CoolIT's 15kW single-phase cold plate could reshape the data center liquid cooling technology roadmap
AI summary card
CoolIT's 15kW single-phase cold plate could reshape the data center liquid cooling technology roadmap
Bernstein believes that if CoolIT's 15kW single-phase DTC cold plate can be commercialized and mass-produced, data center operators may continue using single-phase liquid cooling for a longer period, thereby delaying the large-scale adoption of two-phase DTC and direct-to-die.
- The market had previously believed single-phase DTC was nearing its limit around GPU TDP of approximately 2-2.5kW, and that Rubin Ultra might require more aggressive solutions such as two-phase DTC or direct-to-die.
- CoolIT's announced 15kW single-phase cold plate was validated under relatively standard conditions; if it can be scaled, it would push out the need for two-phase DTC to beyond 2030.
- Two-phase DTC offers higher heat flux advantages, but still faces challenges including gas-liquid flow management, control complexity, PFAS regulation, and early-stage commercialization.
- Direct-to-die offers lower thermal resistance and stronger hotspot management capability, but comes with high manufacturing, cleanliness, packaging reliability, and cost complexity; the report expects it to be difficult to scale before 2030.
- The impact on the CDU and cold plate ecosystem is two-sided: near-term demand remains strong, while medium- to long-term innovation premiums may decline and create commoditization pressure, but the risk of cold plates being eliminated also decreases.
Report interpretation
Overview
This report focuses on the data center liquid cooling technology roadmap and discusses two possible successor paths after single-phase direct-to-chip (DTC): two-phase DTC and direct-to-die. The key variable in the report is CoolIT's 15kW single-phase cold plate announced in June 2026. If this product is not only technically feasible but also capable of commercial-scale production, single-phase DTC may continue to cover most of the cooling needs of next-generation AI infrastructure, thereby reducing the need for the industry to migrate quickly toward two-phase DTC or direct-to-die.
Core views
The report believes that the current mainstream liquid cooling approach is still single-phase DTC, and the market had previously worried that it would approach its thermal limit once GPU TDP reached around 2-2.5kW. Two-phase DTC can remove more heat through refrigerant evaporation and may approach commercial viability over the next 12-18 months, but it still faces engineering issues such as reverse flow, system control, refrigerant selection, and PFAS regulation. Direct-to-die theoretically offers stronger heat flux performance, but its R&D and manufacturing remain very early stage, and the report does not expect it to scale before 2030. CoolIT's 15kW single-phase cold plate is the key uncertainty: if it can be mass-produced, it will support data center operators in continuing to use the lowest-resistance single-phase cooling architecture and may compress the commercial window for two-phase DTC and direct-to-die.
Analysis framework
The report uses a technology pathway comparison framework, first explaining the cooling mechanisms of single-phase DTC, two-phase DTC, and direct-to-die, and then comparing them across heat flux, system complexity, manufacturing feasibility, refrigerant/material risks, commercialization timeline, and implications for the equipment ecosystem. In the investment implications section, the report maps technology pathway changes to CDUs, cold plates, and related listed equipment suppliers, emphasizing that tight near-term supply-demand dynamics still support revenue and profit, but medium- to long-term technological differentiation and pricing power may be affected.
Methodology notes
Assess the scalability of data center liquid cooling pathways through the state of the cooling medium, heat transfer path, and system integration method.
Single-phase DTC keeps the coolant in liquid form throughout; two-phase DTC allows the refrigerant to evaporate inside the cold plate to absorb heat; direct-to-die integrates cooling channels directly into the silicon or near the package, reducing thermal resistance but significantly increasing manufacturing and reliability requirements.
Two-phase DTC uses the latent heat of phase change to absorb more heat with a smaller temperature change.
Using water as an example, the report explains that single-phase heating mainly relies on sensible heat, while phase change during boiling allows a liquid to absorb a large amount of latent heat, which explains why two-phase cooling usually has higher heat flux than single-phase solutions under similar conditions.
Laboratory validation does not equal commercial-scale deployment.
The report acknowledges the technical credibility of CoolIT's 15kW cold plate, but still treats scalable manufacturing, deployment economics, competitors' replication speed, and order conversion as key areas for subsequent validation.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- VRTSupplier of data center power and cooling equipment; rated Outperform by the report with a target price of $416.
- Strengths
- The report notes that Vertiv is relatively advanced in two-phase DTC commercial readiness, and that data center cooling demand remains strong at present.
- Weaknesses
- If the lifecycle of the single-phase cold plate pathway is extended, the innovation premium from two-phase DTC may decline.
- Comparison
- An earlier participant in two-phase DTC development, but it may also be affected if breakthroughs in single-phase technology change the pace of migration.
- Risks
- Delay in technology roadmap shift, cooling component commoditization, and competitors replicating CoolIT's solution.
- NVTCooling and electrical equipment-related company; rated Outperform by the report with a target price of $220.
- Strengths
- As an NVIDIA reference design partner, the report believes it should have relevant product capabilities when needed.
- Weaknesses
- Its publicly disclosed progress in two-phase DTC is less clear than that of some leading players.
- Comparison
- Compared with VRT, Accelsius, and others, its visibility in two-phase DTC is lower.
- Risks
- Insufficient product differentiation, cold plate and CDU commoditization, and technology pathway changes.
- TTHVAC and data center cooling-related company; rated Outperform by the report with a target price of $555.
- Strengths
- Expanding data center cooling demand provides medium-term support.
- Weaknesses
- The report does not show that it has made clear public announcements regarding two-phase DTC.
- Comparison
- Compared with companies publicly advancing two-phase DTC, it has disclosed less on the technology.
- Risks
- Declining innovation premium in liquid cooling components and uncertainty around the technology pathway.
- JCICompany related to the data center cooling ecosystem; rated Outperform by the report with a target price of $173.
- Strengths
- The report notes that JCI participates in the two-phase DTC ecosystem through Accelsius.
- Weaknesses
- If CoolIT's single-phase solution is commercialized, demand for two-phase DTC may be delayed.
- Comparison
- Its visibility is higher than companies that have not disclosed progress, as it entered two-phase DTC through investment or partnership.
- Risks
- Pace of two-phase DTC commercialization, PFAS regulation, and slowing customer adoption.
- CARRCooling equipment-related company; rated Market-Perform by the report with a target price of $75.
- Strengths
- The report notes that Carrier invested in ZutaCore and participates in the two-phase DTC ecosystem.
- Weaknesses
- Its rating is more neutral than other covered companies, and the public visibility of ZutaCore's products is limited.
- Comparison
- Similar to JCI in participating in two-phase DTC through external investment, but the report says visibility into specific products remains insufficient.
- Risks
- The commercial window for two-phase DTC being delayed by single-phase cold plates and declining customer interest.
- ETNElectrical equipment and data center infrastructure-related company; rated Outperform by the report with a target price of $534.
- Strengths
- Benefits from demand for data center power and cooling infrastructure.
- Weaknesses
- The report provides less disclosure on its specific liquid cooling technology pathway.
- Comparison
- More of a broad beneficiary of data center infrastructure rather than a pure-play exposure to a single cold plate technology.
- Risks
- Data center buildout pace, technology pathway changes, and equipment commoditization pressure.
- CoolITThe core company behind the 15kW single-phase cold plate technology.
- Strengths
- Its Split-Flow microchannel architecture validated 15kW thermal capacity under conditions including 45°C inlet water temperature, PG25 coolant, and standard thermal paste.
- Weaknesses
- It is still mainly at the concept or test-platform validation stage, and scalable manufacturing and deployment economics remain to be proven.
- Comparison
- If the technology can be mass-produced, it would offer a lower-resistance continuation path relative to two-phase DTC and direct-to-die.
- Risks
- Competitor replication, effectiveness of patent protection, manufacturing cost, order conversion, and production ramp-up.
Key data
- Expected thermal limit of traditional single-phase DTCApproximately 2-2.5kW GPU TDPThe report says the market generally believes this range may be reached around Rubin to Rubin Ultra.
- CoolIT single-phase cold plate thermal capacity15kWThe report says this capability is about 4 times the thermal capacity of existing cold plates and was validated under laboratory/thermal test platform conditions.
- CoolIT test conditionsApproximately 1.2 L/min flow rate, 45°C inlet water temperature, PG25 coolant, standard thermal pasteThe report emphasizes that these conditions are relatively close to industry-standard operating parameters.
- Assessment of two-phase DTC commercialization timingMay move closer to commercial viability over the next approximately 12-18 monthsHowever, the report also emphasizes that reverse flow, refrigerant selection, system control, and PFAS regulation have still not been fully resolved.
- Assessment of direct-to-die scaling timelineLarge-scale adoption not expected before 2030Microsoft, Corintis, TSMC, and others already have proof-of-concept or R&D activity, but deployment remains largely theoretical or in early pilot stages.
- Example of direct-to-die thermal performance>1000 W/cm2The report mentions that bionic microchannel prototypes from Microsoft and Corintis demonstrated thermal capacity performance at this magnitude.
- VRT rating and target priceOutperform, $416Disclosed in the Bernstein ticker table and investment implications section.
- NVT rating and target priceOutperform, $220Disclosed in the Bernstein ticker table and investment implications section.
- CARR rating and target priceMarket-Perform, $75Disclosed in the Bernstein ticker table and investment implications section.
Impact & implications
In terms of investment implications, the report believes that demand for data center cooling equipment still far exceeds supply in the near term, so CoolIT's 15kW cold plate is unlikely to significantly depress revenue or margins over the next few years. However, if the technology scales, the industry's structural migration from single-phase to two-phase or direct-to-die could be delayed, reducing innovation premiums for CDUs and cold plates and increasing commoditization pressure. For cold plates, the downside is that differentiated pricing power may weaken, while the upside is that the risk of direct-to-die making cold plates obsolete over the long term declines. For private companies in two-phase DTC and related investors, customer interest and the commercial window may be squeezed.
Risks
- CoolIT's 15kW cold plate still needs to validate commercial-scale manufacturing capability and deployment economics.
- Two-phase DTC still faces engineering challenges including two-phase flow, local hotspots, system pressure, refrigerant selection, and control algorithm complexity.
- PFAS refrigerants may create compliance, disclosure, and long-term supply chain risks.
- Direct-to-die has high manufacturing complexity and requires very high standards for silicon processing, fluid cleanliness, package sealing, and long-term reliability.
- If single-phase cold plate capability is rapidly replicated, innovation premiums in the CDU and cold plate industry may decline and commoditization may accelerate.
- If future GPU power consumption continues rising above 6kW, two-phase DTC or direct-to-die may still become necessary again.
What to watch
- Whether CoolIT's 15kW cold plate secures commercial orders and scaled deployment.
- Whether CoolIT's Split-Flow patents can create sustainable differentiation, and the speed at which competitors replicate them.
- The pace of GPU TDP increases from Rubin and Rubin Ultra to subsequent Feynman/Kyber platforms.
- Product launches and customer validation from two-phase DTC players such as Vertiv, Accelsius, ZutaCore, CoolIT, Boyd, and Schneider.
- Changes in PFAS regulation and industrialization progress of alternative refrigerants.
- Validation progress and manufacturing yield of direct-to-die within the Microsoft, Corintis, and TSMC CoWoS ecosystem.