The energy and sustainability conundrum
Rising power density requirements are colliding with grid constraints, sustainability expectations and public scrutiny. As a result, energy is shifting from a technical consideration to a core strategic and commercial priority for data centre development.
With power under severe constraints in Tier 1 markets, it is an economic imperative for data centres to be as energy efficient as possible. This is increasingly challenging as demand for power density increases: 79% of respondents expect occupant demand for power density to increase in the next three years – and 40% of those respondents expect it to grow ‘significantly’.
But the pressure to boost sustainability – from customers, regulators, investors and the public – is growing too. Some operators, such as hscale, have investors with strict ESG commitments. Meanwhile, public scrutiny of the data centre sector has intensified in the past 12 months, especially over the environmental impact of AI facilities, prompting even more stringent reviews from environmental agencies.
For many, though, the most stringent environmental demands come from occupants, especially hyperscalers. Occupants’ demand for access to renewable energy is growing almost as intensely as power density, respondents say (Figure 4).
Figure 4. Occupants’ demand for both power density and renewable energy are intensifying. How do you expect demand from data centre occupants for the following features or capabilities to change over the next three years? (percentage of operators and contractors)
For the second year running, ‘reducing carbon emissions and increasing energy efficiency’ is the number one design priority among operators, ahead of both ‘meeting the needs of AI’ and ‘reducing cost’.
Improving overall energy efficiency is also the primary driver behind operators’ plans to retrofit their capacity in the next five years – 66% expect to retrofit at least a quarter of their current facilities in that time.
Respondents are optimistic that sustainable energy measures will ease grid constraints: 69% agree that growth in alternative energy sources, such as private PPAs, on-site solar and battery storage, will help reduce reliance on constrained public grid infrastructure in the next three years.
Slow progress on sustainability
Despite all this, progress on energy sustainability is slow. Since last year’s survey, the adoption of key measures such as on-site renewable generation, the use of materials that boost energy efficiency, and incorporating on-site heat/energy management innovations, has barely progressed (Figure 5).
Figure 5. Adoption of many sustainability measures has barely progressed in the past year. Which of the following has your organisation adopted in data centre projects, specifically in response to environmental regulation and/or to meet sustainability commitments? (percentage of operators and contractors, ‘adopted’)
This inertia may reflect the high capital cost of these measures. But data centre operators should think long term: grid constraints and high energy costs in Europe mean that almost any investment in energy generation or efficiency will provide a return. And a study by the International Renewable Energy Agency found that 91% of renewable energy projects are cheaper than fossil fuel alternatives. In other words, the energy decarbonisation agenda and the need to cut costs are aligned.
Operators may also be cognizant of the risks these new energy solutions bring. “If you're being more innovative – with battery solutions, for example – you're taking on more construction to deliver your own power,” explains Lockton’s Baker. “That introduces additional construction and technology interface risk, and increases the proportion of risk retained by the owner.”
Pioneering approaches to energy sustainability are easier in some locations than others. For example, ‘zero gen’ facilities, which have no or reduced onsite back-up generator (often the greatest source of emissions) are emerging in the Nordics, where multiple external sources of renewable power are available.
This approach is currently attractive for AI training workloads, Mercury Engineering’s Byrne explains, for which a power outage is not as critical as cloud computing. “If it’s machine learning only, there is potential for a reduced redundancy.”
Elsewhere, grid constraints are driving other energy innovations. AVK has been involved in some of the earliest data centre microgrid projects in Europe. “We’ve seen circumstances where developers need to monetise a piece of land and the hyperscalers want it, but the grid can’t support it, so they had to find a solution,” explains Pritchard. “That has brought everyone on board with microgrids.”
Currently, these local grids are powered by a mix of energy sources, including gas-powered engines, BESS and renewables.
These engines could be fuelled by hydrogen, but the infrastructure is not yet mature enough to make this viable, says Pritchard. “The reality is that the fuel availability for hydrogen at scale isn’t there yet.”
Power partnerships
The power and sustainability conundrum is a complex problem that must be solved collectively. Survey respondents are hopeful: 64% agree that collaboration between developers, OEMs and energy providers will unlock new capacity and reduce grid access delays over the next three years.
hscale is working with utilities to secure access to power that is committed to other customers but not currently in use, says Stephenson. “If a competitor next door has secured 60MW to be delivered in phases over three to four years, it will take at least six months per phase for them to ramp up to 60% or 70% utilisation. That means there is ‘spare’ capacity in the grid, i.e. contracted but not being utilised. We see an opportunity to secure non-firm power to overcome shortfalls while the utility increases our own supply, or indeed whilst we permit and construct our own self-generation solutions.”
“Not every utility is flexible or willing to treat that as an opportunity, but some European markets are open to it,” he adds.
The most common form of partnership is a long-term commitment between utilities, energy providers and government bodies. These allow the energy provider to build capacity while the operator secures its long-lead equipment. Given the volume of demand, energy providers have started to introduce limitations on how long they will reserve power, e.g. for two to five years. But staying in close communication allows all parties to understand one another’s evolving needs.
Beyond collaboration, energy constraints might also give rise to entirely new types of development: 63% of survey respondents agree that in future, data centres will increasingly co-locate with other advanced tech facilities, such as semiconductor fabs, AI/quantum labs and clean tech storage sites. The idea is that sites with abundant clean power should be put to maximum use.
There is a lot to resolve before this type of facility becomes reality: multi-purpose sites will require a complex design and, therefore, a complicated construction process. Sustainability issues also increase with the scale of a facility.
But adjacencies are already emerging, with power generation or storage facilities being developed next to data centre sites, enabling direct, ‘behind the meter’ supply.
Pritchard expects this opportunity to attract new investors, such as infrastructure funds, who have experience in backing power plants, in the near future. “That will allow data centre investors to focus more on building the actual data centres, without their investment being dwarfed by the power piece that needs to be built,” he says.
Liquid cooling is still a work in progress
The power profile of AI facilities is still being determined, but it is certain that AI workloads outweigh the ability of air-cooling techniques to dissipate heat.
Meanwhile, the water consumption of data centres has become the focus of public opposition to data centres. The largest hyperscale facilities can reportedly consume up to 5 million gallons of water a day, roughly equivalent to the daily consumption of a town of 50,000 people.
Liquid cooling offers a solution to both these challenges, dissipating heat more efficiently and – despite the name – consuming less liquid than traditional methods.
Last year, however, the industry was still evaluating how exactly liquid cooling would be implementing and to what extent. This helps to explain why, despite the race to build AI-ready capacity, a long-awaited breakthrough on liquid cooling adoption has still yet to happen. Operators expect 44% of their data space to be liquid cooled in the coming year, on average. This figure has barely changed since this study began in 2024 (Figure 6).
Now, though, liquid-to-liquid (L2L) cooling is part of the baseline design for every new facility, Byrne says. "We don't see designs without the capacity to provide L2L cooling anymore.”
hscale’s Stephenson adds that the leasing pause gave operators that had planned facilities for air-cooling the chance to pivot their designs. “It’s allowed competitors to catch up.”
There is still much to be resolved, however. For example, there is uncertainty around commissioning. Testing liquid cooling requires electric boilers to simulate the heat of IT equipment. This is straightforward for the global loop, which moves water around the whole facility, but testing the connections to individual servers is harder and riskier. It is currently not clear where a contractor’s responsibilities begin and end, says Byrne.
Byrne expects these issues to be resolved in the near future. “In 12 to 18 months' time, people will know what they want,” he says. When that happens, operators can expect liquid cooling equipment and skills to be in high demand.
Figure 6. A breakthrough on liquid cooling has yet to happen. What proportion of the data centre space your organisation operates will be liquid cooled [in the coming year]? (percentage of operators)
Share this on social media:
RLB © 2026 Rider Levett Bucknall.