Underwriting the residual on a 6 kW per rack building in a 60 kW world

Question

Chasing a question I can't settle. A facility built around 2014 for something like 6 to 8 kW per rack, air cooled, decent connectivity, on a feed with modest headroom. Fully leased today at rents that look fine. The seller's residual assumption is a modest cap rate expansion off in-place rent at year ten.

My problem is that the workloads driving all the demand everyone talks about don't fit in that building. High density AI racks want liquid cooling and a power density the floor was never designed for, and the retrofit isn't a chiller swap. It's floor loading, containment, distribution, sometimes the roof. So the residual buyer in year ten is either an enterprise tenant with ordinary workloads, which is a shrinking but real pool, or a developer buying the parcel and the interconnection and gutting the building.

Two defensible ways to underwrite that. Value the shell and the fit-out on remaining useful life, accept obsolescence, and hold rents flat with a wide exit cap. Or ignore the building almost entirely and value the interconnection capacity plus the fiber routes plus the land, on the argument that in a market where power delivery takes years, an energized site with existing capacity is worth more than the box sitting on it, and the box just has to cover carry until somebody wants the site.

Those two paths give me residuals about 40% apart on the same asset. Which one would you actually put in the model?

nicole_adeyemi · Answer

Land and interconnection, with one condition. Check whether the capacity is transferable and what happens to it if the load drops. Some utility arrangements tie capacity rights to the existing customer or reduce them if usage falls off for a period, and those terms differ by utility and by state tariff, so it's a document question rather than a general one. If the capacity walks out with the tenant, your whole second path evaporates and you're back to valuing a 2014 box.

The other thing I'd price is the zoning and the neighbors. Gutting a building for high density in ten years is a permitting exercise, and if it sits somewhere that's grown residential around it since 2014, the developer buyer you're counting on may not get approval for the noise and water side of a liquid cooled build.

brandonlee36 · Answer

I'd model the first path and treat the second as upside you don't pay for. Not because obsolescence is certain, it's that a residual built on a hypothetical redevelopment buyer ten years out is the same assumption sponsors were making about suburban office in 2015. parcel's point on transferability decides how much of that upside is even real, and I'd want the utility letter before I let it move the number by 40%.

Underwriting the residual on a 6 kW per rack building in a 60 kW world

2 replies