Vertiv

IThe business

A chip turns electricity into heat. Everything around the chip exists to get the electricity in cleanly and the heat out. That is Vertiv’s business. It makes the uninterruptible power supplies, switchgear and busbars on the power side, and the chillers, air handlers and liquid cooling units on the thermal side, then installs and looks after them.⁴ If a data center is a body, the chips are the brain and Vertiv sells the heart and lungs.

The company is older than it looks. Ralph Liebert started the business in 1946 and built the first computer room air conditioner. Emerson Electric bought Liebert in 1987 and grouped it into Network Power, sold it in 2016, and the new owner, Platinum Equity, named it Vertiv. It listed on 7 February 2020 by merging with a blank-check company chaired by Dave Cote, the former Honeywell chief.⁴¹⁰ The first close was $12.88.¹ For most of its history this was a steady, slightly dull supplier to banks and telephone companies. Sales were $3.9B in 2017 and $4.4B in 2019.¹³¹⁴

Then the buyers changed. Data centers were about 85% of sales in 2025, rounded by the company, with commercial and industrial customers about 10% and telecom networks about 5%. Power was 35% of 2025 sales, thermal 30%, services and spares 20%, IT systems 8% and infrastructure solutions 7%.³ The Americas were 62%, Asia Pacific 20% and EMEA 18%.⁴ This year the Americas is pulling further ahead: $3.0B of the $4.3B in the third quarter, up 75%.²

What a megawatt is worth

The cleanest way to size Vertiv is by the megawatt. Management puts its opportunity at $3.25M to $3.75M for each megawatt of IT power, up 10% to 15% on its earlier range, because it now sells more of the chain.³ It expects the industry to add about 140 GW over 2026 to 2030, with annual additions rising from about 20 GW to about 35 GW. Those are management’s own estimates.³

The arithmetic is simple. Twenty gigawatts at $3.5M is about $70B a year. If 85% of this year’s guided sales are data centers, Vertiv takes about $12.5B, or a little under a fifth C (calculated).³² At 35 GW and the same content, the same share gives about $22B in 2030. Add the other businesses and you are close to management’s plan. So the plan does not need a heroic share. It needs the gigawatts to arrive.

Management sizes its markets another way, at about $75B in 2025 after adding fluid services, converged infrastructure and DC power, growing 16% to 18% a year to 2030. Cloud and colocation, about $29.5B of that, is growing 23% to 25%.³ Those figures draw on Omdia, Dell’Oro and others, and on management’s judgement. I can’t check them independently. I use them as a ceiling on what is sensible.

How the sale works

Vertiv mostly sells direct, through about 3,000 salespeople, to the people who own and run the sites.⁴ Hyperscale clouds, colocation operators such as Equinix and Digital Realty, the newer AI clouds such as CoreWeave and Nebius, and large enterprises all buy.⁴ Management says its customer mix matches the market’s, and Cote says orders look strong with or without the neoclouds.⁵ The 10-K does not give a customer concentration figure that I could find, so I can’t tell you how much rests on the top five. The merger proxy said the top 50 customers were about 35% of 2018 sales.¹³ That is old, and I would expect it to be higher now.

A big order works like this. The customer reserves capacity 12 to 18 months ahead and often pays in advance. Vertiv records that money as deferred revenue, which was $4.12B at 30 September, up from $1.81B at the end of 2025.⁶ It builds, ships, commissions, and then, if it does its job, keeps a service contract for years. Services and spares were $719M in the third quarter, up 41.5%, or 18.1% before acquisitions.² Management says about three quarters of service sales are lifecycle work and a quarter project work.³

Why the order book went dark

Backlog was the number everyone watched. It went from $7.2B to $15.0B in 2025, helped by a fourth quarter with a book-to-bill of about 2.9 times, meaning orders of about $8.4B against sales of $2.9B C (calculated).⁴¹⁵ Vertiv stopped publishing orders and quarterly backlog after that report. On the third-quarter call Cote said the company does not plan to share orders data going forward.⁵ It gave one exception: third-quarter organic orders rose more than 100%.²

Some of the hidden number can be worked out. In the third quarter of 2025 book-to-bill was about 1.4 times on sales of $2.68B, so orders were about $3.7B.¹⁸ More than double that is more than $7.5B. Sales this quarter were $4.3B, so backlog rose by at least $3.2B in three months C (calculated). That is a floor, and an estimate, since it ignores currency and acquired orders. Large orders are lumpy, and I can see why a company would tire of explaining each quarter’s swing. Still, a shareholder now gets the key number once a year, in the 10-K.⁴

What counts as backlog also matters. Albertazzi says only a purchase order counts, and that a capacity agreement does not.⁵ The 10-K says the majority is firm and due within 12 to 18 months, that orders can be cancelled or rescheduled, and that customers can cut or defer firm orders in some cases, usually with a penalty.⁴ It is a strong signal, though a weaker one than a contract.

IIThe moat

The test of a protected business is whether it keeps its margins when things go wrong. Vertiv has had one clean test as a public company, and failed it. In 2021 and 2022 it had taken long orders at fixed prices, then copper, freight and chips went up. Adjusted operating margin fell to 7.7% in 2022 and adjusted free cash flow was −$260M, in a year when backlog rose 49%.³⁸ On 23 February 2022 the stock lost 37% in a day, from $19.57 to $12.38.¹ Shareholders are still suing directors over what was said about it.⁶ A business with strong protection would have repriced. This one was stuck with what it had signed.

It has fixed the process since. Today the 10-K still warns about long-term fixed-price contracts, but management reports positive price against cost in 2026, including tariffs.⁴⁵ What the episode shows is the kind of protection this is. Vertiv’s customers are large, expert and few. The 10-K says plainly that large customers often demand better terms.¹⁷

Look at gross margin, which is the best single read on pricing power. It was 33.8% in 2017 and 36.3% in 2025. For the first nine months of 2026 it was 37.9%, in the strongest demand this industry has ever seen C (calculated).¹³⁴⁶ Almost all of Vertiv’s margin gain has come from spreading fixed costs over more sales. That is good, and it is real, but it tells you the customer still has a voice in the price.

Are competitors closing in?

Vertiv names its rivals in the 10-K: Schneider Electric, Eaton, Legrand and Huawei at scale, and Delta, Stulz, Johnson Controls and Socomec as specialists.⁴ The strongest evidence that the field is healthy for everyone is that they are all growing fast. nVent’s organic sales rose 44% in its third quarter, with its largest segment up 54%.⁷ Trane’s backlog reached $12.5B, up 73%, with Americas commercial HVAC up about 80%.²⁰ Eaton’s Electrical sector backlog was up 43% in June, and Eaton has paid $9.55B for Boyd Thermal, $1.43B for Fibrebond, a builder of power enclosures, and $86M for a solid-state transformer startup.¹⁹²¹ When rivals spend that kind of money to enter your niche, it says the niche is good. It also says you will have company.

On that table Vertiv does not look out of line. It has the fastest growth and pays about two more turns of earnings than Eaton or Trane. The market does not treat it as a rare asset. It treats it as the best of a good group, and the group’s multiple is where the risk sits.

The technology could reset the game

The harder question is what happens when the rack changes. Liquid is taking a larger share of heat removal at the chip, though Albertazzi says air is not going away.⁵ Power is moving to 800-volt DC, which could remove parts of today’s AC chain. Albertazzi says 800-volt starts to bring revenue in 2027 and the full native version deploys in 2028.⁵ Cote’s line is that the more technology changes, the better it is for Vertiv, and so far that has been true: on its own estimate, the latest round raised content per megawatt by 10% to 15%.⁵³ But an incumbent’s edge is thickest in what already exists. A clean-sheet architecture gives every rival a fresh start, and Eaton has already bought its way into solid-state transformers.¹⁹ I don’t know how this ends. I would watch who is named on the first big 800-volt deployments.

IIIThe competitive advantage

Section II asked what keeps others out. This one asks what Vertiv does better today. There is one idea at the center of it: a customer’s most expensive problem is time.

AI compute costs millions a rack and earns nothing until it runs.³ Management’s phrase is time to first token. It claims that its work with NVIDIA on reference designs and factory-built modules can cut design time by a factor of ten and deployment time by up to half, and at one Norwegian site, adding up to 30% more AI output in the first 12 months.⁵ Those are company claims and best-case numbers. The direction is plain, though. If power and cooling can arrive as tested blocks from a factory, the customer needs fewer electricians, fewer trades and less time.

From the field to the factory

This is the part of the story I find most interesting. A traditional data center is built on site by many contractors. Vertiv’s OneCore and SmartRun products pull that work into its own plants, with up to 50% shorter deployment, by its own figures, and up to 25% lower lifetime cost.³ The effect on Vertiv is that it takes a larger slice of each megawatt, which is part of why its figure per megawatt has risen.³ And it turns a set of parts into a larger system that only a few firms can deliver.

It also brings the risk inside the factory. In the second quarter, sales of $3.27B missed expectations, and Vertiv blamed multi-phased project execution and temporary supply chain congestion.²² Management told investors these were the first very large projects with this level of complexity, as The Motley Fool reported it.²³ The stock fell 17.3% in a day.¹ Chamberlin says $100M to $150M of that slipped into the third quarter, which beat its guide by more than $560M.⁵²²² So the problem was real and it was temporary this time. But notice what kind of problem it was. Vertiv is learning to deliver a project like a product, and a mistake on a gigawatt site is expensive. On the October call management introduced Mike Resha as its new head of manufacturing and logistics; his first Form 4 is dated 5 October.⁵¹¹

Upstream, one acquisition at a time

Vertiv is buying its way along the chain. Great Lakes added racks in 2025. PurgeRite, $1.15B gross, added the flushing and fluid work that liquid cooling needs. In 2026 BMarko added modular structures, Strategic Thermal Labs added cold plate design, ThermoKey added heat rejection in Europe, and King Environmental took fluid services to EMEA.^624252627 Then on 27 October it closed UtilityInnovation Group, or UIG, for about $1.45B in cash and up to $1.15B more over two years if earnings targets are met.⁹ UIG designs microgrids and the controls that let a site run on its own power. Vertiv says this extends it from grid connection to chip, at about 13 times UIG’s expected 2027 EBITDA before the earn-out.⁹

Chamberlin’s reason is that UIG gets Vertiv to the table sooner, when the site’s power design is decided.⁵ That is a sensible argument. Power is the bottleneck, and whoever designs the site early can specify what follows. It is also a move into a young, crowded and fast-moving field, at a time when every asset near AI power carries a high price.

Where the advantage is thin

Analysts raised a new subject on the third-quarter call. Evercore’s Amit Daryanani spoke of investor concerns about product quality and share loss. Melius’ Scott Davis asked about talk of burned-out compressors and overheating UPS systems.⁵ I could not find a primary source for that talk, and it may be nothing. The quotes in this section come from the third-party transcript. Albertazzi did not deny early-life issues. He said a higher share of new products means more of them in that first phase, that the data are improving, and that “we are not perfect.”⁵ The current warranty accrual was $57.2M at 30 September, up from $43.2M at year-end, in line with sales.⁶ That is no alarm. But Vertiv’s advantage rests on the word reliability, so I treat this as the thing to check every quarter.

IVManagement

Two people matter. Dave Cote, 73, ran Honeywell from 2002 to 2017, then raised the blank-check company that bought Vertiv, and has been Executive Chairman since the listing.¹⁰ Giordano Albertazzi, 60, has spent more than 25 years in the business, ran Europe and then the Americas, became chief operating officer in October 2022 and chief executive on 1 January 2023.¹⁰ He took over at the bottom. Craig Chamberlin, the CFO, joined from Wabtec in November 2025 and is 45.²⁸

The record under Albertazzi is very good. Adjusted operating margin went from 7.7% in 2022 to a guided 24% this year, and adjusted free cash flow from −$260M to a guided $2.7B.³² In May he raised the five-year plan to 20% to 22% organic growth and a margin above 27%. Six months later the first year alone is running at about 38%.³² His own plan of November 2024 said 12% to 14% and about 25%.³ Management’s habit has been to promise less than it delivers.

Follow the pay

Albertazzi’s pay for 2025 was $18.3M: $1.25M of salary, a $4.0M bonus and $13.0M of stock options, valued at grant.¹⁰ It was $6.2M in 2023. His ratio to the median employee is 658 to 1.¹⁰ The size is unremarkable for this company’s value. What matters more is what the pay is for.

The annual bonus turns on adjusted operating profit and adjusted free cash flow, and for 2025 paid out at 135% of target.¹⁰ The long-term piece, Strategic Performance Awards granted in 2022, pays on adjusted operating profit alone. The targets were $1,000M, $1,400M and $1,750M for 2023 to 2025. Each $50M above target earned another 20% of the award, and the results were 120%, 160% and 220%. The shares cliff-vest on 1 January 2027.¹⁰ Annual equity is time-vested stock options.¹⁰ In 2026 the committee began a new award that pays on earnings per share instead: targets of $6.25 for 2026, $8.15 for 2027 and $10.55 for 2028, with an extra 20% of a bonus target for each $0.15 above target, and vesting in January 2030.¹⁰

So what does that pay them to do? Grow adjusted profit, keep cash coming in, and get the share price up. All three have been good for owners so far. But look at what the measures leave out. Adjusted operating profit excludes amortization of acquired intangibles, earn-out charges, and deal costs.² Adjusted free cash flow counts customer prepayments as if they were earned. Nothing in the pay I can find measures return on capital, and the proxy says the new EPS targets are not adjusted for acquisitions or buybacks, so a deal that adds EPS counts whatever it cost. A management paid this way, with a balance sheet this strong and $28B of planned capital to deploy, has every reason to buy things.³ The price paid for those things doesn’t touch its bonus. Its deals so far look sensible. Even so, that is the way the incentive leans.

How much of their own money is in

Less than you might think. At the April proxy date Albertazzi held 166,101 shares, about $39M at today’s price, and options on 1.46M more that he could exercise.¹⁰ On 8 September he gave 118,523 of them to a trust for his immediate family, leaving about 47,600 in his own name.²⁹ The stake stays in the family. I found no open-market sale by him since January 2024. Cote’s beneficial holding was 481,587 shares: 22,258 held by his wife and the rest options.¹⁰ All directors and officers together held 1.10%, options included.¹⁰ These are people with large option gains. They don’t own large stakes bought with their own cash, and my search of Form 4s since 2024 found no open-market purchase by anyone.¹¹ The 2024 warrant exercise by Cote’s sponsor vehicle, Cote SPAC I LLC, produced 4.8M shares.⁴ They are not in his proxy line, and I can’t say who holds them.

The insiders’ selling says more. In the days after the February results, directors and one senior officer sold heavily near $250 a share, which is above today’s price.

Directors have many reasons for selling, and these sales were a part of large gains. Still, Cote closed his October remarks by telling investors, as he has before, to act while supplies last, eight months after selling 40,000 shares at about $255.⁵¹¹ I put more weight on what people do with their own money. The ownership guideline is five times salary for the CEO and five times the cash retainer for directors, which is low.¹⁰

What they did with capital over a full cycle

The company has had one full cycle, and it is a mixed record with one very good decision in it.

Cash spent on acquisitions from 2021 to now, UIG included, comes to about $4.1B C (calculated). Against that the shares repurchased cost $0.6B and dividends have been small. That mix is reasonable while each dollar put into the business earns so much. On the third-quarter call, asked about buying back stock in a choppy market, Chamberlin said the most value lies in Vertiv’s own capacity and in acquisitions.⁵ A few days earlier the stock had been 48% below its May high. That answer told me where management’s instincts lie: expansion first.

In May management planned about $28B of capital to deploy from 2026 to 2030, of which about $4B is for dividends and buybacks and about $24B is unallocated, assuming leverage of 1.5 times.³ That compares with $12.5B in the 2024 plan. In this industry, at this point in the cycle, $24B of dry powder is the single largest judgement call this team will ever make. I would like it to be smaller, or to see more of it come back to owners.

One note on the chairman. Cote’s record at Honeywell is why many investors own this stock, and he has brought that operating system with him. He has also been Executive Chairman for over six years, and a good deal of the culture is his. How the handover ends is something no filing tells me.

VFinancials

The record divides into two companies. Before 2023 Vertiv was a low-margin, indebted carve-out. Since then it has been something else. Figures for 2017 and 2018 come from the merger proxy, and I start cash flow in 2019 because the audited record for a clean calendar year begins there. Vertiv’s predecessor had September fiscal years through 2016, so ten comparable years don’t exist.¹³

In 2026 so far, the first nine months brought net sales of $10.24B, up 39%, operating profit of $2.06B and net income of $1.65B.² The guide for the full year is $14.65B of sales, 24% adjusted margin, $7.15 of adjusted EPS and $2.7B of adjusted free cash flow.² GAAP EPS is guided to $6.29.² The Q3 tax rate was 23%, and management tells modelers to use about 24% from 2027.²³ That rise is one reason the path of adjusted EPS will lag operating profit.

Very little capital

This is the most attractive feature of the business, and it surprised me. Net plant was $1.32B at 30 September against trailing sales of $13.1B.⁶² Capital spending is guided at $575M to $600M this year, about 4% of sales, and that includes a large capacity build.⁶ Depreciation is running near $150M C (calculated). So most of the capital spending is for growth, and a business standing still would need little. An asset-light supplier into an asset-heavy boom is a good place to be.

Return on capital is therefore very high, though the number is distorted. Equity plus debt less cash is about $4.7B, and trailing adjusted operating profit is $3.0B C (calculated). Take out the customers’ deposits as well and the capital base is nearly nil.

Customers are funding the growth

Working capital is the thing to study here. In 2022 receivables plus inventory less payables and deferred revenue was 24% of sales. It was 10% in 2025. At 30 September it was 2%, or $303M C (calculated).¹⁷⁴⁶ That is a fine thing for a company to achieve. It comes from two sources: better collection, and customers paying ahead. Deferred revenue rose $2.31B in nine months. Inventory rose $1.81B and receivables $1.28B over the same period, and payables rose $1.41B.²

There’s a catch. That cash is real, but it is a loan from customers against goods not yet shipped. If orders flatten, prepayments stop arriving while old ones turn into sales. Free cash flow would then fall below earnings for as long as that lasted, just when an owner would like it most. Management’s own 2026 guide already has conversion at about 96% of adjusted net income, and targets 95% to 100%.³² The quarter also showed the turn in small: third-quarter working capital used $138M.²

What an owner can really take out

Reported and adjusted numbers are generous here for three reasons, so I worked out the cash an owner could take out and keep the business where it is.

The stock line deserves a word. Vertiv expensed only about $50M of stock pay in nine months, about 0.5% of sales. That sounds small, and it is the value at grant, when the shares were a fraction of today’s price. But the company’s own modeling note says to expect about 3M shares of new dilution a year from employee equity.³ At $232.50 that is $698M a year, about ten times what the income statement shows C (calculated). An owner who wanted to keep the share count flat would have to pay that. In my model I charge 2.5% of sales, about the cost of buying 3M shares a year near my value.

So owner cash this year is about $2.2B, or $5.65 a share. That is about 90% of GAAP net income, and 41 times at the price. It’s also a year of very fast growth, so it undersells what comes next. Reported nine-month cash tells a similar story from the other side: operating cash flow less capital spending and stock pay was $2,180M, of which $540M was working capital. Without that, owner cash was $1,639M, almost exactly the $1,652M of net income C (calculated).²

The balance sheet

Vertiv had $3.51B of cash, $199M of short-term investments and $2.95B of fixed-rate debt at 30 September, or about $0.76B of net cash, which is what its −0.2 times net leverage reflects.⁶⁵ That was before UIG. After $1.45B for UIG, $269M of accrued earn-outs, and my estimate of what UIG’s earn-out costs, plus guided fourth-quarter cash, I count about $1.3B of net debt for the base case E (estimate). The purchase price of King Environmental has not been disclosed. Bonds run to 2066, with $850M due in 2028.⁶ Bank guarantees were $415M.⁶ This balance sheet can take a bad year. The 2022 version could not have done so as easily. Moody’s rates it Baa3 and S&P BBB−.³³

One small unbooked item. In February the Supreme Court invalidated the IEEPA tariffs, and refund claims opened in April. Vertiv says it will recognize any refund only when realized.⁶ It gives no amount, so I give it no value. It could be a pleasant surprise of unknown size.

VIValuation

First, how this goes badly wrong

Before any numbers, I asked what would most likely ruin this investment, and worked back from there. Three things can.

Customers stop building. Vertiv’s sales are about 85% data centers, and the money comes from a small group of very large buyers and a set of newer ones that borrow to build.³ The four largest spent $496.7B on capital goods in the first nine months of 2026 and Meta alone guides to $135B to $145B for the year.³⁴³⁵ In mid-October Vertiv fell 14% in a week without bad company news, reportedly as yields on CoreWeave’s 2032 bonds neared 15%.³⁶ Orders can be deferred, and prepayments stop when orders do. A pause would mean falling sales, new factories half used, and cash flowing back out.

A project goes wrong at scale. It has happened. In 2022 fixed-price contracts cut the adjusted margin to 7.7%.³ This year a much smaller slip cost the stock 17% in a day.¹ Vertiv now delivers projects far larger than anything in its past, with new products, new factories and a new head of manufacturing, and it provides performance guarantees.⁴⁵ One serious failure at a gigawatt site would harm the word it sells most, which is reliability.

The money goes in at the top. Management plans to deploy about $24B beyond dividends and buybacks through 2030.³ It counts recent deals at about $3B to $4B, before UIG, and all at prices set in a boom.³ Booms in capital goods tend to end with too much capacity bought too late.

These three are linked. A pause in AI spending would arrive at the same moment that new factories and new deals are at their largest. That is my bear case, and it is why I put a down year in my base.

What the price implies

At $232.50 the market values Vertiv at $91.3B on 392.7M diluted shares.²¹ That is 32.5 times this year’s adjusted EPS and 25.4 times Nasdaq’s 2027 consensus of $9.14, which rests on eight analysts. The consensus for 2028 is $12.31 and for 2029 $14.32, the last from two analysts.¹² On those numbers the stock doesn’t look wild.

The owner’s view is colder. I take my base cash through 2030 and ask what the business must earn after that. Those four years of cash, discounted, are $10.3B. The price pays for $82B more. That needs $11.5B of owner cash a year from 2031, growing 3% for good C (calculated). My base reaches only $6.9B by 2036, and this year’s owner cash is about $2.2B.

Put another way, on my sales the settled margin would have to be 44%, which no one in this industry earns. Or growth would have to run about 1.8 times my base in every year, which takes 2030 sales to about $46B, above my bull. So the market is pricing something beyond my bull, not the plan. The plan itself, the one management set out in May, gets you to about my base value.

What I can’t judge

Three inputs matter most, and I can’t know any of them. I can’t tell you when the AI buildout digests. Nobody can. I have put it in 2031 because a date has to go somewhere, and my stress case shows what an earlier and deeper pause does. I can’t see the backlog, since the company no longer shows it. And I can’t tell you what margin this business earns once capacity catches up with demand, because that has never happened at this scale.

So here is the range on the margin, and read my $123 with it in mind.

The discount rate

I use 12.0%. The capital asset pricing model starts with the 4.94% ten-year Treasury yield in Q4 2026 and Damodaran’s 4.20% implied equity risk premium of 1 October.³⁷³⁸ My beta on month-end returns since the February 2020 listing, 80 months, is 1.83, which gives 12.6%. The five-year beta of 2.10 gives 13.8% C (calculated).¹³⁹ I sit just under that range because the downturn is already in my cash flows, and I don’t want to charge for it twice. The rate is above the 10.0% on my NVIDIA page and 10.5% on my Broadcom page, which earn far more on each dollar of sales and have steadier cash.

Three cases

The model is simple on purpose. Each year has a sales growth rate and an adjusted margin. I take stock pay at 2.5% of sales in place of the 0.5% already in adjusted profit, tax at the 24% management advises for 2027 to 2030, capital spending at 3.5% of sales, the middle of its 3% to 4% range, and depreciation rising from 1.2% to 2.5% of sales.³ Working capital rebuilds from 2% toward a set share of sales. My 2027 margin of 24.5% sits below the 25.5% to 26.5% that management’s 30% to 35% incremental framework gives on my sales. I take the lower figure because new capacity costs margin while it ramps, as Chamberlin said, and because the Nasdaq consensus of $9.14 implies something similar.⁵¹² I count no future acquisitions, in sales or in cost. I reconcile the base with the May plan like this: about $26B in 2030 becomes about $27.7B after this year’s beat, and about $29B to $30B with what Vertiv has bought since E (estimate). That is my $29.5B.

The weighted value, 25%, 50% and 25%, is $140.

There is also a stress case with no weight: a pause in 2028, sales down 35% in 2029, a margin of 9%, near the 2022 low, and working capital back to 20% of sales. It is worth $24 a share. It sits outside the weighted value and outside the margin-of-safety price. I include it because this company has been there once, on a much smaller stage.

Why the market may be right

I want to be fair about what the market sees. A ten-year cash model with a 3% tail, at 12%, leaves Vertiv in 2036 at about 12 times owner cash. Very good industrial companies rarely trade that low. If Vertiv in 2029 earns my $14.28 and still fetches 22 times, about what Eaton and Trane fetch today, that is worth $224 today at 12.0% E (estimate). So the market’s case is simple: Vertiv keeps today’s multiple through a further three years of growth. I also tried to be kind to my own inputs: no down year in 2031, a 27% margin for good, a 10.5% rate, 4% growth after 2036, stock pay at 1.5% of sales and working capital at 5%. That gives $213 a share E (estimate), still under the price. If the multiple holds, $232.50 is about fair. My model says that multiple can’t last forever, and that it would fall quickly at the first sign of a pause. Vertiv’s five falls of 30% or more since listing suggest the market agrees on the second point.¹

Compared with what?

The first alternative is the Treasury at 4.94%, and my base return of 3.8% a year doesn’t beat it. That is a harsh test, and the return uses a 12% rate for the year-ten value. But a business that gives a Treasury-like return in my central case, with far more risk, hasn’t paid me for the risk.

The better test is the best nearby use of the same dollar. For someone who wants exposure to AI building, my own pages give other options.

On my numbers NVIDIA offers more than twice the base return, and it sits at the point in the chain where the most profit is made. The risk is the same pause in AI spending, and its bear case is also bad. Vertiv’s claim against NVIDIA is that power and cooling are needed whichever chip wins, and that Vertiv needs little capital. That’s a real claim. At these prices I don’t think it’s worth paying more for.

Quality and price are different questions

How good is the business? Very good, and much better than its history. It has a real franchise in service, a place at the center of a large buildout, and little need for capital. How good is the price? That is a separate matter. At $232.50 a buyer pays for the plan, then for much more than the plan. This looks to me like a fine business at a full price. The fall of 48% from May to October took out a lot of excess, but the stock has risen 20% from that low and the excess isn’t gone.

The margin-of-safety price

My margin-of-safety price is $58, the price at which my $44 bear value is three quarters of what you paid, the same rule as my other pages. At $58 my base returns 23% a year and my bear 8.2% E (estimate). The stock last closed there in January 2024, so it describes a collapse, and I don’t expect to see it without a serious reason. There’s no margin of safety at $232.50.

Where this could be in 5 and 10 years

These are my model’s values at later dates. Each is the present value, at 12.0%, of the owner cash that comes after that date, less today’s net debt held flat, per today’s 392.7M shares and in that year’s dollars. I count no future deals. Dates are 31 October.

For the base to stand at 5 years, sales have to reach $29.5B in 2030 at a 27% margin, then take one down year in 2031 and recover. To stand at 10 years, sales reach $41.4B in 2036 at a 26% margin, which is $6.9B of owner cash. Value then grows about 4% a year. The return is 3.7% because the price starts $110 above my base value.C (calculated)

The bear has the AI build pausing in 2029. Sales fall to $13.1B, the margin to 16.5%, and working capital goes back to 16% of sales. The bull has no pause, with sales of $79.4B in 2036 at a 30% margin. Sales growth, the settled margin and working capital drive the gap. Two points of margin move my base by $12 a share today. The 12.0% rate comes next.

The most likely way my base is wrong is the margin. I have it settling at 26%, above the 20.4% of 2025, because rivals will add capacity and new plants cost margin while they ramp. If it settles at 22%, my base is $99 today, $140 in 2031 and $175 in 2036.E (estimate)

These are estimates of value, not forecasts of the share price. The market price can sit above or below value for years.

VIIThe case for the price

What most people miss

Three things cut in Vertiv’s favour and get little notice. First, the hidden order book can be roughly rebuilt. Third-quarter orders were more than double a year earlier, when book-to-bill was about 1.4 on $2.68B of sales, so backlog grew by at least $3.2B in the quarter in which the stock hit its low C (calculated).²¹⁸ Second, the May plan is already out of date. It aimed at about $26B of sales in 2030. After this year’s beat, that needs only about 15% a year from here C (calculated).³² Third, the customers are paying for the growth. Deferred revenue of $4.12B means Vertiv carries almost no working capital, which is why so little of each dollar of growth needs new money.⁶

The bull case

In plain words: power and cooling are now the limit on how fast AI gets built, and Vertiv is the one firm that sells all of it, services it everywhere, and can build it in a factory. Albertazzi says we are in the middle of the largest infrastructure buildout, and that it is accelerating. He expects “significant above-market” organic growth in 2027 and further margin gains.²⁵ Cote says the gap with everyone else keeps widening.⁵ Management’s plan has 20% to 22% organic growth and a margin above 27% by 2030, and its record is of beating its own plans.³

The sell-side is warmer still. TD Cowen’s Lance Vitanza called both of the May targets conservative, and RBC’s Deane Dray expects organic growth of at least 20% to 25% a year through 2030, as Yahoo Finance reported in May.⁴⁰ In October Morgan Stanley cut its target to $330 from $350, as reported by TIKR.³⁶ Among the investors I checked, Coatue held 806,594 shares at 30 June, down from 981,633 in March.⁴¹ Altimeter, Atreides, Duquesne, Appaloosa, Situational Awareness and Berkshire held none in their last three 13Fs.⁴² I could not find a public argument about Vertiv from Brad Gerstner, Gavin Baker, Cathie Wood or Bill Ackman, so I don’t cite one.

Against reality

The bull case is right about the present. Sales grew 61% last quarter, margins rose, and cash came in. If Vertiv does what it says, my base follows it, and that is worth $123. The price needs more. On my model it needs sales of about $23B in 2027, $33B in 2028 and $46B in 2030, above my bull, or a settled margin near 44%, or a market that keeps paying today’s multiple through the next pause.

It helps to multiply it out. Say AI building keeps rising through 2029: 60%. Vertiv converts its orders without another serious slip: 80%. It keeps its share through the move to 800-volt power: 70%. Margin reaches 28% or more despite rivals adding capacity: 50%. And the buildout runs past 2030 without a down year: 50%. Those are my guesses, and they aren’t independent. Multiplied, they give 8% E (estimate). Allowing for the overlap, perhaps 20% to 25%. That is about the weight I give the bull, and well short of the 75% on the bull that the price implies. Each step looks likely enough. All five together do not.

What a 10x would take. Ten times $232.50 is $2,325, about $913B of market value. At 25 times earnings Vertiv would need about $37B of net income. At an 18% net margin, close to this year’s adjusted level, that means about $203B of sales. Reached by 2036, that is 30% a year for ten years from this year’s $14.65B C (calculated). Management’s own estimate of its whole served market, after the recent additions, is about $75B in 2025, growing 16% to 18% a year, or about $160B to $170B in 2030.³ So a tenfold price needs Vertiv’s sales to exceed its entire served market today, and most of that market by the mid-2030s, against Schneider, Eaton and the rest. Tenfold runs from about $90B have happened. NVIDIA did it from August 2019 to May 2023, and Broadcom from March 2020 to December 2024.⁴³⁴⁴ Both make chips with gross margins above 60%. Vertiv’s is 38%. I know of no maker of electrical and cooling equipment that has done it from this size. From here, tenfold is close to impossible.

Where I come down

The cash discipline moved me, and so did how little capital the business needs: without them my base would be well below $123. The orders news did not move my values much, because I can’t see the backlog and my base already follows the plan. Two years of margins above 27% with backlog still rising would raise my values most.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• Customer concentration. The 10-K warns that fewer, larger customers mean any one of them can hurt results, and that they demand better terms.⁴¹⁷ It doesn’t say how large the largest are.
• Neocloud credit. Newer AI clouds borrow to build. If their funding closes, their orders and prepayments go with it. Management says orders are strong without them, but doesn’t show the split.⁵
• Disclosure. No quarterly orders or backlog. The third-party transcript I used has errors, and the company’s investor site blocked my reader, so I could not confirm every line against the webcast.⁵
• Quality. Analysts asked on the call about failing compressors and overheating UPS units. Management described early-life issues on new products as normal and improving.⁵ A recall or a large failure would cost far more than its direct expense.
• Tariffs and trade. Vertiv makes and services equipment at about 300 locations in more than 40 countries.⁴ It reports positive price against cost in 2026, including tariffs.⁵ The IEEPA ruling helps; new tariffs could come under other laws.⁶
• Technology. 800-volt DC power and liquid cooling at the chip change what a data center needs. Vertiv expects 800-volt revenue in 2027 and full deployments in 2028.⁵ A rival winning the first large orders would hurt share at the wrong moment.
• Acquisitions. Up to $1.15B of UIG earn-out and $250M at PurgeRite depend on targets, and $24B of capacity is unallocated.⁹⁶³
• Legal. A securities class action filed in 2022 over statements about inflation and pricing is still pending in part, and derivative suits filed in 2023 and 2024 are pending, the first stayed.⁶ The 10-K also reported an SEC subpoena and a parallel request from the US Attorney in New York received in November 2023. The latest 10-Q does not mention them.⁴