Vertiv GB300 Liquid Cooling Backlog Analysis: Why a $15 Billion Order Book Points to 26% Upside in the AI Power Trade

Table of Contents

When investors think about the artificial intelligence build-out, the conversation almost always starts and ends with the chip. Nvidia designs the accelerator, TSMC fabricates it, and the hyperscalers buy it by the tens of thousands. But an AI accelerator that cannot be powered and cannot be cooled is an expensive paperweight. As rack densities climb from a comfortable 10–15 kilowatts toward the 142 kilowatts that Nvidia’s GB300 NVL72 platform demands, the physical infrastructure that surrounds the chip — the power distribution, the uninterruptible supply, the switchgear, and above all the liquid cooling — has quietly become one of the tightest bottlenecks in the entire AI supply chain.

Vertiv Holdings (NYSE: VRT) sits directly on that bottleneck. The company is the closest thing the public market offers to a pure-play on data center power and thermal management, and it has just told the market that its combined order backlog reached $15.0 billion, a figure management says gives it revenue visibility all the way into 2027. Trading at $300.53 with a consensus 12-month price target of $378.23, the stock offers roughly 26% upside to the average Wall Street estimate even after a ferocious 2026 run that took it from a 52-week low of $110.06 to a high of $379.93.

This article makes the case that the market is still underpricing three things. First, Vertiv’s deepening co-engineering relationship with Nvidia — most recently the cooling and power reference architecture for the GB300 NVL72 platform — is not a one-off press release but a structural moat that gets reinforced with every new accelerator generation. Second, the thermal management segment, and specifically direct-to-chip liquid cooling, is inflecting from a niche product line into the fastest-growing category in the entire data center capital stack, growing at a 31% compound annual rate by one industry estimate. Third, the recurring services business attached to an ever-larger installed base is a margin and durability story that the “it’s just a cyclical industrial” bears consistently miss.

We will walk through Vertiv’s business model and segment mix, spend the most time on the industry structure that governs its future, dissect the economic moat, pressure-test the financials, build a valuation with explicit bull/base/bear scenarios, lay out the real risks, and finish with a concrete rating and exit plan. The core question is simple: is a company earning a 45% return on equity, growing orders at triple-digit rates, and holding the top share in the single most supply-constrained corner of AI infrastructure worth 34 times forward earnings? Our answer is a qualified yes.

1. Company Overview

Vertiv designs, manufactures, and services the critical digital infrastructure that keeps data centers, communication networks, and commercial and industrial facilities running. In plain terms, the company sells the equipment that delivers clean, uninterrupted power to computing hardware and removes the enormous quantity of heat that hardware generates. It is a “picks-and-shovels” business: Vertiv does not care whether the winning AI model comes from OpenAI, Anthropic, or Google, because every one of those models runs on racks that need Vertiv’s power and cooling to function.

The revenue engine breaks into three broad offering categories. Power management — the largest — includes uninterruptible power supplies (UPS), intelligent rack power distribution units, switchgear, and busbar trunking systems that carry electricity from the utility feed to the server. Thermal management — the fastest-growing — spans precision room cooling, chillers, coolant distribution units, and the rapidly expanding direct-to-chip liquid cooling portfolio. Integrated rack solutions and services bundles modular and prefabricated data center systems together with the installation, monitoring, maintenance, and spare-parts business that recurs over the multi-decade life of the installed base.

Because Vertiv reports geographically as well, its revenue splits across the Americas (its largest region), Asia Pacific, and Europe, Middle East & Africa. The approximate offering mix below illustrates where the money comes from; readers should treat the split as indicative rather than audited to the decimal.

Offering category	Approx. share of revenue (est.)	Growth character
Power management (UPS, switchgear, busbar, rPDU)	~45%	Steady, AI-accelerating
Thermal management (air + liquid cooling, CDUs)	~30%	Fastest-growing
Integrated rack / modular solutions	~12%	Cyclical, project-driven
Services & spares	~13%	Recurring, high-margin

Vertiv’s customers are the hyperscale cloud operators, colocation providers, enterprise data center owners, and increasingly the “neocloud” GPU-as-a-service firms such as CoreWeave, with whom Vertiv has publicly described joint infrastructure work. On market position, the third-party research firm MarketsandMarkets identifies Vertiv and Schneider Electric as the two leading players in data center liquid cooling, and reported that Vertiv held over 11.3% share of that market in 2025 — the top position in a fragmented field. Ownership is dominated by institutions, a legacy of the company’s 2020 public listing via a Platinum Equity-sponsored transaction; insider ownership is modest, and the float is highly liquid.

Trailing twelve-month sales stand at $10.84 billion with net income of $1.56 billion, and the company generates an exceptional 45.1% return on equity — a figure we will scrutinize in the financial section, because such a high ROE can reflect either genuine capital efficiency or an equity base thinned by buybacks and leverage.

2. Industry Analysis

2-1. Market Size & Growth Trajectory

The market Vertiv addresses is best understood as two overlapping layers: the broad data center power and thermal infrastructure market, and the far faster-growing subset of liquid cooling within it.

Start with liquid cooling, because that is where the inflection is sharpest. Independent estimates vary in magnitude but agree on the shape. MarketsandMarkets projects the data center liquid cooling market growing from roughly $4.07 billion in 2026 to $27.65 billion by 2033, a 31.5% compound annual growth rate. Grand View Research pegs a similar endpoint — around $29.5 billion by 2033 — off a larger 2025 base of about $6.65 billion, implying a still-brisk ~20% CAGR. The dispersion in these numbers is itself instructive: analysts know the direction is up and to the right, but the terminal size is genuinely uncertain because it depends on how quickly rack densities force air cooling into obsolescence. For a supplier with the leading share, uncertainty skewed to the upside is a favorable place to sit.

The reason liquid cooling is inflecting now is physics, not fashion. Air cooling becomes thermodynamically and economically impractical somewhere around 30–50 kilowatts per rack. Nvidia’s GB200 and GB300 NVL72 systems push rack density toward and beyond 100 kilowatts, and the GB300 reference architecture Vertiv co-developed targets densities up to 142 kilowatts per rack. At those densities, direct-to-chip liquid cooling is not an optional efficiency upgrade — it is the only way to keep the silicon within its operating temperature envelope. That converts cooling from a discretionary line item into a mandatory, spec-driven component of every new AI cluster.

The industry sits in the acceleration phase of its cycle, not the early-adoption phase and not yet maturation. Adoption has moved past pilot projects into standardized, at-scale deployment, but penetration of liquid cooling across the global installed base remains low, leaving a long runway. That combination — proven demand, mandatory adoption at the technology frontier, low base-rate penetration — is the profile you want in a growth-stage capital equipment market.

2-2. Structural Growth Drivers

Driver one: rack density is rising faster than anyone forecast, and it is a one-way ratchet. Each new accelerator generation packs more transistors and draws more power, and the networking topology of AI training — where thousands of GPUs must communicate as one machine — rewards cramming compute physically closer together. That pushes power-per-rack up relentlessly. Vertiv benefits twice: higher density means more power delivery equipment per rack, and it means the cooling method must shift from cheap air handlers to higher-value liquid systems, coolant distribution units, and manifolds. The dollar content of Vertiv equipment per megawatt of compute is rising, so Vertiv’s revenue can grow faster than raw data center square footage. This is a multi-year, structurally reinforcing dynamic rather than a one-time step change.

Driver two: the shift from “buy a box” to “buy a reference architecture.” Historically, a data center operator bought cooling and power components à la carte and integrated them itself. The complexity of 142-kilowatt racks has broken that model. Operators now want a validated, pre-engineered power-and-thermal architecture that is guaranteed to work with a specific accelerator platform out of the box, because a design mistake at these densities can literally melt hardware or delay a multi-billion-dollar buildout by months. Vertiv’s GB300 reference architecture — which management claims delivers up to 50% faster on-site builds in 30% less physical space with up to 70% better annual energy efficiency — is the answer to that demand. Selling an integrated architecture rather than a box raises the average deal size, deepens the customer relationship, and makes Vertiv far harder to displace mid-project. Over the long run, this is the driver that most changes Vertiv’s competitive economics.

Driver three: the services and installed-base flywheel. Every megawatt of Vertiv equipment shipped becomes an annuity. Liquid cooling systems in particular are more complex to commission, monitor, and maintain than the air handlers they replace, and downtime in an AI training cluster is astronomically expensive, so operators pay for premium service contracts. Industry research expects the data center liquid cooling service segment to grow at a 36% CAGR, outpacing even the hardware. Because service revenue is recurring, higher-margin, and stickier than equipment sales, the mix shift toward services should structurally lift Vertiv’s margins and smooth the cyclicality that has historically made the market discount infrastructure names. This is the slowest-moving but most durable of the three drivers.

2-3. Competitive Landscape

Vertiv competes across two adjacent battlefields — power and thermal — and the competitor set differs on each. In power, the heavyweights are Schneider Electric, Eaton, and ABB, all far larger diversified electrical conglomerates. In thermal and specifically liquid cooling, the field includes Schneider, Eaton (which added liquid cooling via its Boyd Thermal acquisition in March 2026), Modine Manufacturing, and a long tail of specialists. Industry research indicates the five leading vendors — Vertiv, Schneider Electric, Eaton, Rittal, and Airedale — collectively account for roughly 52–58% of global revenue, meaning the market is consolidating but still fragmented enough to leave share up for grabs.

The comparison below is directional; figures for the diversified competitors are approximate and drawn from public knowledge of far larger, multi-industry parents, so they should be read as scale context rather than precise financials.

Company	Focus	Approx. scale (est.)	Key advantage	Key limitation
Vertiv (VRT)	Data center power + thermal pure-play	~$10.8B sales	Integrated portfolio, Nvidia co-design, top liquid-cooling share	Concentrated in one end-market
Schneider Electric	Broad electrical + automation	Multi-tens of billions (est.)	Global scale, energy management breadth	Data center is one of many segments
Eaton	Diversified power management	Large-cap industrial (est.)	Boyd Thermal deal adds liquid cooling	Later to integrated liquid cooling
Modine	Thermal specialist	Mid-cap (est.)	Focused cooling engineering	Smaller power/integration footprint

Vertiv’s differentiation is precisely its concentration. Schneider and Eaton are enormous, but data centers are one vertical among many, and their liquid cooling capabilities were assembled partly through recent acquisitions that still need integration. Vertiv wakes up every morning thinking about nothing but critical digital infrastructure, and it can offer a single integrated stack — power and thermal, designed to work together and validated against the latest Nvidia platform — that a component vendor cannot easily match. That said, the competitive intensity is real and rising: Eaton’s Boyd Thermal acquisition is an explicit move onto Vertiv’s turf, and the diversified players have the balance sheets to buy their way to parity. Vertiv’s defense is time-to-market and depth of co-engineering, not permanence.

3. Economic Moat Analysis

Moat Type 1: Switching Costs and Design-In Lock-In

Vertiv’s most powerful moat is the switching cost embedded in a validated reference architecture. When a hyperscaler or neocloud standardizes its next-generation AI hall around Vertiv’s power-and-cooling design for a GB300 NVL72 deployment, that choice propagates through the entire project: the mechanical layout, the coolant plumbing, the electrical distribution, the monitoring software, and the service contract all get specified around Vertiv equipment. Ripping that out mid-build to save a few points on a competitor’s bid is not merely expensive — it introduces schedule risk and thermal risk into a facility where a delay of even a quarter can strand hundreds of millions of dollars of idle GPUs. The cost of switching is measured not in the price of the hardware but in the opportunity cost of a delayed AI cluster, which is enormous.

The concrete evidence for this moat is the backlog itself. A combined order backlog of $15.0 billion against trailing revenue of $10.84 billion means customers have committed to more than a full year of forward purchases, and management states this provides visibility into 2027. Backlog of that scale does not accumulate for a commoditized box that buyers swap on price; it accumulates when customers have designed you into projects they cannot easily re-engineer. The 109% year-over-year backlog growth and 57% sequential jump reported for the period underscore that the design-in dynamic is strengthening, not fading.

Moat Type 2: Integrated Portfolio and Co-Engineering with the Platform Setter

The second moat is the breadth of Vertiv’s portfolio combined with its privileged position as Nvidia’s power-and-thermal co-designer. Because Nvidia effectively sets the specification for the entire AI data center through its platform road map, a supplier that co-develops the reference cooling and power architecture for each new Nvidia generation gets a structural head start. Vertiv did exactly this for the GB300 NVL72, delivering SimReady assets inside Nvidia’s Omniverse blueprint for AI factory design. That means data center architects modeling their next facility inside Nvidia’s own design tools encounter Vertiv’s equipment as the validated default. Few competitors can offer both the full power stack and the full thermal stack, and fewer still are embedded at the platform-definition layer. The integration of power and cooling into one warrantied system — rather than two separately sourced subsystems that the customer must make play nicely together — is a genuine engineering advantage at 142-kilowatt densities where the two disciplines are physically intertwined.

Moat Durability Assessment

Will this moat hold for five to ten years? The honest answer is that it is durable but not impregnable, and it must be continuously re-earned. The design-in and co-engineering advantages reset with every accelerator generation: Vertiv has to win the reference-architecture role for the platform after GB300, and the one after that. If a competitor were to leapfrog Vertiv on a future Nvidia generation, some of the lock-in would erode at the next refresh cycle. The most serious threats to durability are Eaton and Schneider deploying their vastly larger balance sheets to buy capability and undercut on price, and the small but non-trivial risk that hyperscalers vertically integrate cooling design in-house, as they have done in other parts of their stack. The counterargument is that Vertiv’s focus, installed base, service network, and accumulating co-engineering know-how compound over time, and that the sheer pace of density increases keeps raising the engineering bar in ways that favor the incumbent specialist. On balance we judge the moat to be wide today and likely to remain at least narrow through the current AI capex cycle, but investors should monitor the reference-architecture wins on each new Nvidia platform as the single most important leading indicator of moat health.

투자 분석 이미지 — Photo by Taylor Vick on Unsplash

4. Financial Analysis

Vertiv’s financial profile is that of a company whose growth is accelerating while its profitability is stepping up simultaneously — an unusual and attractive combination. Trailing twelve-month sales are $10.84 billion and net income is $1.56 billion, for a net profit margin of 14.37%. Gross margin runs at 35% and operating margin at 19.06%, both healthy for a hardware-centric industrial and both trending higher as the mix shifts toward software-enabled, service-attached, higher-value integrated systems.

The growth figures are striking. Sales grew 30.13% quarter-over-quarter, EPS grew 135.90% quarter-over-quarter, and EPS this year is up 54.53%. Management raised its full-year 2026 outlook to adjusted EPS of $6.30–$6.40 (above the roughly $6.16 consensus at the time), revenue of $13.5–$14.0 billion, and adjusted free cash flow of $2.1–$2.3 billion. That revenue guide implies well above 20% year-over-year growth off the trailing base, and the free-cash-flow guide is the number income investors should anchor on, because it demonstrates the earnings are converting to cash rather than accruing on paper.

The illustrative multi-year trajectory below frames the acceleration. The trailing-twelve-month column uses the fetched authoritative figures; the earlier-year and full-year 2026 columns are drawn from company guidance and disclosures and are marked as such.

Metric	FY2024 (est.)	FY2025 (est.)	TTM (actual)	FY2026 guide
Revenue	~$7.8B (est.)	~$9.0B (est.)	$10.84B	$13.5–14.0B
Operating margin	mid-teens (est.)	~18% (est.)	19.06%	expanding
Net income	~$0.5B (est.)	~$1.0B (est.)	$1.56B	—
EPS	—	—	EPS ttm $3.98	adj. $6.30–6.40
Order backlog	—	—	$15.0B	into 2027

The story behind the numbers is the AI capex wave translating directly into orders, orders converting to shipped revenue, and operating leverage on a largely fixed manufacturing base driving margins and EPS up faster than sales. The 135.90% quarter-over-quarter EPS jump is the operating-leverage signature: when revenue accelerates 30% against a cost base that does not scale one-for-one, profit growth outruns sales growth dramatically.

On the balance sheet, the key operating metric to watch is that $15.0 billion backlog, which functions as a de-risking mechanism for the next two years of revenue. Leverage is moderate: debt-to-equity is 0.77, comfortable for a company generating $2 billion-plus of annual free cash flow, and the balance sheet has been deleveraging as cash generation has scaled. One caveat worth naming: the 45.1% return on equity and 27.18 price-to-book partly reflect a relatively thin book-equity base — a byproduct of the company’s capital structure and buyback history — so ROE overstates the underlying returns on tangible invested capital. Return on assets of 13.64% is a cleaner read on capital efficiency and is genuinely strong for an industrial. The path from here is continued margin expansion driven by the services mix shift and operating leverage, rather than a turnaround story; Vertiv is already solidly profitable and the debate is about the slope of the improvement, not its existence.

5. Valuation

Vertiv is not cheap on any conventional trailing metric, and any honest valuation has to start there. At $300.53, the stock trades at 75.4 times trailing EPS of $3.98, 10.65 times sales, and 27.18 times book value. Those are growth-stock multiples, and they price in years of continued execution. The relevant question is not whether Vertiv is expensive on the past — it plainly is — but whether the forward earnings power justifies the price.

Because the trailing P/E is distorted by how rapidly earnings are ramping, the appropriate anchor is forward earnings. Consensus next-year EPS is $8.80, which puts the stock at a forward P/E of 34.2 (that is, $300.53 ÷ $8.80). We build our fair-value scenarios off that $8.80 forward figure, applying a range of exit multiples that reflect Vertiv’s growth rate, moat, and the risk that AI capex normalizes.

Base case — $355 (approx. 40x forward EPS of $8.80): This assumes Vertiv sustains 20%-plus revenue growth through 2027 on the strength of its backlog, holds or modestly expands margins as services mix up, and the market continues to award a premium multiple appropriate for a category leader with visible growth. Forty times forward earnings is demanding but defensible for a business compounding earnings north of 30% with a wide moat. The base case implies roughly 18% upside from the current price.

Bull case — $425 (approx. 48x forward EPS of $8.80): Here liquid cooling adoption runs ahead of even optimistic forecasts, Vertiv extends its reference-architecture wins across the next Nvidia generation, services grow at the 36% pace the industry expects, and the multiple re-rates higher as investors treat Vertiv as an indispensable AI-infrastructure toll booth. This is roughly 41% upside and is broadly consistent with the most bullish sell-side targets, such as Bernstein’s $416 initiation.

Bear case — $250 (approx. 28x forward EPS of $8.80): In this scenario, hyperscaler capex growth decelerates in 2027, competition from Eaton and Schneider compresses pricing, and the market de-rates the entire AI-infrastructure complex toward more normal industrial multiples. Note that even in the bear case the absolute EPS may not fall much — Vertiv’s backlog cushions near-term revenue — so the pain comes primarily through multiple compression rather than an earnings collapse. This represents roughly 17% downside.

Weighing these, our base case of $355 sits modestly below the sell-side consensus target of $378.23, reflecting a deliberately conservative stance on the exit multiple. We agree with the consensus direction — the stock has upside — but we are slightly more cautious than the average analyst on how much multiple the market will pay two years out. The consensus $378 implies roughly 43 times forward EPS; we prefer to underwrite at 40 times and treat anything above that as a bonus delivered by the bull case rather than a requirement of the base case. Against the current $300.53, the blended risk/reward is favorable: a base case of +18%, a bull case of +41%, and a bear case of -17% is an attractive asymmetry for a category leader with a $15 billion backlog.

6. Risk Factors

Risk 1: Valuation and the fragility of high-multiple stocks. At 75 times trailing and 34 times forward earnings, Vertiv is priced for sustained excellence, and high-multiple stocks are unforgiving of disappointment. A single quarter of orders that merely meets rather than crushes expectations, or a guidance raise that is smaller than the market has been trained to expect, could trigger a sharp de-rating even if the long-term thesis is intact. The stock has already traveled from $110 to nearly $380 within a year, so a meaningful portion of the future is in the price, and the margin of safety on valuation is thin. Investors sizing a position must accept that drawdowns of 20–30% are entirely possible on sentiment shifts alone, independent of business fundamentals. This is the single most important risk for anyone buying at the current level.

Risk 2: Concentration in a single, capex-cyclical end-market. Vertiv’s fortunes are tightly bound to data center capital spending, and that spending is concentrated among a handful of hyperscalers and a growing set of AI-focused neoclouds whose own funding depends on continued enthusiasm for AI. If the return on AI investment disappoints and the hyperscalers throttle capex — as capital-intensive technology cycles have done repeatedly in history — Vertiv’s order growth would decelerate quickly. Unlike a diversified industrial such as Eaton or Schneider that can lean on utility, building, or automation demand when data centers slow, Vertiv has far less to fall back on. The very focus that gives Vertiv its moat also gives it its cyclicality, and the backlog, while reassuring for the next two years, does not protect against a multi-year capex downturn beyond that horizon.

Risk 3: Intensifying competition and potential customer in-sourcing. The competitive moat is real but must be continuously defended. Eaton’s March 2026 acquisition of Boyd Thermal is an explicit assault on Vertiv’s liquid cooling position, and Schneider Electric brings comparable scale and its own thermal capabilities. Both rivals are far larger and can subsidize aggressive pricing to win reference designs. Separately, the hyperscalers have a demonstrated history of vertically integrating parts of their infrastructure once volumes justify it, and if the largest customers decide to design cooling in-house or dual-source aggressively, Vertiv’s pricing power and share could erode. The company must win the power-and-thermal reference-architecture role on each successive Nvidia platform; losing even one generation to a competitor would meaningfully weaken the design-in lock-in that underpins the entire investment case.

7. Conclusion & Exit Plan

Vertiv is one of the highest-quality ways to own the physical layer of the AI build-out. It combines a genuine economic moat rooted in switching costs and platform co-engineering, a $15 billion backlog that de-risks the next two years of revenue, a 45% return on equity and 19% operating margin that prove the model prints cash, and leadership share in liquid cooling — the fastest-growing, most supply-constrained corner of data center infrastructure. The counterweight is valuation: at 34 times forward earnings the stock already reflects a great deal of good news, and its concentration in hyperscaler capex makes it more cyclical than its diversified peers.

Weighing quality against price, we rate Vertiv a Buy — not a Strong Buy, because the valuation leaves limited margin of safety, but a Buy because the growth, moat, and backlog visibility justify a premium multiple and the risk/reward remains skewed to the upside.

Entry price range: We would accumulate between $260 and $300, with the more attractive risk/reward below $285. Given the stock’s volatility, scaling in rather than buying a full position at once is prudent; a pullback toward the low-$260s on a broad AI-sentiment wobble would be an opportunity to add rather than a reason to sell.

Exit conditions:
– Target achieved: Trim into strength as the stock approaches the base-case target of $355, and take further profits toward the bull-case $425.
– Fundamental break: Sell if the order backlog contracts for two consecutive quarters, if Vertiv loses the power-and-thermal reference-architecture role on the next major Nvidia platform, or if operating margin rolls back below the mid-teens, any of which would signal the moat or the capex cycle is turning.
– Time-based: Reassess the full thesis in six months or immediately after the next two quarterly reports, with particular attention to backlog trajectory and services growth.

Item	Detail
Company	Vertiv Holdings (VRT)
Current Price	$300.53
Target Price	$355 base / $425 bull / $250 bear
Upside	+18% base / +41% bull
Rating	Buy
Key Thesis	Pure-play leader in AI data center power + liquid cooling with a $15B backlog and Nvidia co-design lock-in
Main Risk	High valuation (34x forward) concentrated in cyclical hyperscaler capex

—

This article is for informational purposes only and does not constitute investment advice. All data sourced from public filings, analyst reports, and news as of the publication date. Invest at your own discretion.

This content is general investment information provided to an indefinite/unspecified audience by a quasi-investment advisory business registered under Korea’s Financial Investment Services and Capital Markets Act, and is not personalized 1:1 investment advice tailored to any individual investor. This analysis is for informational purposes only and is not a solicitation to invest. All investment decisions and their consequences rest solely with the investor. The estimates and assumptions in this report are as of the writing date (2026-07-03) and may not materialize depending on market conditions and geopolitical variables. Financial data used reflects sources such as company filings and analyst consensus, and the scenarios and price targets represent the author’s conservative assessment. All investments carry the risk of principal loss, and past performance or analytical track record does not guarantee future results. As of the writing date, the author does not hold a position in this stock. The author’s holdings and positions may change without prior notice depending on market conditions.