DECARBONISATION
rejected using evaporative cooling, because it is the cheapest method available, at real cost to local water supplies. UK data centres generally rely on dry cooling rather than evaporative systems, which is a genuinely better starting position, but we are still, in effect, throwing away a large and usable heat resource. The electricity feeding a data centre is decarbonising over time, just like the rest of the grid. So why would the NHS not capture that heat directly, by hosting data centre capacity on its own sites?
AI workloads could be processed on-site.
of gas a year would need something in the order of an 8MW solar PV array to generate the electricity required to power an equivalent heat pump system, and most of that generation would arrive in summer, not winter, when heat demand is lowest. That is not necessarily wasted (the surplus can be exported to the grid), but the land take involved is stark: in the region of 16 hectares. According to NHS England’s 2024/25 Estates Returns Information Collection, the average land area owned by a general or specialist acute hospital site is just 12.6 hectares. In other words, for a typical acute site, decarbonising heat through PV alone would require more land than the hospital actually occupies, before a single ward, car park or access road is accounted for. That comparison alone explains why PV has never been able to carry the weight of heat decarbonisation on its own.
The AI opportunity nobody has quite joined up At the same time, we are living through an explosion in artificial intelligence. AI tools are increasingly being developed specifically for the NHS estate, to streamline operations and drive efficiency, and medical imaging diagnostics are becoming ever more reliant on AI to deliver higher accuracy. All of this depends on a huge and rapidly growing quantity of data centre capacity to process it.
Data centre development is the next big infrastructure boom, in the UK and globally. But data centres come with an unavoidable physical fact: the computer chips inside them are, in thermodynamic terms, small electrical heaters that happen to do maths. A hyperscale data centre might sit in the region of 500MVA; a smaller, locally sized facility perhaps 10MW. In a well-designed facility, up to 90 per cent of that input power comes back out as heat, and that heat has to go somewhere. This is where the environmental cost of data centres becomes visible. In water-stressed parts of the world (Nevada is a widely reported example), that heat is often
What co-location could look like The opportunity, and it is a growing one, is for data centre operators to co-locate modest, medium-scale facilities on NHS sites, using the heat generated directly within the hospital to provide clean, low-carbon heating. There are a variety of contractual mechanisms that could support this, and the sector will need to think carefully about how they are structured. In broad terms, the NHS has an opportunity to lease land, generating a new income stream from an asset it already owns but does not currently monetise, alongside revenue linked to the data centre’s own use of the site, not unlike a turnover-linked lease arrangement more familiar from retail property. Priced so that the heat supplied is comparable to, or cheaper than, natural gas, this begins to look like a genuinely attractive proposition without requiring the NHS to fund expensive new heat pump infrastructure of its own. A data centre of this kind will typically need its own dedicated electrical connection to site, and it is worth being honest that relatively few hospitals currently have the 33kV incoming supply capacity this requires, though some do, and more could, with the right planning. Structured well, this becomes a mechanism that can be delivered off the NHS balance sheet, protecting Trusts’ own capital headroom. The NHS will still need to retain a proportion of its existing boiler plant for resilience, and it is important that this is tracked and genuinely used only for that purpose, but with the right controls in place, that is a perfectly acceptable and manageable position.
Getting the monitoring right None of this works without rigorous, independent monitoring and verification. The off-balance-sheet position depends on being able to demonstrate, genuinely and continuously, that any retained boiler plant is being used only for resilience, not quietly propping up day-to-day heat demand that should be coming from the data centre. The turnover-linked elements of the commercial model depend on accurate, trusted metering of what is actually being delivered and consumed. And funders providing patient capital over 25 or 30-year horizons will rightly want assurance that the underlying performance data is robust, not self-reported. This is a discipline in its own right, and one the sector
should take seriously from the outset rather than treating as an afterthought. Independent monitoring is not just a compliance box to tick – it is what makes this kind of shared-infrastructure model genuinely bankable. Everyone at the table, the Trust, the operator, the funder, needs to trust the same numbers. That only happens if the metering, the reporting, and the verification sit with a party who is not marking their own homework. Done well, this kind of independent oversight also
Co-locating heat and compute solves two distinct NHS problems with a single piece of infrastructure.
32 Health Estate Journal November 2026
generates a valuable by-product: a genuine, auditable record of how low-carbon assets are actually performing across the estate, over time, exactly the kind of evidence base that has often been missing from earlier generations of NHS energy projects, and that Trusts, funders and regulators alike will increasingly expect to see.
AdobeStock / Sirat
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72
