US energy firm, Convalt Energy has signed a non-binding MoU with Lesotho’s Ministry of Energy to develop a 1.2GW hydropower facility and an integrated AI data center, the Kobong Hydropower and AI Data Center Project, at an estimated cost of $5.92 billion. If built, it would generate more than ten times Lesotho’s current installed generation capacity of roughly 72MW. Mandatory feasibility studies are yet to begin. Construction, if it happens, is targeted for 2029.
Convalt has never completed its announced 500MW data center projects in the US, but wants to build a 1.2GW data center in Lesotho.
The announcement fits a pattern that Africa’s infrastructure community knows well. For over a decade, the continent has been a reliable producer of outsized data center announcements, headline figures in the billions, MoUs signed at investment summits, construction dates set comfortably beyond the news cycle that never progress to ground-breaking. Pan-African research firm Xalam Analytics, which is the most cited independent tracker of African data center markets, described the post-2022 investment environment in its 2024 State of Africa Data Center report as a market “sobered up by the practical realities of building critical digital infrastructure in tough operating environments,” where “early results falling short of expectations” had tested investor patience and “sowed seeds of doubt” in the African data center case. Its 2026 edition was more direct: the market had moved on from “build-it-and-they-will-come enthusiasm” to confronting demand that has not kept pace with announced supply.
Between 2022 and 2023, new facility completions declined nearly 40%, in a market simultaneously flooded with new project announcements. The disconnect between what gets announced and what gets built is most acute outside South Africa, Kenya, Nigeria, and Egypt which are the four markets that account for the majority of Africa’s operational capacity. Africa hosts 229 operational data centers with a combined capacity of roughly 300MW, accounting for less than 1% of global capacity, against a continent that represents 18% of the world’s population. The gap is not for lack of announcements. Upcoming announced capacity across the continent now exceeds 800MW on full build, more than twice the current installed base. The question, consistently, is how much of that pipeline is real.
The structural explanations for this are consistent. A non-binding MoU carries no capital commitment and no penalty for non-delivery. The feasibility study that typically follows is a holding pattern, not a development milestone. Beyond the paperwork, power is the deciding variable. Three recent examples illustrate how comprehensively that variable defeats projects that looked credible on paper.
Will history repeat itself?
In May 2024, Microsoft and UAE-based G42 unveiled plans for a US$1 billion geothermal-powered AI data center campus in Kenya’s Olkaria region, leveraging a power source that supplies roughly 40% of the country’s electricity. The project was billed as one of Africa’s most ambitious AI infrastructure investments.
However, by August 2025, officials had confirmed the project would miss its original May 2026 completion target. In May 2026, President William Ruto suspended the initiative, citing power constraints. The proposed first phase alone required around 100MW, with plans to scale to 1GW – a significant draw on Kenya’s roughly 3GW installed generation capacity. Industry reports also suggested discussions around demand and payment guarantees may have contributed to the delay.
Morocco presents the same dynamic at a macro level. Private operators have announced projects that could bring Morocco’s total installed data center capacity close to 2 gigawatts, a figure that would exceed Africa’s entire current data center footprint across its five leading markets. Texas-based Iozera announced a $500 million, 386MW facility in Tetouan in 2024. South Korea’s Naver unveiled a 500MW project at the same site in 2025, partnering with Nvidia. Morocco’s domestic demand for cloud services cannot absorb anything approaching that capacity. Both projects are premised on exporting compute to European and Middle Eastern clients. Neither has broken ground. Grid readiness and renewable stability may not reach required levels until 2035, analysts note, while water scarcity adds a second constraint for cooling-intensive AI workloads. We were at GITEX Marrakech earlier this year, and there was no mention of both projects.
However, the Kobong project has two features that distinguish it from a standard conference-room MoU. Lesotho’s eastern highlands are among Africa’s most viable hydropower geographies; the existing Muela station has operated reliably for decades. And if the hydropower gets built at anything close to 1.2GW, the economics of exporting surplus power to South Africa, a market with chronic supply constraints and neighboring Lesotho on three sides are not implausible. African Business noted in January 2026 that Kenya’s Microsoft-G42 project’s fate remained an open question 18 months after announcement, with little visible progress and Kenya had a functioning geothermal sector, an identified site, and two of the world’s most capable infrastructure companies. Lesotho has a signed MoU and a first-time developer conducting feasibility studies.
The Kobong project may eventually be built. But the path from a non-binding MoU to a commissioned data center in a landlocked frontier market runs through financial close, environmental approvals, power purchase agreements, transmission infrastructure, anchor tenant commitments, construction finance, and often four years or more of execution in a country whose entire existing generation capacity could fit inside a single hyperscale campus.
Africa undoubtedly needs more data centers. Demand for cloud services, AI workloads, digital public infrastructure, enterprise digitization, and local content hosting continues to grow faster than available capacity. What the continent lacks is not demand. It is execution.
The challenge is no longer convincing investors that Africa needs digital infrastructure. The challenge is converting announcements into operational facilities, megawatts into revenue-generating capacity, and ambition into deployed infrastructure. Africa does not need more press releases about billion-dollar projects. It needs more projects that reach financial close, break ground, switch on, and serve customers. In the AI era, success will be measured not by announced megawatts, but by commissioned megawatts.
