CoolIT Systems announced the development of a 15kW single-phase direct liquid cooling (DLC) coldplate, according to a PR Newswire press release published June 1, 2026. The company says the design delivers nearly 4x the capacity of its prior 4kW coldplate and more than 10x the cooling required for current-generation AI GPUs, with validation using a water-glycol coolant at 1.2 L/min/kW and system-level suitability for 45°C warm-water environments, per PR Newswire and coverage in StorageReview and DataCenterDynamics. The announcement includes direct quotes from CoolIT CTO Kamal Mostafavi and an industry comment from Dylan Patel of SemiAnalysis. Public reporting frames the result as evidence that single-phase DLC retains scaling headroom for future high-density AI accelerators.
What happened
CoolIT Systems announced the development of a 15kW direct liquid cooling (DLC) coldplate design, according to a PR Newswire press release published June 1, 2026. The company reported that the new design delivers nearly 4x the capacity of its previously demonstrated 4kW coldplate and more than 10x the cooling required for current-generation AI GPUs, per the PR Newswire release and coverage by StorageReview and DataCenterDynamics. The PR release includes a direct quote from Kamal Mostafavi, CTO of CoolIT Systems: "Single-phase DLC is already cooling millions of AI accelerators today. This achievement shows it is also the architecture to cool AI infrastructure well into the future," as printed in PR Newswire.
Technical details
Per CoolIT's public materials and reporting in DataCenterDynamics and StorageReview, the 15kW coldplate is based on the company's Split-Flow microchannel architecture and was validated on a thermal test vehicle using a standard water-glycol coolant at 1.2 L/min/kW. Coverage states the design produces system-level thermal performance suitable for 45°C warm-water cooling environments, aligning with industry movement toward higher supply temperatures to improve data center efficiency (PR Newswire; DataCenterDynamics).
Editorial analysis - technical context
Single-phase DLC relies on turbulent flow through engineered microchannels to move heat from silicon into a liquid loop without phase change. Observed patterns in similar transitions: increasing coolant supply temperatures to 45°C and above is a common approach to raise facility-level PUE efficiency and enable heat reuse. Companies rolling multi-kilowatt coldplates typically pursue higher microchannel hydraulic performance and conservative flow-per-kW targets; the reported 1.2 L/min/kW validation number fits within that engineering tradeoff space rather than representing an operational specification for deployed racks.
Context and significance
Editorial analysis: Public coverage frames the 15kW figure as extending the thermal headroom of single-phase DLC, which matters because several next-generation accelerators operate at much higher thermal design points. Reporting in DataCenterDynamics notes that Nvidia Blackwell GPUs can run up to 1,200W and Rubin-class parts have TDPs up to 2,300W, creating pressure on per-device and per-slot cooling capacity. For infrastructure teams, a validated 15kW coldplate suggests that single-phase DLC architectures may remain viable for a broader set of high-power devices, delaying or reducing immediate need for alternative approaches such as two-phase systems or full-immersion cooling in certain deployment scenarios.
What to watch
For practitioners: Watch for independent third-party thermal validation data and system-level demonstrations (rack, CDU, and server integrations) beyond vendor test vehicles, as those are the indicators that a coldplate design moves from lab validation to field-ready deployment. Also monitor coolant loop designs and connector standards in customer trials, and follow server OEM and hyperscaler integration announcements that reference compatibility with 45°C supply-temperature single-phase DLC.
Scoring Rationale #
The announcement is a notable infrastructure development that materially affects cooling design choices for high-density AI deployments. It provides meaningful technical headroom but still requires system-level validation and OEM integrations to shift industry deployments.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.