Summary
- Salute says customer engagements now cover more than 20GW of AI infrastructure globally.
- Its direct-to-chip operations model covers coolant chemistry, leak detection, safety, monitoring, maintenance, procedures, and commissioning.
- The scale illustrates how liquid-cooling expertise is becoming an operational requirement, not only a construction discipline.
Salute says customers have engaged it to support more than 20GW of AI infrastructure globally, as the operational requirements around high-density and direct-to-chip liquid-cooled facilities expand alongside construction.
The figure describes customer capacity supported by Salute rather than data centres owned by the company. Salute provides facility operations, commissioning, staffing, and lifecycle services across hyperscale, cloud, colocation, edge, and AI infrastructure.
Its direct-to-chip liquid-cooling operations service covers coolant chemistry, leak detection, safety procedures, systems monitoring, maintenance, commissioning, and operating procedures. Those activities sit behind the more visible mechanical equipment being installed in new AI data halls.
Liquid cooling changes the daily operating model because the technology cooling system reaches much closer to the IT equipment. Coolant condition, pressure, flow, connectors, manifolds, pumps, sensors, and leak-management procedures all become part of routine data centre operations.
The boundary between facilities teams and IT teams can also become less clear. A conventional air-cooling problem may be addressed largely within the mechanical plant, but a direct-to-chip fault can involve equipment inside the rack as well as pumps, heat exchangers, valves, and facility-side systems.
Salute says its service uses emergency operating procedures, methods of procedure, and standard operating procedures tailored to the design and demarcation points of individual sites. It also provides specialist training and commissioning support.
Those procedures become increasingly important as single cooling systems support larger blocks of compute. A poorly controlled intervention can affect multiple racks, while leaks or coolant-quality problems can create risks very close to high-value accelerator hardware.
The 20GW claim follows Salute’s agreement earlier this month to acquire T5 Operations, a transaction that would increase its wider facility-management footprint. Salute said that deal, once completed, would take the combined operational estate to more than 15GW across more than 150 markets. The two measures are not directly interchangeable: one concerns broader capacity under management, while the latest announcement refers to customers engaging Salute around AI infrastructure support.
The distinction is important in a sector full of large gigawatt figures. Operational-service capacity does not represent new supply, but it does indicate the scale of infrastructure requiring trained people and repeatable maintenance processes after construction.
AI data centres are creating a labour and procedures challenge alongside the power and cooling challenge. The hardware can be deployed quickly only if operators can also recruit, train, and retain staff capable of maintaining unfamiliar systems safely. As liquid cooling becomes more common, the availability of that operational expertise is likely to become another constraint on how quickly new capacity can move into stable production.
