Nvidia’s Power Surge, Powers Vertiv

Cooling is already a core determinant of data center economics. The next generation of Nvidia (NASDAQ: NVDA) AI systems will make cooling even more critical, making Vertiv’s (NYSE: VRT) partnership with NVDA a huge opportunity for investors.

With Blackwell-class racks already pushing toward 120–140 kW and forthcoming Rubin Vera platforms expected to reach hundreds of kilowatts per rack, air cooling cannot keep up and is everywhere yielding to more expensive liquid cooling systems. The liquid-cooling bill for a single GB300 NVL72 rack (integrating 72 Nvidia Blackwell Ultra GPUs and 36 Nvidia Grace CPUs into a single platform) approaches $50,000. That rises to more than $55,000 for Nvidia’s comparable forthcoming Rubin-generation systems as GPU power levels climb toward 1.8 kW and beyond per device. Every incremental watt of compute brings a disproportionate burden on power delivery, heat rejection, and total cost of ownership.

Vertiv (VRT) has moved aggressively to sit at the center of this transformation. Over the past two years, the company has repositioned itself from a generic critical-infrastructure vendor to a co-architect of Nvidia-class “AI factories.” Vertiv is co-developing power and liquid-cooling reference designs for Nvidia’s GB200 NVL72 Blackwell systems. But the big target is 800-volt direct current (800 VDC) platforms designed for future Rubin Ultra racks.

Traditional 54-V in-rack power distribution, designed for kilowatt (KW)-scale racks, isn’t designed to support the megawatt (MW)-scale racks required by the Rubin generation coming soon to modern AI factories. 800 volts DC will support 1 MW racks and beyond, and save space, economize on copper, improve energy efficiency, and could reduce the total cost of ownership by up to 30% (compared to current technology for the same workload.)

This is not a loose marketing tie-up; it is a blueprint-level partnership that defines how megawatt-scale AI clusters will be powered and cooled.

As AI infrastructure capex accelerates into the hundreds of billions of dollars per year, a rising share of that spend will flow into power and cooling rather than chips. Vertiv’s numbers already reflect that. Its competitive position versus Schneider Electric, Eaton, and smaller liquid-cooling specialists is being reshaped by its Nvidia collaboration.

Vertiv Rises on the AI Cooling Cycle

As illustrated by Table 1, Vertiv’s financial trajectory tracks the AI datacenter build-out. Revenue has grown from $5.7 billion in 2022 to over $9 billion on a trailing 12-month basis. Margins have expanded with the shift toward high-value AI infrastructure. This pattern confirms that Vertiv is capturing incremental value per rack rather than relying on volume alone for growth. Adoption of liquid-cooling is translating into tangible financial leverage and operational profitability. Vertiv is now one of the few hardware infrastructure players growing at a rate comparable to the AI chip suppliers themselves.

* TTM = trailing 12 months based on latest reported figures.

Vertiv and Nvidia design the Rubin-class AI factory

As Table 2 shows, Vertiv’s technical integration with Nvidia distinguishes it from competitors, including Schneider Electric (SPGSY) and Eaton + Boyd Thermal. Eaton (ETN) acquired Boyd Thermal in late 2025 for $9.5 billion to expand its data-center liquid-cooling capabilities and integrate thermal management with its power-infrastructure business.

Vertiv is directly embedded in Nvidia’s GB200 NVL72 and Rubin-class reference architectures, providing end-to-end power and cooling systems that support rack densities approaching 130 kW and eventually over 500 kW on the way to megawatt scale. Vertiv’s 360AI blueprint and its participation in the Department of Energy’s COOLERCHIPS program further reinforce this leadership by combining direct-to-chip, immersion, and hybrid cooling approaches within standardized, scalable systems. Nvidia’s Omniverse DSX now includes Vertiv’s designs as digital-twin models, enabling customers to simulate Rubin-era deployments before construction. These co-developed systems place Vertiv at the heart of how hyperscalers and sovereign AI projects plan to build their infrastructure.

Extraordinary demand

As Table 3 shows, the potential demand for the product of the Vertiv/Nvidia partnership is extraordinary. Global datacenter power demand is forecast to rise by roughly 50% through 2027, with AI-related spending projected to reach $371 billion in 2025. Within that, the cooling segment is expanding at nearly 17% annually, led by the rapid adoption of liquid systems to handle racks exceeding 30–50 kW. Cooling is now a foundational enabler of AI computing. For Vertiv, that addressable market is compounding as both capacity and per-rack cooling intensity rise.

Investor takeaway

Vertiv’s transformation from a legacy data-center supplier to a critical enabler of Nvidia’s AI infrastructure is visible in the stock’s 2025 market performance. The chart below compares year-to-date share price gains for Vertiv Holdings (VRT) and Nvidia (NVDA).

Vertiv’s stock has surged approximately 49.3% year to date, decisively outperforming Nvidia’s 29.3% gain over the same period. The relative strength of Vertiv’s performance is notable given that Nvidia’s own rally has driven the broader AI sector. Clearly investors recognize Vertiv as a second-order AI winner.

Chart 1

Each new Nvidia platform—consuming greater wattage and demanding higher rack density—translates directly into incremental revenue opportunities for Vertiv, not just in equipment sales but in long-term service contracts and retrofits. The company’s ability to convert thermal complexity into consistent profit expansion has made it a top performer among AI infrastructure stocks in 2025.

Looking ahead, Vertiv’s outperformance may persist as hyperscaler and sovereign AI projects enter large-scale build-out phases. The company’s 49% stock appreciation in 2025 reflects what my Gilder colleague John Schroeter has called “the great AI Inversion,” with revenues and profits migrating to the physical layer of AI. Power, cooling, and energy efficiency are the next frontier for sustained investment returns. As GPU wattage and rack densities continue to climb toward the Rubin Ultra era, Vertiv stands to be one of the most direct beneficiaries of this thermally constrained AI supercycle.