The Data Centre Boom Is Becoming Another African Infrastructure Story

Data centre infrastructure is emerging as another major African infrastructure story, driven by AI demand, power, water, connectivity and construction needs.

The global artificial intelligence boom is creating a new kind of infrastructure client: the hyperscale data centre. What was once a specialised digital facility is rapidly becoming a major construction and infrastructure undertaking, bringing together power generation, transmission, buildings, cooling systems, fibre networks, water, batteries and specialist engineering.

Two developments this week illustrate the scale of the transformation. Samsung announced a US$1 billion investment in Helix Digital Infrastructure, an AI infrastructure company whose activities span hyperscale data centres, power generation and transmission, and fibre infrastructure. The investment also brings several Samsung businesses, including construction, data-centre operations, cooling and battery technologies, into an increasingly integrated AI-infrastructure ecosystem. Samsung's announcement

OpenAI provides another measure of how physical the AI boom has become. Its PORTS-Pike project in Ohio involves approximately 8 GW of IT capacity and is expected to generate 35,000 construction jobs during its six-year build-out. The company says the development will require additional power generation, transmission and water infrastructure as it expands. OpenAI's PORTS-Pike announcement

The scale is also evident in Michigan, where OpenAI and its partners broke ground in June on The Barn, a 1 GW data-centre campus. The project is expected to create more than 2,500 construction jobs, while its closed-loop cooling system is designed to limit water consumption. OpenAI's Michigan project announcement

These are no longer conventional building projects. An AI data centre requires substations and grid connections, backup power, batteries and UPS systems, sophisticated cooling, fire protection, fibre connectivity, security, roads, drainage and other specialist mechanical and electrical infrastructure.

Data centres are an emerging  new kind of infrastructure client

Africa is entering the infrastructure race

The same transition is beginning to emerge in Africa.

Zimbabwean entrepreneur Strive Masiyiwa has spent years investing in African digital infrastructure, but the proposition is now moving from connectivity and conventional data hosting towards domestic AI-compute capacity. Cassava Technologies has announced an AI Factory initiative using NVIDIA accelerated computing and AI software in its South African data centres, with planned expansion to facilities in Nigeria, Kenya, Egypt and Morocco. The stated objective is to give African businesses, governments and researchers access to computing capacity for developing and deploying AI while keeping data within African borders. Cassava Technologies' AI infrastructure announcement

East Africa is also moving into larger-scale facilities. In Kenya, construction began on a 44 MW data centre at Tatu City, developed by Nxtra by Airtel Africa. The Kenyan Ministry of ICT describes the facility as a major new data-centre investment intended to support growing demand for cloud and digital services. Kenya Ministry of ICT

The significance for Africa is not simply the number of data centres being constructed. These facilities create demand for the wider infrastructure systems that support them, particularly reliable electricity, transmission capacity, fibre networks, cooling and industrial services.

That creates both an opportunity and a constraint.

AI is arriving on top of existing infrastructure pressures

Africa is entering the AI infrastructure race while many of its existing infrastructure systems remain under pressure.

Electricity generation may be expanding, but reliable supply and transmission capacity remain challenges. Transformers and other grid equipment can be difficult to secure. Water systems remain under pressure. Industrial sites do not always have adequate supporting infrastructure. Specialist engineering and construction skills are limited, while permitting and environmental requirements can add further complexity.

The issue is therefore not simply whether African countries have enough land. In countries such as Uganda, land in absolute terms is not necessarily the fundamental constraint. Ownership, acquisition, location, zoning and access to supporting infrastructure are more important questions.

A large site becomes considerably less useful to a data-centre developer if it lacks reliable electricity, fibre, water, roads or an efficient connection to the transmission network.

And AI is now adding another major industrial demand for those same systems.

Uganda offers an early glimpse of the challenge

Uganda already has an established data-centre layer serving banking, telecommunications, cloud and enterprise requirements. The more interesting development for the AI story is the proposed Aeonian Project at Karuma, which is being positioned as a high-performance computing and AI facility.

According to its developer, Synectics Technologies, the proposed facility would be a hyperscale data and high-performance computing centre drawing 100 MW of excess pre-transmission electricity from the Karuma Hydropower Plant. The project includes a 30 MW AI facility, with additional cooling using water from the nearby river and fibre connections linking it to Kampala and international connectivity through Kenya. Synectics Technologies' Aeonian Project description

The strategic attraction is obvious: locating energy-intensive computing close to a major hydropower source provides an important energy advantage.

But Karuma also illustrates why AI infrastructure cannot be separated from its physical environment.

Recent reporting by Nature Africa has raised questions about water use and discharge at the proposed AI centre, with researchers seeking more information about the project's water footprint. Synectics says its direct-to-chip cooling technology will minimise water consumption, but the wider water requirements of a major data centre remain an important consideration. Nature Africa's reporting on the Karuma AI centre

The project's eventual physical footprint and its interaction with the surrounding environment will also matter. Karuma sits in a region known for being a significant wildlife habitat, so any large industrial development would need to address site selection, environmental assessment and habitat considerations as part of the normal planning and permitting process. At this stage, however, the precise scale and footprint of the proposed AI facility are not sufficiently established to draw conclusions about the extent of any such impact.

What the Karuma case already demonstrates is simpler: even a project strategically located beside abundant hydropower still has to solve questions of water, cooling, discharge and environmental management.

The resource problem is bigger than electricity

This is likely to become one of the defining infrastructure questions around AI.

Data-centre developers are competing for many of the same physical inputs required by existing industries and communities: reliable power, transmission capacity, transformers, water, serviced industrial land, construction capacity and skilled labour.

The global experience is already demonstrating the problem. OpenAI's PORTS-Pike project incorporates additional power generation, transmission and water infrastructure into its development plans. Even where some infrastructure already exists, expansion to hyperscale capacity creates additional requirements. OpenAI's PORTS-Pike announcement

OpenAI's Michigan project similarly demonstrates the need to build infrastructure requirements into the project itself. The company says the energy and infrastructure needed for The Barn will be paid for by the project, while its closed-loop cooling system is intended to limit pressure on local water resources. OpenAI's Michigan project announcement

For Africa, the challenge is more pronounced because this new demand is being added to infrastructure systems that already have substantial unmet requirements.

But there is an opportunity within that constraint.

Large data-centre investments can justify new substations, transmission connections, generation capacity, fibre networks, industrial infrastructure, specialist cooling systems and technical skills. Infrastructure developed for AI can subsequently support manufacturing, financial services, telecommunications, research and other digital industries.

The question, therefore, is not whether Africa should build data centres.

It is whether the continent can build the physical infrastructure ecosystem required to support them without simply transferring pressure from one infrastructure deficit to another.

The AI economy may be digital, but its foundations are decidedly physical.

For Africa's construction and infrastructure industries, the data-centre boom is becoming another infrastructure story, one that will increasingly compete for the same power, water, equipment, industrial capacity and skilled labour needed by the rest of the economy.

And that makes the construction industry part of the AI story whether it intended to be or not.