When the data centre becomes the next office tower

The office tower was not a bad building. It was a good building for a demand pattern that changed. That is the general form of the commercial property accident: capital committed to a specific use, underwritten on a demand curve that held right up until it did not.

Data centres are being built at a pace that only makes sense if demand continues compounding. Much of that capacity is genuinely contracted. Some of it is not, and the two are increasingly hard to tell apart from outside.

What makes the asset fragile

• Single use. A shell built for high-density compute, with power and cooling designed around it, converts to almost nothing else at a sensible cost.
• Single tenant. Concentration that would be flagged in any other property type is normal here, and the covenant strength varies enormously between a hyperscaler and a specialist compute provider funded by private credit.
• Fast obsolescence. Power density requirements have moved repeatedly. A shell designed for the last generation of hardware is not automatically suitable for the next.

Office space was rescued, where it was rescued at all, by conversion. There is nothing to convert a purpose-built compute hall into.

The financing is the risk, not the concrete

Speculative development funded with short-dated construction debt is the part that breaks first, exactly as it did in every previous property cycle. Leases signed at today's rents get refinanced at whatever rate exists at maturity, and the developer's equity is the buffer between those two numbers.

The lenders are frequently the same private credit vehicles carrying exposure to the operators renting the space. That is a correlation nobody underwrites, because it only shows up when both sides go wrong together.

The public backlash arrives before the financial one

Data centre construction competes with households for electricity and, in some regions, water. Retail rate increases attributable to large loads are already a political question. Planning refusals and grid connection delays are the mechanism by which the political question becomes a financial one, and they arrive years before any repricing.

What to watch

• Contracted versus speculative capacity in new development, disclosed by developers rather than inferred.
• Covenant quality of the tenants, and how much rent depends on companies that are themselves funded by private credit.
• Construction loan maturities relative to lease commencement dates.
• Utility rate cases and grid connection queues in the largest build markets.