Africa Data Center Capacity: Read the Evidence

Africa Data Center Capacity: Read the Evidence

Africa data center capacity is often presented as a single, confident number. In practice, that number may refer to live IT load, utility power, a full-campus ambition, a grid request or simply the upper limit of a masterplan. Those are not interchangeable.

I have been reminded of this while reviewing facilities country by country across Africa for PowerlandMap. The hard part is rarely finding an announcement. The hard part is deciding what the announcement proves today. A market-entry decision built on announced megawatts can look precise while resting on very thin evidence.

My view is straightforward: coverage should expand only as quickly as the evidence allows. A blank capacity field is better than a plausible but unsupported number.

Africa Data Center Capacity Starts With Status

The first distinction is not between large and small facilities. It is between live, under construction and planned supply.

A live facility needs evidence that it is actually serving customers or has entered operation. An operator page can help, but dated operational evidence is stronger. In Zambia, for example, a Paratus case study published on 18 June 2025 says that CFAO had been hosted in its Lusaka data centre since 2021. That supports an operating-status conclusion. It does not, however, disclose usable MW IT capacity.

Construction requires a different standard. Digital Parks Africa said on 8 September 2026 that it was building NDC1 in Lagos and that the facility was undergoing a Tier IV Constructed Facility assessment. The company announcement supports an under-construction classification. Neither source gives a capacity figure that I would treat as verified IT load.

A planned campus deserves a third label. Land control, a memorandum with a municipality or a grid application can all be meaningful. They do not make a facility live. This is why the PowerlandMap coverage view separates project evidence and development stage instead of collapsing everything into one supply total.

One MW Figure Can Describe Several Different Things

The expression “100 MW data centre” may describe at least four different scopes:

• IT load available to servers;
• electrical capacity at the facility boundary;
• total campus power across several phases;
• generation, grid connection or requested capacity upstream of the site.

The commercial implications are very different. A buyer looking for 20 MW of near-term IT capacity cannot rely on a 100 MW masterplan unless the delivery phase, conversion assumptions, cooling design and grid milestones are understood.

The global energy context makes this distinction more important, not less. The IEA’s data-centre findings updated in June 2026 underline how quickly electricity demand from digital infrastructure is rising. But a national or regional power-demand trend still does not prove that a particular campus has energised capacity.

This is also why I avoid normalising every disclosure into MW IT. When a source says “site capacity” or “power,” I preserve that wording. When it says “client IT load upon completion,” I record the future scope and keep it separate from today’s inventory. The same discipline applies to AI factories, where I have previously set out the difference between headline capacity and data center IT load.

Certification Is Evidence, but Only of What Was Certified

Certification language is another common source of false precision.

A design certification shows that submitted design documents satisfy a standard. It does not, by itself, prove that a facility was built, commissioned or opened. Uptime Institute itself has explained that design goals and proven constructed achievement are not the same. Its discussion of Constructed Facility certification makes the distinction practical: the constructed assessment verifies the built facility against the design and tests its performance.

That difference matters when mapping emerging markets. A Tier III or Tier IV reference can be useful evidence of intended resilience. It should not silently become proof of current operations, spare capacity or commercial availability.

For NDC1, the public wording says that the Constructed Facility assessment is underway. I therefore treat it as a real project milestone, not as an awarded certification. That may sound conservative. In my experience, it is simply accurate.

A Country Map Is Not a Capacity Map

Africa is not one data-center market. The commercial conditions in Johannesburg, Lagos, Nairobi, Cairo, Casablanca or Lomé are too different for one regional supply number to be useful on its own.

A good country layer should answer several separate questions. Which facilities are live? Which projects are genuinely being built? Which operators disclose current capacity, and which disclose only full-build ambition? Where is the network footprint visible? What evidence exists for grid access, redundancy and expansion land?

This is the logic behind the PowerlandMap product: connect facilities to market, power and project evidence without pretending that every record has the same confidence level. It also explains why the powered-land due diligence framework starts with deliverability rather than marketing terminology.

There is a practical cost to leaving uncertainty visible. Some country pages will initially show fewer MW than market presentations suggest. I am comfortable with that. The purpose of the database is to support decisions, not to win a contest for the largest aggregate number.

How I Assess Africa Data Center Capacity Evidence

When I review a facility, I use a simple hierarchy.

First, I look for a dated primary source from the operator, authority, utility or project sponsor. Second, I separate the event from the current state: an announcement date is not an operating date. Third, I preserve the original capacity basis. Fourth, I look for corroboration when the status or number materially affects an investment conclusion. Finally, I record contradictions rather than averaging them away.

The same approach applies to carrier neutrality. An operator can describe a site as carrier-neutral, but an investor still needs to understand the actual ecosystem, access routes and concentration risk. The carrier-neutral data center due diligence framework covers that second layer.

Inferences still have a place. A large substation, an active construction site and a credible operator may together indicate a serious project. But I label that as an inference until the relevant facts are public. My opinion is that market intelligence becomes more valuable when it shows where the evidence stops.

What This Changes for Market Entry

For an investor, the immediate benefit is a cleaner denominator. Acquisition price per MW, development cost per MW and future market share are meaningless if the MW basis changes from one asset to the next.

For a developer, the benefit is sequencing. A market with limited live supply may be attractive, but only if grid access, delivery partners, customer demand and permitting can support the timing. The power-procurement analysis explains why upstream power evidence has become part of site strategy.

For an end user, the key question is not how much capacity has been announced in a country. It is what can be contracted, by when, under which technical and commercial conditions.

That is the standard I will continue to apply as PowerlandMap deepens its African coverage: evidence first, capacity basis second, aggregation last.

Africa data center capacity is growing, but the investable picture will not come from one headline number. It will come from disciplined records that keep live supply, construction, planned phases, certification and power evidence separate.

If you are assessing an African market or comparing specific facilities, you can request access to PowerlandMap and review the underlying project and market evidence.

*Matthieu Gallego*