On September 8, Egypt's Ministry of Communications and Information Technology stood alongside Vodafone Business, Cassava Technologies and Elsewedy Electric to announce Africa Data Centers Egypt, a joint venture built around a $1 billion sovereign AI data center programme, with roughly $200 million of foreign direct investment committed to its first phase. Days earlier, MTN Group had confirmed the scope of a new venture of its own, Africa Data Hub Holding Limited, built with Dubai based data center developer Tarek Al Ashram. The target there is 150MW of AI-ready capacity across Nigeria and South Africa in phase one, with Al Ashram's platform providing the bulk of the capital and MTN holding a minority stake.
Read the way most of the continent's tech press has read them, as proof that Africa's AI infrastructure moment has finally arrived, the two deals feel like confirmation of a boom. Two of the largest telecoms and infrastructure players on the continent are backing sovereign and regional AI compute inside the same fortnight, with headline figures running into the hundreds of millions and low billions of dollars.
They are real deals, with real capital and real government backing behind them. But hold them against the number that actually matters, the megawatts Africa has installed today measured against the megawatts McKinsey estimates it will need by 2030, and both announcements shrink considerably. The gap between the language of these deals "sovereign," "AI-ready," "regional hub" and their first phase numbers 20MW & 150MW is the real story here. Africa's AI infrastructure announcements are multiplying faster than its megawatts
Why is everybody building now ?
The urgency behind both deals is not manufactured. AI workloads, training and increasingly inference, are compute hungry and latency sensitive in ways African markets have historically routed around by leaning on capacity in Europe or the Gulf. That workaround gets harder to justify as AI moves from research curiosity into the operational core of banking, telecoms and government services on the continent. Discovery Bank and Nedbank made that shift visible in public at the CNBC Africa AI Summit this month. Over half of Discovery Bank's client interactions now run through its AI channel. Nedbank has cut its fraud case capture time from 40 minutes down to two. This is production grade AI, the exact kind of deployment that turns "nice to have local compute" into "need local compute, now, for reasons of latency, cost and data residency."
The numbers that actually matter
According to the Africa Data Centres Association's 2026 Economic Report, the continent's active data centre capacity stands at roughly 360MW, with a further 238MW under construction and 656MW sitting in the pipeline. Africa's data centres hold about 0.6% of global capacity on a continent that holds roughly 20% of the world's population, a ratio flagged independently by both the Africa Data Centres Association's own reporting and CNBC Africa's coverage of the sector this year, which put Africa's share at "less than 1%" of global capacity. For scale, the United States alone hosts something close to 45% of the world's data centres, and global active capacity sits around 55GW, with another 15GW under construction and 135GW in the pipeline elsewhere in the world.
South Africa is the standout on the continent by a wide margin. It has 55 facilities, an expected R50 billion in investment over the next three years, and it is the only market with all three global hyperscalers AWS, Google and Microsoft Azure present, alongside operators like Teraco, Vantage and Open Access Data Centres. Even there, the constraint is not appetite or capital. Multiple operators and analysts point instead to grid reliability and water availability as the binding constraint on how fast that pipeline can actually be energised.
McKinsey's November 2025 analysis, "Building data centers for Africa's unique market dynamics," puts a number on what "enough" would actually look like. Demand for data centre capacity on the continent is projected to rise from about 0.4 gigawatts today to somewhere between 1.5 and 2.2 gigawatts by 2030, a 3.5x to 5.5x increase, requiring an estimated $10 billion to $20 billion in new investment to build out. Put the two headline deals from this fortnight against that number and the arithmetic turns sobering. Egypt's programme targets 200MW at full build out. MTN's targets 150MW in its first phase. Combined, and assuming both hit their stated targets in full, that is 350MW, meaningful, but a fraction of the roughly 1,100MW to 1,800MW of additional capacity McKinsey's range implies the continent needs by 2030, and a small slice of the $10 billion to $20 billion in capital that build-out is estimated to require.
What the Egypt deal actually commits to
It is worth being precise about what was actually signed, because the framing around sovereignty and "regional hub" status is doing a lot of work in the coverage. Africa Data Centers Egypt's first phase is 20MW, not the $1 billion headline figure, and not the 200MW long run target either. The 200MW figure is where the venture is aiming to get to, not where it starts. Egypt's Communications Minister, Raafat Hindi, framed the deal in explicitly sovereignty forward language. "Digital sovereignty should not mean isolation, but rather the ability to engage with international technology companies." Vodafone Egypt's CEO, Mohamed Abdallah, described it as demonstrating commitment to "supporting Egypt's digital transformation through infrastructure investment." The venture is expected to create roughly 200 direct jobs and around 2,000 indirect ones.
None of that is a criticism of the deal on its own terms. Starting at 20MW and building toward 200MW over several years is a normal, sensible way to de-risk a capital intensive, multi-year infrastructure build. The criticism is of how it gets reported and repeated, a $1 billion, sovereign AI, regional hub headline, attached to a first phase that is one tenth of its own eventual target and a rounding error against McKinsey's 2030 demand estimate.
What the MTN deal actually commits to
The MTN structure is more interesting precisely because it is more candid about its own limits. MTN Group CEO Ralph Mupita was explicit on both the geography and the staging. "For now, only in South Africa and Nigeria...looking at 150MW as phase 1. We will build out as demand requires." That is a company stating, on the record, that it is not yet committing capital beyond two markets or beyond a first tranche of capacity, a useful contrast to the more expansive "sovereign hub" language showing up elsewhere in the sector's PR this year.
The deal's real innovation is structural, not just physical. MTN is not funding this alone. Al Ashram's platform, built on his experience at Gulf Data Hub, a Dubai based developer established in 2011, provides the majority of the capital, with MTN taking a minority stake and its connectivity subsidiary, Bayobab, contributing pan-African network reach and market access. Al Ashram summarised the rationale plainly. "We see significant potential to bring our experience in developing and operating large scale data centre platforms to African markets." A second executive involved in structuring the venture, Mroué, described the combination as bringing "investment, infrastructure expertise, connectivity and market access into a single platform," effectively an admission that no single African telecom operator currently has the balance sheet to fund AI scale data center capacity alone, and that Gulf capital and operating expertise are becoming a structural precondition for these builds, not an optional accelerant.
The bottlenecks nobody's press release mentions
Strip away the announcement language and three constraints recur across almost every serious analysis of why Africa's data centre pipeline of (656MW, per the ADCA report) consistently under delivers against its own targets.
Power. Data centres, and AI workloads especially, need dense, reliable, continuously available power, the opposite of what grid instability in markets like South Africa and Nigeria has historically supplied. Facilities can and do build around this with dedicated generation and renewables, but that adds cost and time to every megawatt, which is one reason first phase numbers (20MW, 150MW) are so much smaller than headline ambitions.
Water. AI scale compute runs hot, hot compute needs cooling, and cooling at scale needs water, in a set of markets where water stress is already a live policy issue quite independent of data centres. Reporting on South Africa's build out has been blunt about this trade off, framing the AI data centre opportunity explicitly against the water and power "catch" that comes bundled with it.
Predictability. Steven Santini, a Schneider Electric vice president focused on the sector, put the third constraint simply. "We need predictable permitting" to accelerate data centre development and faster execution across the continent. BCG Managing Director Kesh Mudaly frames the broader coordination gap the same way. "We do have a bit of a way to cover" on infrastructure, alongside weak inter-ministerial alignment on AI strategy and underdeveloped data governance rules. Mudaly's sharper warning concerns what happens if that gap is never closed, a pattern where "African data is processed overseas and then returned to local users through foreign owned applications," sovereignty in name, dependency in practice.
A framework to assess actuality from PR hype
Every AI infrastructure announcement out of Africa for the next several years is going to arrive wrapped in some version of the same language, sovereign, regional hub, AI ready, billion dollar. Three questions, asked of any such announcement, separate the ones that actually move the needle from the ones that are mostly a press release.
1. What's committed versus what's aspirational? Is the headline figure the financing that has actually closed, or the long run target the venture is "building toward"? Egypt's $1 billion covers the eventual build. Its committed first phase FDI is roughly a fifth of that.
2. What's the first phase megawatt number, not the target megawatt number? Egypt's first phase is 20MW against a 200MW target. MTN's is explicitly capped at 150MW "phase 1," with expansion contingent on demand. Both companies are, to their credit, reasonably transparent about this distinction. It is the surrounding coverage that tends to collapse target into commitment.
3. Are power and water secured, or merely assumed? A deal that has answered how it gets reliable power and water at the scale of its ultimate target, not its first phase, is a fundamentally different proposition from one that has not.