What I will cover:
- Why the physical asset data in a telco estate is so poor, incomplete and out of date
- For the people trying to exit sites: how to consolidate and get out, rapidly and without risk
- For the people trying to monetise sites: how to sell the capacity you already own
- Where to start
What I am not talking about is the logical layer — bandwidth, throughput, the graphs of how much light is travelling down a given fibre. That is not the focus of my blog.
I am talking about the physical layer. The buildings. The exchanges, the central offices, the cell sites, the street furniture. The equipment inside them, the power feeding it, and the cooling units that extract the heat.
Telcos were the original data centre operators. Some of these sites have been running since the 1960s, long before anyone used the phrase, and they sit on property in city centres that nobody could assemble today at any price. That estate is worth more than it has been in decades — more so with the arrival of AI and the demand for edge and inference capacity close to the user. Very few operators can describe it accurately enough to act on it.
Why the data is patchy, incomplete and stale — more than anywhere else
Every physical estate drifts away from its record over time. A telco estate drifts hardest, and there are three reasons for it.
- Age: These are among the oldest estates in operation — buildings, cabling and power systems installed and altered over sixty years, by people long gone, across generations of technology. Records assembled over that span, often transferred between operators along the way, were never complete to begin with.
- Distribution: A colocation business has a handful of buildings, and someone walks each of them regularly. A telco has hundreds or thousands of sites, many with no permanent staff, where the work is done by field engineers and contractors who arrive, do the job and leave. Nothing was ever built to capture that work at the point it happens.
- You are rarely the only occupant: Your own equipment shares these buildings with other operators, wholesale customers, emergency services, broadcasters and content caches. Different owners, different maintenance regimes, different technology roadmaps — and one record underneath all of it that was unreliable from the start. So, it does not fail dramatically. It drifts, site by site, and every team downstream keeps making decisions on it anyway.
If you are trying to exit sites: consolidate, quickly and safely
One group inside every operator is looking at the estate as a cost. They are paying leases on sites that look superfluous, and the question coming down from above is simple: why are we still paying for these?
The answer is usually consolidation — migrate what matters onto fewer sites, then close, sublet or exit the rest. On paper the business case writes itself. In practice the programme stalls, in the same place every time: nobody is confident what is actually in each building, or what unplugging any of it will break. What is there, what is it connected to, who owns it, what power is it drawing?
That question has real teeth, because consolidation is never simply switching a site off. Services and tenants have to migrate to a receiving site, which may itself need more power, new fibre or structural strengthening before it can take the load. You have to prove the site you are closing is genuinely redundant and not providing quiet resilience or coverage. And the lease itself has its own calendar — break dates, notice periods, reinstatement liabilities — that rarely lines up with the date the site became technically redundant. Miss a break date because the site was not ready to clear, and you hold it for another term.
We work with a large global telco on exactly this. The engineering was never the hard part. The hard part was getting absolute visibility of what they had, where it was and how it was interconnected, so that exits could be planned rather than discovered halfway through. Decisions get made from the desk, not as the result of ten repeat engineer visits to the same building.
Monetising the power and location for AI/Inference/Edge
A different group is looking at the same estate and seeing the opposite of a cost. They have sites with power already allocated — no grid application, no waiting list, no queue — and their own network equipment now draws far less of it than it used to. That leaves headroom that did not exist when the sites were designed, in buildings that sit exactly where compute now wants to be.
That is not theory. Telefónica is converting more than a hundred of its disused copper exchanges into one-to-two-megawatt edge data centres for AI and cloud, chosen for their proximity to the customer. BT is examining which of the roughly thousand exchanges it will keep that could be repurposed for data centre use, and has separately agreed to host fourteen megawatts of sovereign AI compute across three of its sites. In the US, Verizon and others already market edge colocation from thousands of former central offices. The analyst house STL Partners expects the number of network-edge data centres worldwide to more than double, to around 1,800, by 2028.
The customers for this often want nothing more than power and floor space in the right location; they bring their own equipment and do the rest. You own the locations. You have the power. The conversation should be easy.
It stalls for one reason. The spare capacity is real, but it exists in fragments, either within a site, or across a group of sites — a little here, a little there — and none of it is aggregated into something you could confidently price and sell. If you had reliable data on the physical site, you would know exactly what you have, your options for consolidation, and what you could sell.
The same missing record blocks both
Notice that the two groups — the ones exiting sites and the ones monetising them — are blocked by exactly the same problem. Neither can act, because neither can say with confidence what is physically in a given building, who owns it, what it is connected to, and what power is free.
That is not a strategy problem. Both strategies are sound. It is the record underneath them that is missing.
Where to start
Not with a survey of the whole estate. That is a year’s worth of work and it is out of date before it finishes.
Start with one site — the one you would most like to exit, or the one you would most like to sell capacity into. Then ask two questions:
- What is in it, to the level of detail you would put in front of a buyer or a decommissioning team.
- When was that last verified by someone who physically stood in the building.
If you can answer both in an afternoon, your capture is working and this is not your problem. If you cannot — and in my experience almost nobody can — then you have found why the consolidation keeps slipping and why the monetisation conversation doesn’t get past the first meeting.
That is the gap Spire™ was built to close. It captures physical work at the moment it happens — installs, moves, changes, decommissions, connection changes — and holds the physical estate as one record across every site: the location, the equipment in it, what each thing is connected to, who owns it, and the power it draws from source to consumer. Then you scale the method rather than the survey. Once one site is genuinely accurate it becomes the template for the next ten, each usable the moment it is done, so consolidation and capacity sales can begin against the sites that carry the most cost or the most opportunity rather than waiting for the whole estate to be mapped.
The estate is already yours. The power is already allocated. The only thing missing is a record good enough to act on. And that is a fixable problem — and Spire has been solving it for two decades.
If you can already answer those two questions about your least-wanted site, I would be glad to hear it. If you cannot, tell me what you found. That is a better first conversation than a demo — for you and for me.
Martin Docherty (DCIS®) works on data centre asset management at Assetspire.