Single hyperscaler at fifteen years versus forty colocation tenants rolling every three, which data center risk is worth holding for a decade?

A. The entire case for it is that nothing is asked of the owner. Forty tenants and a general manager is a job with a rent check attached, and the moment the market stops marking up you are running an operating company nobody wanted.

Verity, that's fair but the fifteen year lease isn't nothing either. You still own the roof and probably the structure, and at some point somebody wants a density upgrade and the lease decides who funds it. I've watched enough scope grow past its estimate to know that a landlord-funded mechanical rebuild inside a fixed rent is the worst seat in the building.

A question rather than a vote. On deal A, what do the renewal terms look like? If the fifteen years runs to market renewals at the end, bumps in the low twos cost far less than they do when the tenant holds three five year options at a fixed rate. That single detail decides most of the argument about handing over the upside.

Ten years is not a long hold. I'd take A and stop thinking about it, because the thing that makes B work is a supply shortage, and shortages are the one condition nobody sensibly extends out a decade. Power delivery is slow now. Slow isn't forever.

The financing outcome differs more than the equity math suggests. A fifteen year investment grade lease finances at terms a forty-tenant colo with a three year WAULT will not see, and the colo lender will want reserves and probably a cash sweep tied to occupancy. Part of the colo's higher going-in yield is simply payment for the debt being harder to place. Confirm any of that with the actual lender in writing, terms move.

On deal A the assignment and change of control language deserves attention before anything financial. Very large technology tenants sublease and reassign inside corporate families, and if the lease permits assignment to an affiliate without a credit test, a fifteen year investment grade lease can end up sitting against an entity that was never underwritten. Whether that survives depends on how the guarantee is drafted, and a lawyer in the state where the lease sits should be the one to read it.

Warrant's point is the one that would move my vote. I was leaning A entirely on the credit and if the credit can migrate then A's whole case is thinner than the spread suggests. Quill, the option rents matter too. Low twos over fifteen years with fixed options at the same escalator is a long time to be under market in a sector everyone says is short of space.

B has a business attached, and businesses can be sold separately from buildings. That's either an extra exit or an extra thing to mismanage.

Run both to a residual and the argument mostly resolves. On A the exit value is a function of remaining term at sale, so a ten year hold on a fifteen year lease means selling five years of term, which is when cap rates on single tenant assets get ugly. On B the exit is a stabilized operating platform sold at whatever multiple colo trades at then, plus seven or eight years of mark-to-market rent along the way. Underwriting tends to land on B being worth more in the upside case and much worse in the case where a single large customer in that forty leaves and takes 30% of the load with it. So check the concentration inside B. Forty tenants sounds diversified until the top three are half the revenue.

Sextant said it. "Forty tenants" is a headline. I'd want the top ten by megawatt and by revenue before I called that diversified, and I'd want to know how many are resellers who could vanish in a quarter.

Running a small team is the part underweighted in this comparison. The multi-tenant colo structure needs a general manager who understands power, cooling, and customer contracts, and that person is expensive and hard to replace. If they leave in year three, that's when it becomes clear how much of the renewal rate was actually relationships. That isn't a modeling risk, it's a hiring risk, and it's worth knowing who's actually on the payroll before underwriting the renewal assumption.

Reading this thread as someone still working out where they fit, A sounds like real estate and B sounds like a technology services business that happens to own a building. Those aren't the same skill and I don't think they should be compared on cap rate.

Plumb, that's the cleanest framing so far. I've hit my own ceiling by taking on deals where the operations turned out to be the actual asset. I voted A because I know what I'm bad at.

I'd want the interconnection capacity on both before voting. If A's tenant is using 80% of available capacity there's no expansion story and the residual is a leasing problem. If B has headroom on the feed, then the mark-to-market story has room to keep running and the going-in yield is not the whole return. Kestrel, do you have the energized versus contracted numbers for each?

Naive question. On deal B, if a customer leaves, do they take their own servers out and the space is just empty, or does the owner have to rip out cooling and power that was built for them specifically?

Wren, that depends on how the space was fitted, and it is the reason colo churn costs more than office churn per square foot. Custom power density and containment do not transfer cleanly to the next customer. Kestrel, a slower question worth adding to windlass's fast one: what is the historical churn rate in B, by megawatt, for the last five years, and how much of the recent rent growth is renewals versus new space? If growth is mostly renewals, that points to a pricing recovery, which can run its course and finish.

A recurring lesson across ownership of multifamily assets is that turnover cost is the number that decides whether a high yield asset is actually high yield. Forty tenants at three year terms means constantly absorbing turnover. If the fit-out cost per departing customer is meaningful, B's mid sevens is not really mid sevens once that cost is priced in.

Five years into running a service business would change how anyone reads deal B. The seller is selling a team's performance and calling it a yield. Meeting the general manager before the building is the right order of operations.

Nobody's asked what happens to A's building if the tenant leaves and the market has moved to a different power profile. In small markets that's a shell with a substation and a very short list of buyers. A is still the better hold, but it deserves a wider cap than low sixes.