Bought a 1.5 MW colo as a passive hold and inherited an operating business

The chiller line is painful and familiar. "Operational" in a mechanical report means it turned on the day the guy was there. Running a two compressor unit on one for over a year is a load problem waiting for the first hot week, and in a room full of servers there's no forgiving day.

So colocation means the customers own their own servers and they're renting space, power and cooling from you? And a cabinet is one rack of gear? I had to look up half of these terms, and it helped. Trying to picture what 40 customers in 22,000 square feet actually looks like, because that sounds like a lot of empty floor.

@rigging yes, and the empty floor is the point. The building is sized for the power, not the square footage, so a colo running at 40% of capacity looks like a warehouse with a few islands of equipment in it.

cobalt, the number I keep going back to is the 9 cap on $378k. Was that trailing NOI net of a management fee and net of the sales cost of replacing churn? Because in my experience seller-prepared colo financials for a small facility have the owner-operator's own labor sitting at zero, and once you put a real number on running the place the cap rate you actually bought at is two or three points lower.

@meridian that's the whole thing right there. Owner-operator labor at zero is the oldest trick in small commercial and it doesn't look like a trick because there's no line item to point at.

cobalt, did the seller disclose the three migrating customers anywhere, or is that a thing you inferred later? Asking because if there was a written representation about no known intent to terminate, that's a conversation with a lawyer rather than a lesson.

The gap you fell into is the one between two different things both called data centers.

A powered shell or wholesale facility leased to a single credit tenant on a long triple net is close to a passive real estate hold. Rent is contracted, capex responsibility is allocated in the lease, and the value sits in secured power and connectivity. Retail colocation is an operating business with real estate underneath it. Revenue comes from short contracts, mostly one to three years, that reprice and churn, and margin depends on sales, support, and mechanical uptime. A cap rate applied to retail colo NOI is pricing a business, so the multiple has to carry the operating risk. Buyers who do this for a living underwrite it that way.

On your usage point, billable kW under contract with weighted remaining term is the right denominator, and it's worth going further and separating contracted kW from actual drawn kW, because customers frequently contract more than they pull and the delta shows up when you try to sell the remaining capacity against a feeder limit.

The feeder answer is the general lesson. Power delivery timelines and grid capacity are the binding supply constraint across the sector right now, which is why access to electricity drives site selection and why existing energized facilities carry a premium. That premium only exists where the site can grow. A facility capped at 1.5 MW by its service, in a market where the paying demand is wholesale blocks many times that size, doesn't get the benefit of the constraint. It sits on the wrong side of it.

Utility interconnection and upgrade cost allocation practice varies by territory and by state, so confirm capacity and upgrade timelines in writing with the serving utility for the specific address before pricing any expansion into a model.

The $340k battery quote sounds like a straight like-for-like swap of the strings. Did anyone price moving to lithium while the room was open? Longer life, less floor space, higher upfront. I'm not saying it saves the deal, the feeder still kills you, but on a hold you intended to keep it changes the ten year capex picture and it's the kind of thing that never gets modeled because the mechanical report only tells you what's there now.

Adding this one to my list of things I'd have walked into. The usage framing especially. If I'd seen 640 of 1500 I'd have thought "upside" and not "the previous owner already tried."

@bramble that's exactly the reflex. Vacancy in an apartment building is upside because the demand is there and the unit is the product. Vacant kW in a small colo is a symptom, and the question is always why nobody sold it in the ten years before you showed up.

cobalt, out of interest, what did the regional operator do with it? If they consolidated the customers into a bigger facility and dumped the building, then you sold at the price of the customer list and the real estate was worth roughly nothing, which is a grim way to find out what you actually owned.

The sales function point is the one I'd underline for anyone reading this and thinking about a small operating asset in any sector. Recurring revenue with 19% churn needs someone whose whole job is replacing it. You can't add that to an existing team on the side and expect it to happen, because it's the work that has no deadline attached and it loses every week to the work that does.

Month to month contracts at 2016 rates on 180 kW. Did you try raising them? I ask because in management we do this constantly with legacy tenants and the answer is usually that half accept and half leave, and you find out fast which half you had. On 180 kW at old pricing the downside of testing it looks smaller than carrying it.

I'm still trying to understand how private capital actually gets deployed into this sector, and the size question is what stuck with me. 1.5 MW is too small for the buyers with money. So the entry points for smaller capital are basically funds, REITs, or the land and lending side, and the direct ownership of a small physical facility looks like the worst of the three because you get the operating risk without the scale that makes the operating risk worth carrying. Is that a fair read, cobalt, or is there a size where a single facility works for an individual owner?

Let me put the numbers where I can see them.

$4.2m basis. Trailing NOI $378k, so 9.0% going in as presented. Strip out an owner-operator salary at say $110k and a real repairs and maintenance reserve at say $60k on that plant and you're at $208k, which is a 4.95% cap on your purchase price. Then apply 19% churn to a revenue base and it degrades from there without a sales spend, which is another $80k to $120k a year if you're doing it properly.

So the deal was underwater at signing on any honest expense load, and the feeder was the thing that removed the option to grow out of it. cobalt, I don't think the diligence miss was the customer list. It was accepting a seller's NOI with a zero in the labor line, and that's a mistake that has nothing to do with data centers. It shows up in car washes and self storage and small marinas the same way.

The $1.2m is the tuition. The transferable part is that a cap rate quoted on an operating business is a sentence, not a valuation.