Behind-the-Meter Power Generation: Designing Data Centre Campuses for a New Power Reality

The rapid growth of cloud infrastructure and AI workloads is reshaping the scale, pace, and complexity of data centre development. But in many markets, the defining constraint is no longer land, planning, or even demand. It is power.

Grid connection timelines are stretching further out, and in some regions, operators are facing multi-year waits for meaningful capacity. The International Energy Agency has noted that data centres are becoming a growing contributor to global electricity demand, increasing pressure on energy systems worldwide.

For owners and developers, this creates a fundamental question: how do you deliver new capacity when the grid cannot deliver power on schedule?

One solution gaining traction is behind-the-meter power generation (BTM).

BTM generation refers to producing power onsite, supplying the facility directly rather than relying entirely on the public utility grid. In practice, this is often delivered through natural gas generators or turbines, either as a replacement for a delayed grid connection or as a transitional strategy while infrastructure catches up.

What is notable is how quickly this has moved from being an ‘exceptional’ approach to becoming a serious part of mainstream campus planning.

But behind-the-meter power is not simply an energy decision. It is a design decision.

Historically, the grid connection has always been relied upon as an economical means of primary power. The campus is designed around diverse utility supplies, with diesel generation available for operational activities or continuous backup should the utility supply be lost.

Behind-the-meter generation changes that relationship completely.

When power is produced behind-the-meter, generation no longer must only be an emergency layer. It becomes a utility sat within the control of the operator, forming part of the primary infrastructure that contributes to the facilities availability. That shift has major implications for how a campus is located, planned, engineered, and ultimately operated which varies depending on the facility application.

It influences everything from redundancy philosophy, electrical topology and equipment, to spatial planning, maintainability, cooling integration, and long-term flexibility.

This is where early-stage engineering judgement becomes critical.

There have been notable cases where even the best efforts of planning and project management have fallen short to derisk power availability concerns. Developers have secured land, broken ground and signed lease agreements only to find out that the committed power is in fact available but limitations in the transmission systems would prevent the electrons from reaching site. A radical shift had to be made to redefine the project phasing approach and to deploy natural gas bridging power at a reduced capacity.

For those of us supporting lease providers during end user due diligence, a palpable change occurred in their willingness to accept signed letters from utility providers as situations such as that described became more commonplace in the industry. Independent engineering studies were often required by third parties to overcome the hurdle and successfully demonstrate that project power requirements could be met.

These are the realities that do not always show up in early feasibility models, but they define whether a project progresses with confidence.

This conversation is particularly relevant for the next generation of high-density facilities.

AI training and inference environments are driving rack densities well beyond traditional enterprise or cloud norms. With the increase, the margin for inefficiency shrinks and the consequences of early design assumptions become more significant. Power and cooling decisions need to be innovative, forward thinking and able to adapt to future scenarios.

In this context, integrating BTM generation is not simply about producing megawatts. It is about ensuring that generation can be delivered efficiently and reliably into a facility designed for demanding compute environments.

It means thinking carefully about how generation interfaces not only with the wider electrical and mechanical ecosystem, but also the dynamic workloads of the connected chips. IT equipment, Battery strategy, thermal rejection and heat reuse in combined plant systems all become part of the same design equation.

Behind-the-meter power can look straightforward at a concept level: place generation onsite and reduce reliance on the grid. In reality, successful delivery depends on foresight and expert integration.

A campus designed around onsite generation must account for fuel supply resilience, emissions and permitting pathways, acoustic impacts, spatial constraints, and the long-term evolution of the power strategy.

It also requires clarity of intent. Is this a permanent off-grid solution? A bridge to future grid connection? A hybrid approach that will change over time? Those decisions shape the design from day one.

The strongest projects are those where power generation is treated as part of the overall infrastructure masterplan, not an afterthought bolted onto a constrained site later in the process.

Natural gas generation can unlock capacity, but owners are increasingly aware that power strategies must remain adaptable.

Regulatory expectations are evolving, sustainability requirements are tightening, supply chains are stretched and future energy mixes may look very different from today’s.

The most resilient behind-the-meter designs are those that allow optionality over time, whether through future hybridisation, grid integration, or alternative fuels as the market develops.

This is where engineering maturity matters most: designing campuses that solve today’s constraints without creating tomorrow’s limitations.

In this context, behind-the-meter generation is emerging as one of the most practical responses to the current grid reality. For many operators, it offers a route to accelerate delivery, control resilience, and unlock growth in constrained markets.

But it is not a shortcut. It is an engineering-led campus strategy that must be designed holistically, integrated carefully, and aligned with the long-term operational future of the asset.

As the industry moves into higher-density AI-driven infrastructure, the ability to seamlessly integrate onsite generation into overall facility design will become an increasingly important differentiator, not only in speed of delivery, but in long-term performance and certainty.

The data centre projects that succeed in the years ahead will be those that treated power as a core part of the campus design conversation and not as an external dependancy from the very beginning.