Two‑phase liquid cooling poised to transform high-density data centres

As rack densities soar above 70 kW, data centres are increasingly adopting two‑phase liquid cooling to manage heat loads, promising higher efficiency, operational savings, and large-scale deployment prospects amidst a challenging thermal environment.

Data centre operators confronting an intensifying thermal crisis are increasingly turning to two‑phase liquid cooling to manage heat loads that conventional air and single‑phase water systems can no longer handle. According to the original report, rack densities now routinely exceed 70 kW and the physics of two‑phase cooling , using a dielectric fluid that absorbs heat as it vaporises , offers a capability to dissipate heat levels that would overwhelm legacy approaches. ^[1]

The technology’s basic advantage is thermal efficiency through phase change rather than mechanical work: when the refrigerant or dielectric fluid transitions from liquid to vapour it absorbs large amounts of heat at near‑constant temperature, allowing much higher local heat flux removal with far lower flow rates than water‑based systems. Industry vendors highlight operational benefits including reduced pump power, lower fan loads and a safety profile where non‑conductive fluids eliminate the electrical risk associated with water leaks. The original report quotes Matt Roberts, Vice President of Sales at OptiCool Technologies, explaining that two‑phase systems “do not face those limitations because the refrigerant’s phase change from liquid to vapour does the work for you.” ^[1]

Those performance claims are being matched by a shift from proofs of concept to planned production deployments. The most prominent recent example is Accelsius’s NeuCool technology being selected for a 300 MW campus in Ontario by DarkNX , a deployment described by Accelsius and industry press as potentially the largest two‑phase, direct‑to‑chip installation to date, with the first two 65 MW facilities slated for 2026 and 2027. The company says this win signals a move toward full‑scale adoption in large AI‑oriented facilities. ^[2][6][7]

Vendors stress different routes to integration that aim to reduce disruption to existing operations. Modular architectures composed of pumps, overhead manifolds and rear‑door heat exchangers are presented as “Lego block” building blocks that can be configured around a site’s constraints and work with racks from multiple manufacturers. According to the original report, OptiCool’s design allows connection to existing chilled water plants or, where those are absent, DX‑based outdoor units to form closed loops, and Accelsius claims its systems are effectively plug‑and‑play where a facility is already pumped for liquid cooling. ^[1]

Retrofitting remains a technical and logistical challenge. Cleantech consultants note that adding two‑phase at rack level typically requires evaporative cold plates or immersion tanks, secondary condenser loops, new manifolds, containment systems and tighter monitoring to manage phase behaviour , work that is far from trivial in many legacy facilities. The original report quotes François Le Scornet of Carbonexit Consulting noting that retrofitting “is not completely straightforward at all to say the least.” White papers from systems integrators also argue that two‑phase direct‑to‑chip solutions can be easier to adopt than full immersion in some contexts because they use less dielectric fluid and minimise added weight and enclosure changes. ^[1][5]

The economic case is increasingly compelling for high‑density deployments. Data cited in the original report shows two‑phase zones can achieve partial PUEs (pPUE) near 1.05–1.10 versus typical whole‑facility PUEs of 1.3–1.6 for air‑cooled halls, and vendors estimate substantial reductions in annual operating cost and maintenance compared with CRAC/CRAH‑led approaches. François Le Scornet observed that the ROI becomes attractive when high‑density racks allow more compute per square metre and reduce chiller dependence, while other vendors argue lifecycle savings multiply at hyperscale. ^[1]

Proof points and market outreach are multiplying: Accelsius has been demonstrating its technology to European customers via a Thermal Simulation Rack showcased in Global Switch’s London showroom as part of a “Future Now: London Live Liquid Cooling Showcase,” highlighting the vendor’s push to familiarise operators with two‑phase systems in market settings. Meanwhile, other suppliers of two‑phase immersion and direct‑on‑chip solutions emphasise sustainability and long‑term reliability as differentiators from older cooling models. ^[3][4]

Despite momentum, supply‑chain constraints, unfamiliarity among many operators and the organisational complexity of large retrofits will temper adoption rates in the near term. Vendors, early adopters and consultants interviewed in the original report say that the next five to ten years will see significant growth in two‑phase use among AI‑centric and latency‑sensitive sectors , but that the transition will be phased, with new builds and greenfield AI campuses leading the way while many existing facilities pursue hybrid approaches. Accelsius’s Josh Claman told Data Center Knowledge: “Why would anyone expect data centres with some of the highest heat fluxes over the next five to 10 years not to go to two‑phase?” , a sentiment that captures both the technology’s potential and the strategic choices ahead. ^[1]

📌 Reference Map:

##Reference Map:

• ^[1] (Data Center Knowledge) – Paragraph 1, Paragraph 2, Paragraph 4, Paragraph 5, Paragraph 6, Paragraph 8
• ^[2] (Data Center Dynamics) – Paragraph 3
• ^[6] (Data Center Dynamics) – Paragraph 3
• ^[7] (Data Center Dynamics) – Paragraph 3
• ^[3] (GlobeNewswire) – Paragraph 7
• ^[4] (Opteon) – Paragraph 7
• ^[5] (Chatsworth Products) – Paragraph 5

Source: Fuse Wire Services