TL;DR: best liquid cooling solutions and companies in 2026
The best liquid cooling solution is the one the facility, server design, operations team, and warranty boundary can support. Start with the cooling architecture, then compare companies by integration responsibility and proof.
| Company or solution | Best for | Quick list hint |
|---|---|---|
| Vertiv | Best for facility-scale thermal management and infrastructure integration | Verify CDU, heat rejection, service scope, and controls integration. |
| CoolIT Systems | Best for direct liquid cooling components and cold-plate deployments | Match server platform, manifold, coolant, and warranty boundaries. |
| Schneider Electric | Best for integrated power, cooling, and data center infrastructure planning | Confirm end-to-end scope, controls responsibility, and facility retrofit limits. |
| LiquidStack | Best for immersion and high-density liquid cooling evaluation | Verify tank or system fit, operations model, maintenance, and fluid handling. |
| Submer | Best for immersion and modular liquid-cooled deployment paths | Check facility interface, service plan, and hardware compatibility. |
| Accelsius | Best for two-phase direct-to-chip diligence | Confirm refrigerant path, service model, and server support. |
| GRC | Best for immersion cooling retrofit and density comparisons | Verify deployment envelope, operating procedures, and facility heat rejection. |
| ZutaCore | Best for waterless two-phase direct-to-chip evaluation | Require server, warranty, service, and operational evidence. |
Buyer action
Pressure-test the cooling shortlist before vendor outreach
Use these tools after the provider shortlist to compare direct-to-chip and immersion fit, size rack-level cooling load, and prepare quote questions for GPU colocation providers.
Fit depends on rack density, chip type, coolant boundary, facility water strategy, heat rejection, hardware warranty, operations staffing, and whether the buyer is retrofitting an existing hall or building a new one.
How should buyers choose between direct-to-chip and immersion?
Direct-to-chip can fit buyers that need to keep more conventional rack service models while removing heat at the chip. Immersion can fit very high-density or specialized environments, but it changes hardware handling, service workflow, fluid management, and procurement assumptions. The decision should be made with the server vendor, facility operator, cooling vendor, and operations owner in the same diligence loop.
What evidence proves a liquid cooling company can support AI data centers?
Ask for compatible server platforms, CDU or tank specifications, coolant requirements, manifold design, leak detection, heat-rejection path, maintenance procedure, warranty boundary, spares plan, installer scope, commissioning process, service-level assumptions, and references with comparable density. For colocation, also confirm what the landlord allows.
What facility constraints can break a liquid cooling plan?
The common blockers are floor loading, water availability, heat-rejection capacity, power and cooling redundancy, pipe routing, containment, service access, insurance requirements, and tenant-landlord responsibility. A cooling vendor can have a strong product and still be a poor fit if the data hall cannot support the integration.
How should buyer RFPs frame liquid cooling?
The RFP should separate the workload requirement from the preferred solution. Require each respondent to state the supported density, architecture, fluids, facility tie-ins, warranty assumptions, operations tasks, installation timeline, failure modes, monitoring data, and exclusions. Avoid accepting a generic liquid-cooling label without a named architecture and responsibility matrix.