Vertiv Holdings announced a $1 billion-plus acquisition of a Raleigh, North Carolina data center infrastructure firm, expected to close in the second half of fiscal 2025. The target company specializes in liquid cooling systems for high-density AI compute racks, a category that tripled in commercial deployment since Q1 2024 as nvidia H100 and H200 clusters pushed thermal loads past 40 kilowatts per rack.
The deal represents Vertiv's largest acquisition since its 2020 SPAC merger and marks a deliberate shift from power distribution equipment into thermal management. Vertiv CEO Giordano Albertazzi told analysts the acquired firm holds patents on direct-to-chip liquid cooling manifolds that reduce GPU throttling by 18-22 percent compared to air-cooled systems. The Raleigh operation ships 1,200 cooling units per quarter and holds service contracts with three of the five largest hyperscalers. Vertiv projects the acquisition will add $400-450 million in annualized revenue once integration completes in Q1 2026.
This move matters because cooling bottlenecks now delay AI infrastructure deployments longer than power availability or real estate permitting. Hyperscalers building 500-megawatt AI campuses need liquid cooling systems ordered 18-24 months before servers arrive, creating a land grab for suppliers who can deliver at scale. Vertiv's existing UPS and power distribution business gave it facility access but no thermal technology. The Raleigh acquisition solves that gap while hyperscalers still build their own cooling R&D teams. CoreWeave's latest Tennessee buildout postponed rack deployment by 90 days waiting for cooling infrastructure, a delay that costs roughly $2.8 million per day in lost compute revenue.
The acquisition also signals Vertiv expects AI training workloads to stay concentrated in owned data centers rather than distribute to edge locations. Liquid cooling systems require on-site glycol handling and trained technicians, making them uneconomical for facilities under 10 megawatts. Vertiv's capital allocation suggests the next 36 months belong to hyperscale campuses, not distributed inference nodes. That view aligns with Microsoft's recent $80 billion data center capex guidance but contradicts meta's push toward smaller regional hubs.
Allocators should watch Vertiv's Q2 2025 earnings call in late July for integration cost guidance and whether the Raleigh unit's backlog extends past Q4 2026. Hyperscaler capex plans for 2026 will be finalized by September, giving visibility into whether this premium multiple pays off. Competitive responses from Schneider Electric and Eaton will likely appear before year-end if Vertiv's thesis proves correct.
The Raleigh firm's patent portfolio includes 47 active filings in thermal interface materials. Vertiv just bought the next 18 months of AI cooling capacity before the hyperscalers could.