wins, if Meta wins, if anyone wins. The infrastructure is agnostic. That is the real edge.