Can a data centre become a cold store?
This is the first in a two-part series examining what happens to a data centre when its original lease ends and what that means for investors assessing residual value. In this first article, we explore whether a data centre can realistically be repurposed as a cold store, before turning next week to what the asset may be worth to another data centre occupier.
By Sola J., Investment Manager
On paper, a data centre and a cold store look like cousins. Both are windowless boxes defined by abundant power and industrial-scale cooling. So as data centres rise across Africa, cold storage keeps surfacing as the comforting answer to an uncomfortable question. When the data centre tenant leaves, what else is this asset good for? The short answer for practical purposes, is not much. The more useful question is what the failed conversion reveals: the salvage value of a data centre.
Many of the largest facilities are built to suit a single hyperscaler, which leases the entire building on a long-term contract while the developer sells the completed asset. This structure allows both the hyperscaler and the developer to remain capital-light. The long-term risk passes to the financial investor that ultimately owns the asset. If the lease expires, the tenant does not renew or another data centre operator cannot be secured, the investor is left with a fundamental question: what else can the site realistically be used for?
The industry recognises this risk. The prospectus for one recently filed data centre REIT cautions that some properties may be designed around a specific tenant’s operations and may therefore be difficult to sell or re-let to another party. Put simply, a single-tenant data centre is a special-purpose asset and special-purpose assets carry significant re-leasing risk.
The exposure examined here is the single-tenant facility: one tenant, one lease maturity, a renewal option held by the tenant alone and a design built around that tenant's hardware, density and cooling.
African conditions sharpen the cold store intuition. The Food and Agriculture Organization estimates that a substantial share of sub-Saharan Africa's perishable food is lost between harvest and market, partly for want of cold chain capacity. And grids in many markets are unreliable, so a site that has already solved its power problem is unusually valuable to a business whose refrigeration cannot tolerate outages. Does the conversion hold up, physically and commercially?
Start with what a data centre is: not primarily a building. Industry estimates put 60% to 70% of a modern facility's capital cost in its electrical, mechanical and cooling systems and African facilities add the cost of energy self-sufficiency: generator banks, dedicated solar, battery storage. The value sits in energy and connectivity, not walls. Assume a conservative scenario: a US dollar acquisition underpinned by a triple-net dollar lease, a tenant that declines to renew and no comparable compute tenant available to take its place. The facility is ultimately converted into a cold store and sold to an owner-operator. The question is not whether a sale can happen, but how much value is preserved in the process.
Physically, the two share little beyond four walls. A modern cold store needs 12-to-15-metre clear heights, extensive loading bays and deep truck yards; a data hall is filled with mechanical and electrical equipment, has minimal loading access and sits behind a hardened security perimeter. A cold store requires an insulated envelope, a vapour barrier and freeze-protected floor slabs, none of which a data centre offers. The cooling runs in opposite directions: a data centre removes server heat to hold a hall at room temperature, while a cold store's ammonia or CO2 systems drive an insulated space to around -25°C. Data centres follow power, fibre and security; cold stores follow food, ports and distribution corridors. Conversion would therefore require stripping the facility back to its structural frame and rebuilding much of it, leaving a shell that may still be the wrong height, configuration or even in the wrong location.
One piece of the intuition does survive and it is uniquely African: much of a facility's power is captive rather than grid-drawn and those generators, solar arrays and batteries are modular, reusable and answer the cold store's power problem directly. A meaningful advantage.
Commercially, a cold-chain owner-operator will not pay more for a converted data centre than the cost of building an equivalent cold store on industrial land. Replacement cost sets the ceiling, less the retrofit the shell imposes; what the facility originally cost is irrelevant. Even the power premium is smaller than it appears. A site may carry some 20MW with layered redundancy where a cold store of the same footprint needs 3MW to 5MW and the buyer pays for the power it needs, not the power installed. The surplus could in principle be wheeled and sold, but that makes a cold-chain operator a part-time power producer, a different business with its own regulatory risks and a buyer will discount accordingly.
Converted into cold storage, the same footprint would generate substantially less annual EBITDA than a stabilised data centre. Applying typical cold-chain valuation multiples reduces the asset’s value even further, potentially below the cost of developing an equivalent cold store. For an investor that originally acquired the data centre in US dollars, currency movements could deepen the loss.
So yes, a data centre can become a cold store, but only by destroying most of what was built and only at a price that recovers a fraction of what was invested. Could the reverse hold? History suggests that is the only direction the journey runs. Data centres have risen inside former printing plants, factories and, lately, crypto mining sites: buildings acquired cheaply whose power, structure or location could serve a higher-earning use. Conversions move up the income ladder, not down.
The exercise still matters: it puts an approximate floor under terminal value. An asset yielding 6% to 7% over a 15-to-20-year lease roughly returns its capital, so much of the investment case rests on residual value. Investors are in effect underwriting the proposition that the next occupier is another compute tenant. Even that base case carries a cost: retrofitting the facility for a newer generation of equipment can be a meaningful share of the cost of building new. In structurally undersupplied African markets that may be reasonable, but it remains an assumption, not a certainty.
If alternative uses only show the minimum value of a data centre, then its real value at the end of a lease depends on what another data centre operator would pay for it, minus the cost of upgrading it for their needs. The next article will look at how high those upgrade costs can be, why fast-changing technology is increasing them and how more flexible design can help reduce them.