Microsoft and G42’s planned $1 billion data centre in Kenya has reportedly slowed after talks with the Kenyan government over guaranteed capacity payments broke down, putting one of East Africa’s most visible cloud infrastructure projects under pressure.
The project was announced in 2024 as part of a broader digital investment package involving Microsoft, UAE-based AI company G42, and Kenya’s Ministry of Information, Communications and the Digital Economy. It was expected to support a new East Africa cloud region, run Microsoft Azure services, and use geothermal energy as part of Kenya’s push to become a stronger cloud and AI infrastructure hub. (Microsoft)
The delay matters because it exposes a difficult truth: Africa’s AI future will not be built by strategy documents alone. It will depend on power, demand, contracts, data-centre economics, and governments that can attract private infrastructure without taking on obligations they cannot afford.
The problem is not just construction
At first glance, a delayed data centre sounds like a project execution problem. But the Kenya case points to something deeper.
Microsoft and G42 reportedly wanted the Kenyan government to guarantee annual payments for a certain amount of cloud capacity. The government could not meet the requested level, and talks stalled, according to a Bloomberg report cited by Reuters.
That detail matters. A hyperscale data centre is not useful simply because it exists. It needs demand strong enough to justify the investment. It needs customers that can pay consistently. It needs power supply that can support heavy workloads. It needs confidence that public and private sector usage will grow fast enough to fill the capacity.
This is where the economics become harder.
African countries want cloud regions, AI infrastructure, local data hosting, and digital sovereignty. Investors want predictable returns. Hyperscalers want anchor demand. Governments want development without carrying unsustainable guarantees.
Those goals do not always align neatly.
Kenya’s ambition is still serious
Kenya has good reasons to chase this project.
The original Microsoft-G42 plan was one of the largest private-sector digital investment packages announced for the country. It included a geothermal-powered data centre, cloud services, digital skills, AI development, cybersecurity cooperation, and support for local-language AI work in Swahili and English. (Microsoft)
That mix fits Kenya’s broader positioning. The country already has a strong digital finance market, a visible startup scene, improving fibre infrastructure, and a government that has tried to market Kenya as an anchor point for East African technology investment.
The data-centre project was supposed to deepen that position.
A cloud region in East Africa could reduce latency, improve access to enterprise cloud services, support local AI development, and give governments and regulated industries more options for data hosting. It could also help startups build more reliable services for regional users.
The ambition is sound. The financing and demand model is the harder part.
AI infrastructure needs anchor demand
The global AI boom has made data centres more important and more expensive.
Cloud and AI workloads require large amounts of compute, power, cooling, networking, and security. In mature markets, hyperscalers can rely on deep enterprise demand, large public-sector contracts, and strong private-sector cloud adoption. In many African markets, demand is growing but still uneven.
That creates a dilemma.
If governments guarantee too much capacity, they risk paying for cloud infrastructure that the market does not use quickly enough. If they guarantee too little, private investors may decide the project is too risky or scale it down.
This is not unique to Kenya. It is a wider African infrastructure problem.
Many countries want to host data centres. Fewer have the combination of cheap reliable power, dense enterprise demand, cloud-ready businesses, skilled operators, regulatory clarity, and customer commitments needed to make hyperscale economics work.
That is why the Kenya project is important. It shows that attracting big technology infrastructure is not only about announcing capital. It is about proving long-term utilisation.
Power remains the foundation
The power question is unavoidable.
The original Kenya plan leaned heavily on geothermal energy. G42’s EcoCloud collaboration materials described the proposed facility as strategically located in a geothermal-rich region, with an initial capacity of 100MW and potential expansion up to 1GW. (G42)
That is a strong story for sustainability and positioning. Kenya’s geothermal resources give it an advantage compared with countries that rely heavily on fossil fuels or unstable grids.
But clean energy potential is not the same as immediate power availability.
A data centre of that scale must be connected, supplied, cooled, and protected against outages. It also has to coexist with residential, industrial, and public power demand. If power capacity becomes constrained, governments face a difficult political and economic choice: support a prestige cloud project or protect wider electricity access for households and businesses.
This is why power planning must sit at the centre of Africa’s AI strategy.
No country can become an AI hub without treating electricity as digital infrastructure.
What this means for African startups
For startups, the story may seem far away. It is not.
Cloud infrastructure affects cost, latency, reliability, compliance, and product performance. A Kenyan healthtech startup, fintech platform, logistics company, AI assistant, or edtech product may not need hyperscale infrastructure on day one, but it will eventually care about where its data is hosted, how fast its services respond, how much cloud usage costs, and whether enterprise customers trust its infrastructure.
If local or regional cloud capacity grows, startups can benefit from better performance and potentially more suitable hosting options. If major projects stall, many builders will remain dependent on cloud regions outside their markets.
That dependency is not automatically bad. But it has consequences.
It can affect latency. It can affect costs. It can complicate data protection. It can make local AI development harder. It can limit the ability of governments and regulated sectors to adopt cloud services confidently.
African startups cannot separate their product ambitions from the infrastructure beneath them.
Governments need a sharper playbook
The Kenya case also offers a policy lesson.
Governments should not treat data-centre deals as headline investments only. They need a sharper playbook for what they are willing to guarantee, what demand they can realistically anchor, and how infrastructure risk should be shared.
That playbook should include clear answers to several questions.
How much cloud demand can the government commit to without crowding out other priorities?
Which public services are ready to move to cloud infrastructure?
What power capacity is available now, not only promised later?
How will local firms access the new infrastructure?
What safeguards exist around data protection, cybersecurity, and vendor dependency?
How will the project support local skills rather than only import infrastructure?
These questions are not anti-investment. They are what make investment durable.
A country that cannot answer them may still announce large projects, but those projects can struggle when negotiations move from press releases to contracts.
The regional stakes are bigger than Kenya
East Africa needs more cloud capacity. That part is clear.
As AI adoption rises, cloud demand will grow across fintech, telecom, government services, healthcare, agriculture, education, logistics, and enterprise software. The region also needs better infrastructure to support local-language models, digital identity systems, cybersecurity, and public-sector platforms.
Kenya remains one of the strongest candidates to host part of that infrastructure because of its digital market depth and renewable energy potential. But the Microsoft-G42 delay shows that the path will not be automatic.
Other African countries should pay attention.
A data-centre race is beginning across the continent. The winners will not simply be the countries that offer the loudest incentives. They will be the ones that can combine reliable power, credible demand, clear regulation, strong connectivity, and realistic public-private financing.
The implication for African tech
Microsoft and G42’s Kenya project may still move forward in some form. The reported delay does not mean the ambition is dead. It does mean the economics need to work.
That is the lesson.
Africa’s AI and cloud future will not be decided only by who signs the biggest memorandum of understanding. It will be decided by who can keep the lights on, fill capacity, protect data, train operators, and make the business case work for both governments and investors.
For founders, the implication is clear: infrastructure risk is product risk. For policymakers, it is even clearer: AI ambition without infrastructure discipline will keep running into hard limits.
Kenya still has a strong case to become a regional cloud hub. But this delay shows that the next phase of African digital infrastructure will be measured less by announcements and more by execution.
