Patient capital for impatient demand.

The mismatch.

The customer wants the capacity in eighteen months. The asset will run for thirty years. The money, more often than not, wants out in five. That triangle, impatient demand, long-lived infrastructure and short-dated capital, is the defining structural problem of the AI buildout, and it is where most developers will fail. Not because their sites were wrong or their engineering was poor, but because they financed a thirty-year asset with five-year money against a two-year customer.

The symptoms are already visible. Speculative shells built ahead of any offtake, waiting for a tenant who may sign with someone else. Bridge loans rolled into bridge loans. GPU counterparties whose balance sheets are thinner than their ambitions. The demand is real. The capital structures underneath much of it are not.

Match the money to the asset.

Renewables solved this problem twenty years ago, and the solution transfers almost unchanged. A long-dated contract turns a construction site into something that behaves like a bond, and a bond can be financed by the people who own thirty-year liabilities: pension funds, insurers, sovereign wealth. The trick in solar and wind was the power purchase agreement. The trick in Data Centres is the offtake agreement, and the discipline of not building until you have one.

“Renewables taught us the trick: a contract turns a construction site into a bond. Data Centres are the same trade, with a bigger number and a more impatient customer.”Jamie MacDonald-Murray, Chairman & CEO, Eppur

Our capital stack is built for that. Platform equity carries origination and development, the risk-bearing early years. Project-level debt sits against each campus once land, power and consent are secured. Offtake-backed financing, underwritten by hyperscale and sovereign counterparties, funds delivery. Each layer is sized for the stage it funds and the duration it can bear. Capital partners, lenders and compute counterparties move on the same timelines we do, because we designed it that way.

Where the risk actually sits.

The technology is not the risk. Chips will get faster, cooling will get denser, and a well-designed campus will absorb both. The risk sits in the stack beneath: land options that expire before the grid offer arrives, grid offers with deadlines the planning process cannot meet, tenants who need the capacity before the debt can close. Managing that stack is the whole job.

It is why we report our pipeline the way we do. Anchor projects are sites with land and power confirmed. Everything else is pipeline, labelled by the stage it has actually reached. It is a less exciting number than some of our peers publish. It is also the number an investor can underwrite.

What patient capital earns.

Three revenue streams, one compounding asset base. Development margin from de-risking land, power and permits. Recurring contracted income from campuses in operation. And equity participation in the compute itself, so we capture value from the silicon and not only the shell. Every capital decision is tested against a single question: does it strengthen the quality of the asset base and the durability of the income? If it does not, we do not do it, however impatient the demand.

“Impatient demand is a gift if you have patient money. It is a trap if you do not.”Jamie MacDonald-Murray, Chairman & CEO, Eppur

The AI era will be built by developers who can hold two clocks in their head at once: the customer’s, which is counting months, and the asset’s, which is counting decades. We have done this before, in an industry that also wanted everything yesterday. The buildings are different. The discipline is the same.