Tian Tang Meng Xi Feasibility of Building a Compute Centre in Mauritius
Building an Africa-facing compute centre (AI Data Center / HPC Center) in Mauritius is a proposition that holds up strategically but requires careful assessment at the engineering and commercial levels. The conclusion can be stated up front: positioned as a control node, a financial-grade intelligent computing centre or a light inference node, feasibility is fairly good; positioned as a large-scale general-purpose foundation-model training base, it calls for considerable caution.
1. Four fundamental advantages
1. Submarine cables and network position. Mauritius is one of the telecommunications nodes in the Indian Ocean, connecting international submarine cables such as SAFE, LION/LION2, METISS and the T3 route to South Africa, with direct links to East Africa, Southern Africa, India and the Middle East. For businesses that need to route data between Africa, the Middle East and Asia and to run edge inference, it offers a reasonable level of bandwidth and network topology.
2. A foundation in data and financial compliance. The Data Protection Act is closely aligned with the GDPR, the legal system is a mix of common law and civil law, and capital can move in and out freely. For companies that handle cross-border data and value the security of assets and property rights, the compliance foundation here is relatively clear.
3. Industry incentive policy. The EDB is advancing the development of AI and digital special economic zones (SEZs) such as Côte d'Or, and provides pre-tax deductions for capital expenditure and R&D costs on investment in high-technology fields, along with import tax relief on related equipment (subject to the latest official rules).
4. Natural cooling conditions. A considerable share of electricity costs goes to cooling. Mauritius is promoting seawater air-conditioning (SWAC) technology, using cold deep seawater from coastal waters to cool coastal data centres, which clearly helps improve the energy efficiency of high-density racks.
2. Three real constraints
1. Scale of power supply. The island's total installed generating capacity is on the order of 800MW–1000MW (subject to the latest official data), whereas the continuous power demand of a large AI compute cluster often reaches 100MW–500MW. To deploy a large-scale training cluster, power capacity and the green-energy mix would require a dedicated feasibility assessment with the power authorities. Local electricity tariffs offer no clear advantage over the Middle East, North Africa or South Africa; although photovoltaic and energy storage projects are being advanced, the stability of baseload power still needs to be demonstrated separately.
2. Land and water resources. Mauritius has a land area of about 2,040 square kilometres; land resources are limited and are prioritised for tourism and high-end real estate, so obtaining industrial land for a large data room cluster requires case-by-case assessment.
3. Network latency to West and North Africa. Mauritius is physically distant from West Africa (such as Nigeria and Ghana) and North Africa (such as Egypt and Algeria), and the submarine cable routes are long, so latency may exceed 100ms. Scenarios such as real-time inference, cloud gaming and real-time interaction require dedicated assessment.
3. Comparison of major African locations
| Region | Advantages | Points to assess | Suitable compute type |
|---|---|---|---|
| South Africa (Johannesburg / Cape Town) | High concentration of African digital infrastructure and submarine cables; large data centres have industrial-grade backup power | Public security factors need to be built into the operating plan | Integrated primary compute node; training and inference |
| Egypt (along the Suez Canal) | A submarine cable hub for Asia, Europe and Africa; ample power supply; reaches the Middle East and Europe | Exchange rate and compliance environment require case-by-case assessment | Intercontinental compute relay; large-scale data centres |
| Kenya (Nairobi) | A high share of green power, mainly geothermal; East Africa's digital hub | Infrastructure scale is relatively limited | Regional edge compute and application nodes for East Africa |
| Mauritius | Stable political and legal environment, financial freedom, a good data-compliance foundation and deep-seawater cooling conditions | Smaller grid capacity, higher land costs and greater distance from West Africa | Financial-grade intelligent computing centre, high-value data hosting, light inference |
4. Two more viable models
Model A: Offshore compute routing and a financial AI cloud centre. Rather than deploying a large-scale cluster locally, build a small high-performance node (for example, on the scale of a few hundred cards) to provide risk-control AI, data processing and light inference services to Chinese enterprises and financial institutions in Africa and the Middle East, with Mauritius as the data control and compliance centre.
Model B: A "front office, back factory" distributed architecture. Base the compute operations headquarters, data compliance centre and overseas settlement centre in Mauritius to benefit from the local tax regime and free convertibility of funds; deploy the heavy GPU racks with heavy power needs in regions with lower electricity costs, and schedule tasks over the network.
In one sentence: "put the compute equipment where power is abundant, and put compliance control and fund settlement in Mauritius" — this is currently the more realistic combination.
For enquiries: WeChat maoqiutiantangmeng · WhatsApp +230 52535525 · Email tiantangmengxi_mu@163.com · Address: Royal Rd, Goodlands, Mauritius