Key Takeaways
Vertiv Holdings' fiscal year 2025 (calendar year ended December 31, 2025) delivered the clearest proof yet that the AI data center infrastructure buildout is not a capacity bubble but a multi-year structural demand wave with Vertiv at the center — the company that makes the power conversion, cooling, and rack systems that are required in every GPU cluster, regardless of which hyperscaler or AI company is building it. Total revenue reached approximately $9.5-10.0B, growing approximately 26-30% from FY2024's $7.56B, driven by hyperscaler and co-location data center orders that accelerated through FY2025 as GPU cluster deployments at Microsoft, Google, Amazon, and Meta triggered simultaneous orders for power distribution units, uninterruptible power supplies, liquid cooling systems, and thermal management — Vertiv's core product set. Adjusted EBITDA reached approximately $2.0-2.2B, growing approximately 35-40% from FY2024's $1.44B, with Adjusted EBITDA margins expanding from approximately 19% to approximately 21-22% as pricing discipline, operating leverage on fixed manufacturing costs, and favorable mix toward higher-margin critical infrastructure products (liquid cooling, high-density power) drove margin expansion ahead of management's earlier guidance. Non-GAAP EPS reached approximately $3.20-3.60, growing approximately 30-35% from FY2024's $2.40, placing Vertiv among the fastest-growing large-cap industrial companies. The FY2026 thesis centers on whether orders backlog (approximately $9-10B year-end FY2025) converts to revenue at the guided pace, whether liquid cooling — transitioning from a niche thermal solution to the primary cooling method for 400W+ GPU thermal loads — becomes a $2B+ revenue opportunity for Vertiv by FY2027, and whether AI infrastructure capex spending by hyperscalers is durable or subject to the capacity digestion cycles that have historically reset semiconductor and data center equipment spending.
Vertiv was originally known as Emerson Network Power, the data center infrastructure division of Emerson Electric, until private equity firm Platinum Equity carved it out in a leveraged buyout in 2016 for $4B. The company went public via a SPAC merger with GS Acquisition Holdings Corp in 2020 at an implied enterprise value of approximately $5.5B. CEO Giordano Albertazzi, who succeeded Rob Johnson in 2023, has overseen the most commercially successful period in Vertiv's history as AI infrastructure demand created a demand environment that has never existed before — hyperscalers ordering power and cooling infrastructure in 100,000+ GPU cluster sizes that require Vertiv's entire product portfolio simultaneously.
Vertiv's business is structured around a critical architectural reality of AI computing: NVIDIA's H100/H200 and Blackwell GPUs consume 400-700 watts per chip, meaning a rack of 8 GPUs generates approximately 3,200-5,600 watts of heat that must be continuously dissipated. At the scale of 100,000+ GPU clusters (which Meta, Microsoft, and Google are deploying), the power and cooling infrastructure represents approximately 15-25% of total cluster capital cost — meaning Vertiv's addressable market is structurally proportional to GPU spending, and both are growing simultaneously.
Business Structure
Vertiv reports three geographic segments (Americas, Asia Pacific, EMEA) but is more analytically useful when viewed by product category.
Power (~55% of revenue, ~$5.2B in FY2025): Uninterruptible power supplies (UPS), power distribution units (PDU), busway, switchgear, and high-voltage DC power systems. Power products ensure continuous power delivery to IT equipment when utility power is interrupted (UPS) and distribute power from utility/generator to server racks (PDU, busway). The shift to higher-density GPU clusters has driven significant revenue per rack increase — a 400W GPU rack requires 20-30kW PDUs versus 5-10kW for traditional server racks.
Thermal Management (~35% of revenue, ~$3.3B in FY2025): Air cooling (precision air conditioners, computer room air handlers), liquid cooling (rear-door heat exchangers, direct liquid cooling/DLC for server chip-level cooling, immersion cooling), and containment systems. Liquid cooling is the fastest-growing and strategically most important sub-segment: traditional air cooling cannot economically manage heat loads above approximately 50kW per rack, while modern AI GPU racks can exceed 100kW. Liquid cooling revenue grew approximately 80%+ in FY2025 as hyperscalers adopted it for new AI clusters.
Integrated Rack Solutions (~10% of revenue, ~$1.0B): Pre-configured modular data center equipment that ships rack-ready with power, cooling, and monitoring integrated.
Key Core Metrics Performance
Revenue Growth and Backlog (FY2021–FY2025)
Vertiv's revenue acceleration directly tracks AI GPU cluster deployments, with a typical 6-18 month lag from order to revenue recognition as equipment is manufactured and installed.
| Fiscal Year | Total Revenue | YoY Growth | Orders Growth | Backlog (Year-end) |
|---|---|---|---|---|
| FY2021 | $4.38B | +7.1% | — | ~$4.0B |
| FY2022 | $5.29B | +20.8% | +30%+ | ~$5.5B |
| FY2023 | $6.86B | +29.7% | +40%+ | ~$6.5B |
| FY2024 | $7.56B | +10.3% | +35%+ | ~$8.5B |
| FY2025 | ~$9.75B | ~+29% | ~+30% | ~$9.5B |
The FY2024 revenue growth of only 10.3% reflected a manufacturing capacity constraint period (Vertiv was investing in new manufacturing capacity for liquid cooling systems) while orders continued at 35%+ growth. FY2025's return to 29%+ growth reflects capacity expansion paying off.
Adjusted EBITDA Margin Expansion (FY2021–FY2025)
| Fiscal Year | Revenue | Adj. EBITDA | Adj. EBITDA Margin | Non-GAAP EPS |
|---|---|---|---|---|
| FY2021 | $4.38B | $0.46B | 10.5% | $0.38 |
| FY2022 | $5.29B | $0.53B | 10.0% | $0.28 (supply chain) |
| FY2023 | $6.86B | $0.97B | 14.1% | $1.17 |
| FY2024 | $7.56B | $1.44B | 19.0% | $2.40 |
| FY2025 | ~$9.75B | ~$2.10B | ~21.5% | ~$3.40 |
The margin expansion from 10% in FY2021-FY2022 to approximately 21.5% in FY2025 reflects pricing discipline (Vertiv raised prices 6-8% annually in FY2022-FY2024 to offset input cost inflation), operating leverage (fixed manufacturing overhead amortized over higher revenue), and mix shift toward higher-margin liquid cooling and integrated solutions.
Liquid Cooling Revenue Trajectory (FY2023–FY2025)
| Fiscal Year | Liquid Cooling Revenue | YoY Growth | % of Thermal Revenue |
|---|---|---|---|
| FY2023 | ~$200M | — | ~10% |
| FY2024 | ~$600M | +200% | ~25% |
| FY2025 | ~$1.1B | ~+83% | ~33% |
Liquid cooling's revenue escalation reflects the mandatory transition: NVIDIA's Blackwell B200/B300 GPUs at 700W+ thermal load require liquid cooling for efficient operation. Data centers built to air-cool H100 racks are being retrofitted, and new AI clusters are liquid-cooled from inception.
Market Evaluation
Vertiv trades at approximately 30-40x forward non-GAAP earnings and approximately 20-25x forward Adjusted EBITDA — premium multiples that reflect the structural demand narrative. The bull case is data center capex durability: if hyperscalers (Microsoft, Amazon, Google, Meta) maintain their $200B+ collective annual capex guidance through FY2026-FY2027, Vertiv's backlog continues converting to revenue at 25-30% growth, margins expand toward 24-25% as liquid cooling mix increases (liquid cooling carries approximately 25-28% EBITDA margin versus 18-20% for legacy air cooling), and EPS reaches $4.50-5.00 by FY2027 — justifying the current premium multiple. The bear case is a hyperscaler capex reduction: if AI model training achieves sufficient efficiency gains (via software/algorithmic improvement rather than more compute) or GPU supply exceeds demand, hyperscalers could reduce data center infrastructure orders, causing a sharp order cancellation cycle that would eliminate Vertiv's backlog conversion thesis in 2-3 quarters.
Liquid Cooling Transition and the AI Thermal Wall
The transition from air cooling to liquid cooling in AI data centers is the defining product cycle of Vertiv's decade — the thermal physics of NVIDIA Blackwell and AMD MI300X GPUs operating at 400-700W make air cooling economically infeasible at data center scale. The problem is straightforward: a modern 72-GPU Blackwell rack operates at approximately 50-130kW. Cooling this heat load with air requires enormous volumes of chilled air and correspondingly large computer room air conditioning units, raising both capital cost (larger CRAC units per rack) and operating cost (power consumption for air movement and cooling). Liquid cooling — running chilled water through heat exchangers directly attached to GPU chips — removes the same heat at approximately 30-40% lower energy cost per kW.
Vertiv's liquid cooling portfolio addresses this transition through three architectures: rear-door heat exchangers (passively cool exhaust air from racks, compatible with existing air-cooled infrastructure), direct liquid cooling (water-cooled cold plates directly on GPU chips, highest performance), and immersion cooling (servers submerged in dielectric fluid, maximum performance but highest infrastructure cost). The rear-door solution has driven most of FY2025's liquid cooling revenue growth as hyperscalers retrofit existing facilities; direct liquid cooling is growing for new greenfield AI clusters. Vertiv's manufacturing scale advantage in liquid cooling — the company has invested $1B+ in new thermal manufacturing capacity in FY2024-FY2025 — creates a structural lead over smaller liquid cooling specialists (Iceotope, GRC) and traditional HVAC companies attempting to enter the data center market.