Interview With Alex Marshall, Group Director At Clarke Energy, A Rehlko Company On The Data Centre Industry

Why do you believe the data centre industry is dealing with a structural power constraint rather than a temporary power shortage?

Because the mismatch is one of timescales, not volume. Data centre demand is arriving on grids that are already managing the retirement of coal and some nuclear capacity, rising penetration of intermittent renewables, and the electrification of heating and transport. Each of those is a decade-long trend, and they are compounding rather than resolving. On the supply side, the response times are slow, new transmission takes years to consent and build, large transformers have multi-year lead times, and grid connection queues in several markets now stretch into the 2030s. A temporary shortage clears when supply catches up. Here, the demand drivers are growing faster than the delivery mechanisms can respond, and that gap is structural.

Many people see speed and sustainability as competing priorities. Why do you think that has become a false choice for data centre developers?

The choice is only real if the assets deployed for speed, are dead ends. Rushing to power can lock in long-term carbon and forfeit future optionality, but it does not have to. The generating assets most deployed on data centre sites, such as gas engines, have credible routes to de-fossilisation. Combined cooling and power raise the overall efficiency of the plant by putting the engine’s heat to work. The fuel can move from natural gas to biomethane or biomethane blends today, and to hydrogen as supply develops. The operating role can shift from baseload to backup and grid support as cleaner grid supply arrives. Carbon capture, combined with biomethane, can take the plant to net carbon negative. Biomethane addresses a growing credibility gap in the industry: it closes the distance between renewable energy claimed through certificates and offsets and renewable energy physically used on site. The question a developer should ask is not “fast or clean” but “does this asset have a transition path or not.” If it does, speed and decarbonisation are sequenced rather than opposed.

How can energy autonomous strategies help developers bring new AI capacity online faster without abandoning long term decarbonisation goals?

Onsite generation removes the grid connection queue as the binding constraint on delivery. A developer with dispatchable generation on site can energise a facility on the equipment lead time rather than the utility’s timeline, which in constrained markets is the difference between years. The decarbonisation goal survives if the strategy is designed as a bridge rather than a destination: fuel-flexible assets, contracts that allow the operating role to change, and a plan for how the onsite plant transitions from primary supply to backup and flexibility as grid capacity arrives. Energy autonomy done properly is not a rejection of the grid. It is a way of arriving at the grid on your own schedule, with assets that still have a job when you get there.

You mention building infrastructure for transition instead of replacement. What does that look like in practice, and why is it becoming more important?

Replacement thinking assumes each asset is scrapped when something cleaner arrives. Transition thinking asks how the same asset changes role and fuel over its life. In practice that means specifying engines that can move from natural gas to biomethane or hydrogen blends; designing the electrical and control architecture so a plant built as primary power can later operate as backup, peaking, or grid support; recovering heat where there is a thermal use for it (and/or leaving space for integration of absorption chillers); and writing commercial arrangements that preserve those options rather than foreclosing them. It matters more now because the industry is deploying generating capacity at unusual speed. Decisions made in the next three years will sit on sites for twenty. Building for transition is how you avoid owning a fleet of stranded assets in the 2030s.

How can onsite generation, battery storage and hybrid energy systems change the role data centres play within the wider energy network?

They turn the data centre from a passive load into a controllable one. A facility with generation and storage behind the meter can reduce its draw at peak, export or provide balancing services when the grid is stressed and absorb power when it is abundant. That changes the negotiation with the network operator: a flexible load is easier to connect than an inflexible one, because it adds less to peak demand, which is what the system is built for. Most grids carry substantial unused capacity outside peak hours. Facilities that can flex into that headroom get connected faster and become part of how the system balances, rather than another problem it must absorb.
 

The UK has seen investment in onsite generation slow following policy changes. Do you think current grid constraints could lead to a renewed focus on these energy strategies, and what would need to change?

Yes, and the logic has already shifted. The earlier wave of UK onsite generation was built substantially on embedded benefits, and when charging reforms removed much of that value, the investment case weakened. The new case does not depend on tariff arbitrage. It depends on connection timelines: when the alternative is waiting years for a grid connection, onsite generation is valued on speed to power and on the revenue a flexible asset can earn, which is a more durable foundation than a charging methodology.

What needs to change is partly structural. Unlike other markets such as the USA, the UK offers little support for integrated microgrids that combine generation, storage and load behind a single connection. The policy landscape instead favours grid-level balancing: standalone batteries and dedicated solar built for export. That leaves the UK well supplied with flexibility at system level but poorly equipped to serve large loads that need power faster than the connection queue allows. Beyond that, the fixes are mostly about clarity. Planning treatment of onsite generation at data centres remains inconsistent, connection reform needs to reward flexible loads with faster access, and the flexibility markets need enough visibility for developers to underwrite against them.

Looking ahead over the next 10 to 15 years, what should developers be doing today to make sure the data centres they build remain resilient, flexible and commercially viable?

Treat power as a strategic asset rather than a procurement line. Concretely: buy optionality in the generating plant, meaning fuel flexibility and the ability to change operating role over the asset’s life; design for grid interactivity from day one, because a facility that can flex will be worth more to the system than one that cannot; take thermal output seriously where there is a use for it, since rejected heat is unmonetised fuel; and stress-test the commercial structure, not just the engineering, against a decade in which power markets, carbon rules, and grid access regimes will all move. The facilities that stay viable will be the ones designed on the assumption that conditions change, rather than the ones optimised for the conditions of 2026.