Greek energy group AMACO wants to build a $1.5 billion artificial intelligence data centre in Mombasa. The plan comes with an unusual twist. It brings its own power supply.
AMACO’s chief executive, Theodore Theodoropoulos, has been in Nairobi this week. He is discussing the proposal with Kenyan officials. According to Business Daily Africa, the project would pair a large data centre with an independent power system. That system would not rely on Kenya’s national grid. AMACO calls it HERCULES. It processes natural gas and merges electricity generation, cooling, and data centre infrastructure into one platform.
The idea tackles a question many African AI plans have skipped. Where will the electricity come from? Global data centre electricity use stood at around 415 terawatt-hours in 2024. It could reach 945 terawatt-hours by 2030. AI is the main driver of that growth. This puts pressure on power grids everywhere. African grids often carry less spare capacity than grids in wealthier regions.
So what can the rest of Africa learn from AMACO’s approach? A few lessons stand out, even before anyone knows whether Mombasa gets built.
First, build the power alongside the data centre, not after it. A standard data centre might need 10 to 25 megawatts. An AI focused hyperscale facility can need 100 megawatts or more. That is enough to power roughly 100,000 households. The International Energy Agency has warned that grid limits could delay about 20 percent of planned data centre projects worldwide. The fix is simple. Plan power and computing together from day one. Otherwise, governments chasing big data centre investments risk signing deals they cannot power.
SEE ALSO:Why 2026 Could Trigger Africa’s Clean Energy Boom
Second, do not assume the grid can absorb everything. Kenya’s electricity demand hit a record peak of 2,439 megawatts in December 2025. That was up 151 megawatts from the year before. Thermal generation rose 24 percent in the first half of the current financial year. Imports from Ethiopia and Uganda climbed almost 25 percent too. Now add a single 100 megawatt AI facility to those numbers. That one facility would add demand equal to roughly 4 percent of Kenya’s national peak.
To be fair, Kenya is not short on options. The country has strong geothermal resources and a mostly renewable power mix. Still, treating a hyperscale facility like an ordinary commercial customer looks risky. Dedicated generation, as AMACO proposes, is one solution. Requiring large developers to fund extra generation or transmission is another.
Third, ask about cooling early. As AI chips pack more power into each server rack, they generate more heat. AMACO says HERCULES recovers heat and cold within its power system. The company also claims its Mombasa plan would avoid straining local water supplies. That claim matters in Mombasa specifically, since the coastal city already struggles with water shortages and leans on supplies from outside the county.
Data centres vary widely in water use. This depends on their cooling technology. Some closed loop systems use very little water. Others consume a lot. So the smart move is asking hard questions before construction starts, not after. How many litres per kilowatt hour will the facility use? Where will that water come from? What happens during a shortage?
Fourth, location carries its own lesson. Most of Kenya’s data centre capacity sits around Nairobi. That makes sense, since the capital holds the banks, telecom firms, and tech companies that buy most of the capacity. Mombasa offers something different, though. Several submarine cables connecting Kenya to the wider internet land there. The port also suits AMACO’s gas supply plans.
This points to a wider shift. African governments often think about tech hubs in terms of where startups already cluster. But large AI infrastructure shifts that logic toward energy geography instead. Kenya has geothermal power. Ethiopia has hydro. Nigeria and Mozambique have gas. Morocco has invested heavily in solar and wind. In other words, the next generation of data centre clusters does not have to follow the existing startup map.
Finally, do not assume gas is the model to copy. HERCULES starts with liquefied natural gas, which is a fossil fuel. There is a commercial logic here. Gas can run continuously, and data centres need steady power. In fact, the IEA estimates gas already supplies about 26 percent of data centre electricity worldwide.
But Kenya’s situation looks different. The country already has close to 1 gigawatt of geothermal capacity that also runs continuously. Most of its power already comes from renewable sources. That makes the case for imported gas weaker there than it might be elsewhere. Meanwhile, big technology companies face pressure to cut their data centres’ carbon footprint even as AI drives up their power use. So a facility’s emissions profile could shape its appeal to future customers.
AMACO says HERCULES can later shift to renewables and hydrogen. That flexibility may end up mattering more than the gas it starts with.
None of this confirms whether AMACO will actually build in Mombasa. Financing, the final size of the facility, customer commitments, water use, and Kenya’s available generation capacity all remain unresolved. Those details deserve more scrutiny than the headline $1.5 billion figure. Even so, the questions AMACO is raising, about power, water, location, and fuel, are ones the rest of Africa will need to answer soon.
