Addressing the data centre ‘heatwave’ with new cooling solutions

Alongside capacity constraints, the issue of cooling has become a significant limiting factor within the European data centre market. This is certainly true for AI-led facilities because their servers run hotter and denser than traditional cloud racks.

Presently, cooling an average data centre consumes approximately 1.4 million litres of freshwater per day. Facilities relying entirely on legacy air cooling solutions typically record a Power Usage Effectiveness (PUE) between 1.6 and 2.0, meaning that it takes between 60-100% as much power as the servers themselves! Plus – as anybody residing in England this June can attest to – the usage of water only increases due to the droughts and extreme heats Europe now experiences regularly.

The steps being taken

There’s also an environmental factor too, with research from early 2026 revealing that the sheer thermal exhaust from AI-focused data centres can actually raise surrounding land surface temperatures by an average of 2°C – an effect felt up to 10km away. In an age where climate change concerns dominates headlines, this kind of localised heat impact is a PR nightmare for any operator trying to claim environmental responsibility.

Perhaps in part due to this environmental scrutiny, hyperscalers are increasingly investing in closed loop cooling to reduce their reliance on freshwater systems and find new ways to be sustainable. A press release from the European Union’s Directorate-General for the Environment in March 2026 underscored this shift, highlighting how waste heat from AI data centres on the continent can be repurposed for water purification and even carbon capture processes.

This is timely, since a number of European regulators have begun tightening rules on water consumption within the facilities based within their borders. The Netherlands, for example, has introduced major environmental requirements for water-cooled data centre facilities, while the Republic of Ireland now requires several data centres to switch to air-cooling in order to reduce water draw. This has led hyperscalers to consider air-cooling, liquid cooling or hybrid systems, all of which require more electrical power – slightly concerning when capacity issues have already resulted in several key data centre projects being shelved or postponed.

Hotter coolant, cooler data centres

Take NVIDIA as an example, as the company’s latest liquid cooling architecture has marked a major shift in how AI data centres manage heat. By running Rubin-generated servers entirely on 45°C coolant – which is hotter than a hot tub - NVIDIA advised in June 2026 that it has eliminated the need for traditional chilled water systems and fans. This approach dramatically reduces water consumption, cuts cooling energy costs and enables higher‑density racks with more predictable thermal behaviour.

Through this approach, operators are empowered to reject heat using ambient air for most of the year, even in warmer climate conditions, while still opening the doors to waste-heat recovery for nearby buildings.

This breakthrough could be game changing, because we’re already seeing conventional cooling models increasingly become unsustainable. If fully liquid cooled platforms can replace expensive mechanical cooling solutions with closed loop water systems, then businesses have breathing room to deliver hyperscale growth while reducing their environmental impact and simplifying their thermal designs across the entire server. 

 Marketing innovation

Immersion cooling is also gaining traction as a comparable next generation approach. Here, servers are fully submerged in dielectric fluid to eliminate the reliance on chilled water and mechanical cooling. Arguably, immersion systems go one step further than NVIDIA’s liquid cooling solution as they remove cold plates, pumps and complex plumbing entirely. As a result, hyperscalers and colocation providers are increasingly piloting immersion tanks as a way to cut water consumption and future-proof their facilities against rising AI power densities.

For the businesses offering these solutions, their marketers now face a new challenge: in a market crying out for cooling innovation, differentiation has become a brutal task. Every vendor is racing to position its solution as the answer to AI-driven thermal and sustainability pressures, yet the underlying narratives often sound identical – efficiency gains, water savings, density improvements and the delivery of environmentally-friendly operations.

With cooling architectures focused on similar capabilities, the brands that stand out will be those able to translate complex engineering into clear, credible value propositions, and why these matter in an era where thermal performance has become a board-level concern.