$12M A Megawatt: The Number Africa Wants Cut In Half

The session was titled “Africa’s Moment to Deliver Digital Infrastructure at Scale.” What the panel actually delivered, over an hour at the Radisson Blu, was a costed list of reasons that moment isn’t yet being converted. There was also a striking degree of agreement on where the arithmetic fails.

The panel brought together Wole Abu, Managing Director for West Africa at Equinix; David Bunei, Managing Director for Kenya at Oracle; Mugo Kibati, Chief Executive of Telkom Kenya; Chris Wood, Group Chief Executive of WIOCC; Andile Ngcaba, Chairman of Convergence Partners; and Adil El Youssefi, Chief Executive of Africa Data Centres. Between them, they build, finance, connect and sell the layer everything else in the digital economy sits on.

The Number

Ngcaba supplied the session’s sharpest framing, and it was a price. Building data centre capacity on the continent runs at around $12 million per megawatt, he said.

For African cities to become genuinely attractive, that figure needs to fall to roughly $5 million per megawatt. That is the point, he argued, at which the economics start working for neocloud operators, the specialist AI compute providers now taking a growing share of global capacity deployment.

The $12 million reference point is not an African penalty. It sits squarely within global benchmarks. JLL puts the average worldwide data centre construction cost at about $11.3 million per megawatt for 2026, with standard facilities generally quoted between $8 million and $12 million.

That is precisely what makes Ngcaba’s argument pointed. If African builds cost roughly what builds cost everywhere, Africa is competing for the same capital as markets offering better power prices, faster permitting and explicit incentives. And it is losing on the margins.

He named the comparison directly. India has put incentives in place to attract AI data centres, and Africa is competing with those countries on two levers: tax incentives and the cost of equipment, including GPU importation, where duty and clearance treatment can add materially to the cost of the most expensive item in an AI facility.

Power completes the equation. Ngcaba put the threshold at around 10 US cents per kilowatt-hour; above that, a location struggles to attract compute-intensive workloads. It’s a stark test, and a useful one. It converts a diffuse conversation about “energy challenges” into a single number an investment committee can check.

El Youssefi, whose company operates one of the continent’s largest data centre platforms, made the same point from the operator’s seat: to compete, the price issues around building data centres must be addressed. And then came the corollary that ran through the entire session: the industry needs to see workloads.

Where the Demand Is Coming From

On that question, the panel was candid about something rarely stated so plainly. Much of the demand now materialising is not organic market pull but policy.

Abu described compelling demand in Nigeria driven by the Central Bank of Nigeria’s data localisation requirements, alongside NITDA’s localisation rules. Together, these are pushing regulated data onto domestic infrastructure and generating both capacity requirements and migration projects.

Bunei made the parallel case for Kenya, pointing to the consolidation of public sector data as a demand driver, alongside Nigeria’s policies bringing data onshore.

Bunei’s framing of sovereignty was notably pragmatic. The sovereignty question, he argued, will bring local workloads. In the shorter term, that means existing local data centres can be utilised rather than sitting under-contracted.

But he was equally clear about the condition attached. Regulations need to provide confidence, and consistency and stability are what allow an investor to be certain the investment can be recovered. Anchor clients providing initial workloads matter, because they underwrite the first tranche of capacity.

He also sketched the demand curve beyond compliance: local languages, startups and localised industries generating requirements that eventually drive volume, which in turn drives down unit cost.

Abu added the network dimension, the piece most often missed in data centre conversations. When traffic is peered locally, he said, more of it gets aggregated, and that aggregation builds more traffic still. Local interconnection isn’t merely a latency improvement. It’s a compounding mechanism that makes each additional participant more valuable to the next.

The Counterweight

Kibati supplied the session’s necessary corrective, in two numbers. Around 90 percent of the continent is covered by mobile networks, he noted, but only about 28 percent of people use the internet.

That is the usage gap, and Kibati located its cause in affordability: of smartphones, of access, of data. Most people on the continent, he pointed out, are not at the economic level of consumers in other markets. The appetite for technology exists; the usage remains low.

Set against the localisation-driven demand the other panellists described, the tension is clear. Regulation can compel banks and government agencies to place workloads locally. It cannot make seventy percent of a continent’s covered population into paying digital consumers. One demand stream is being manufactured by policy; the other is constrained by household income. Only the second scales indefinitely.

The Money Problem

Wood, whose company builds subsea and terrestrial networks, was blunt about capital. Investment at the scale currently being celebrated, a $300 million programme, for instance, is a drop in the ocean measured against the entire continent’s requirements.

He pointed to genuine demand in Nigeria and the Democratic Republic of Congo, and to a need that runs beyond subsea capacity into metro and terrestrial networks, the domestic distribution layer that determines whether international bandwidth reaches anyone.

What investors want, he said, is unglamorous and consistent: a return, reliable power, and political and currency stability. His honest summary: it is never easy to raise money for investment in Africa.

Abu closed the loop with an investor’s checklist for any market seeking digital infrastructure capital: ecosystem density, power, exit gateways, capital repatriation, enforceable contracts, standard and predictable government policy and regulation, speed to market, and standardised templates for power. It’s a list of the things that make a jurisdiction legible to a capital allocator. The point of listing them is that most are policy choices rather than natural endowments.

What the Panel Actually Asked For

Read together, the six positions form a coherent proposition. Bring the build cost down toward $5 million per megawatt through tax and equipment treatment, including GPU importation. Get power below roughly 10 cents per kilowatt-hour and make its procurement standard rather than bespoke. Keep regulation consistent enough that a fifteen-year investment can be underwritten. Use public sector and regulated workloads as anchor demand. Peer locally, so traffic compounds. And address affordability, because compliance-driven demand has a ceiling that consumer demand does not.

None of that is a request for subsidy. Most of it is a request for predictability, the same commodity Africa’s infrastructure conversation has been asking for across successive cycles.

The measurable questions from here are the ones the panel implicitly posed. Does the cost per megawatt of African builds actually fall? Do the localisation policies in Nigeria and Kenya translate into contracted capacity rather than announced capacity? And does the 28 percent usage figure move? That, more than any policy instrument, determines whether the infrastructure being built has a business model underneath it.