03 October 2025
Bernie Malouin, JetCool Founder and Flex Vice President
Artificial intelligence (AI) is transforming every industry and putting unprecedented demands on the infrastructure that powers it. In the UK, data centre operators face a perfect storm of rising energy costs, tougher sustainability mandates, and the challenge of keeping legacy systems in step with next-generation hardware. Among all these pressures, a major challenge looms: cooling.
With major chipmakers and hyperscalers now talking about rack power densities that could reach 1 MW — and AI workloads driving 24/7 demand — it’s clear that traditional air cooling won’t be enough for what’s coming. The question for operators is no longer if they should modernise their cooling strategy — but how quickly they can prepare, and whether they’re choosing the right technologies to stay ahead of the curve.
Is there a clear winner in cooling technologies?
Yes — single-phase direct-to-chip liquid cooling is rapidly becoming the industry standard for high-density AI deployments. It provides the thermal performance, reliability, and scalability needed to keep pace with the next generation of processors and GPUs.
That said, many operators are starting with a hybrid approach - combining direct-to-chip liquid cooling with liquid-to-air heat exchangers - because their facilities aren’t yet plumbed for facility-wide liquid distribution. While this setup is less efficient than full facility-level liquid cooling, it’s a practical first step that lets operators pilot the technology, build internal expertise, and get comfortable with liquid-cooled infrastructure before a full rollout.
In the UK, where energy efficiency is critical, direct-to-chip solutions have a clear advantage. Their ability to operate at inlet coolant temperatures above 45 °C unlocks year-round free cooling in most environments and dramatically reduces reliance on mechanical chillers — cutting both operating costs and carbon footprint while future-proofing data centres for the demands of AI.
Why the liquid cooling market is accelerating
The liquid cooling market is evolving faster than ever. We are seeing a surge in demand for comprehensive liquid cooling providers, not just vendors selling cold plates, but partners offering full-stack solutions including coolant distribution units (CDUs), controls, and service.
Beyond cooling, the real constraint is power. Many UK data centres are reaching the physical limits of air cooling just as their energy budgets are tightening. In some cases, GPUs are being throttled not because of compute constraints, but because there simply is not enough power available, and worse, too much of it is being wasted on inefficient air cooling systems. Using liquid cooling, even with hybrid systems, operators can get more compute in their fixed power budgets or can use the power savings to save on energy costs.
This power bottleneck demands immediate action. Improving system-wide efficiency and reclaiming stranded power is no longer a future goal; it is a strategic imperative essential to unlocking better performance, greater utilisation, and an optimized total cost of ownership.
Why delaying cooling upgrades can be costly
Looking five years ahead, data centre leaders will regret one thing: waiting. Waiting for the next chip generation before modernising infrastructure. Waiting to test liquid cooling at scale. Waiting to move beyond legacy air systems. Waiting to start deploying liquid cooling, even in a small capacity, to start building processes, procedures, and team knowledge.
The cost of inaction will be evident in stranded capacity, underutilised GPUs, and missed sustainability targets. Fortunately, modern cooling solutions do not require a complete overhaul. Self-contained or hybrid systems allow operators to implement liquid cooling quickly while planning full-scale deployments.
Early adopters will be the ones who scale fastest when demand peaks, and they will be best positioned to meet the UK’s ambitious carbon reduction goals.
Top tips when implementing direct-to-chip cooling
1. It is not all or nothing: Businesses do not need to rebuild their facility to get started. Retrofit solutions can integrate seamlessly into existing air-cooled racks, enabling a phased approach to modernisation.
2. Hot coolant is a feature, not a flaw: High-temperature liquid cooling systems thrive at >45°C inlet coolant temperatures, unlocking free cooling opportunities even in warmer environments. This is especially valuable in the UK, where energy efficiency and sustainability are top priorities.
3. Cold plates are just the beginning: Peak performance requires every part of the cooling loop to deliver. Cold plates alone will not solve the problem, meaning operators must look at the full solution, from CDUs and controls to monitoring and service.
The time to act is now
AI is not waiting. Neither should UK data centres. The digital infrastructure must evolve to meet the demands of tomorrow’s workloads, and that evolution starts with smarter, more efficient and simpler to deploy cooling strategies.
Whether an organisation is building a new facility or retrofitting an existing one, the key is to start now, start smart, and scale fast. The future of AI-ready data centres will be defined not just by compute power, but by how well they manage heat.
