Africa’s AI infrastructure challenge is not simply a shortage of capital. Investors need credible customers, predictable demand and contracts showing how computing capacity will be used and paid for. Until fragmented interest becomes measurable workloads and firm commitments, many projects will remain difficult to finance.

Africa’s startups, banks, telecommunications companies, universities and public institutions are experimenting with artificial intelligence. Each may require computing capacity, but their individual requirements are often too small, temporary or uncertain to support a major infrastructure investment.

A fintech may need specialised processors to test a fraud-detection model. A university may require capacity for research during part of the year. A bank may have a larger workload but remain unwilling to sign a long-term contract while its AI strategy is still developing.

Viewed separately, these requirements appear fragmented and difficult to finance. Combined, they could support a viable computing platform.

That was one of the strongest conclusions from the “From AI Consumer to AI Producer: Building Africa’s Intelligence Infrastructure” panel at Hyperscalers Convergence Africa 2026.

Moderated by Prof. Nkem Ihenachor, Professor of Strategic Management, Lagos Business School, the session featured Gary Chomse, Sales Director, Vertiv; Wole Abu, Managing Director, West Africa, Equinix; Kanwulia Okafor, Director, Industry Services, GSMA; Lars Johannisson, Chief Executive Officer, Rack Centre; and Guy Zibi, Chief Executive Officer, Xalam Analytics.

The panel’s message was straightforward: interest in AI is growing, but interest alone cannot support infrastructure. It must be converted into demand that investors can measure and finance.

Consumption is not yet effective demand

African businesses and consumers already use AI, frequently through foreign-hosted applications and global cloud services. Local developers are also creating products for financial services, agriculture, health, education and customer support.

This activity demonstrates interest, but it does not necessarily tell a data centre operator how much capacity to build.

“There is some consumption of AI applications and tools, and several use cases are being developed and tested,” Guy Zibi told the panel. “But we do not yet see the scale of consumption that translates into effective demand.”

Effective demand means more than potential users. It requires customers that can define their computing requirements, agree on a price and commit to purchasing capacity for a sufficiently predictable period.

Infrastructure investors must estimate how many processors will be used, when the demand will begin, how quickly it will grow and whether customers can meet their contractual obligations.

Without that information, a proposed AI facility remains speculative.

Capital follows bankability

There is capital available for African infrastructure through commercial banks, development finance institutions, infrastructure funds, sovereign investors and private-equity firms. But these institutions are not investing in technology announcements. They are financing projects expected to generate sufficient and predictable revenue.

“How can capital be the constraint when we have sovereign wealth funds and multilateral finance institutions willing to take the exposure?” Wole Abu asked. “What is missing is bankability – something an investment committee can sign off.”

A bankable AI infrastructure project requires a credible business model. Investors need to understand the construction and operating costs, power arrangements, expected utilisation, customer mix, pricing, foreign exchange exposure and likely returns.

They also need evidence of demand.

This evidence may take the form of anchor-customer agreements, advance capacity reservations, minimum-spend commitments or long-term contracts. Without them, a facility can be built before the market is ready, leaving expensive processors and data-centre capacity underused.

The World Bank and International Finance Corporation identify market size and potential demand as central considerations in private investment in cloud and data infrastructure. They also note that governments can stimulate markets by digitalising public services and migrating them to the cloud, signalling confidence and encouraging private-sector adoption. Yes, Africa needs computing infrastructure.  But can it develop a commercially credible market?

One fintech is a workload; 50 fintechs are a project

Many African organisations or countries (as Kenya recently recently demonstrates) cannot independently commit to the capacity levels expected by major infrastructure providers. Their requirements may also fluctuate as products are developed, tested and launched.

Aggregation provides a solution.

“One fintech workload is too small; 50 fintechs is one project,” Abu said.

A shared platform could combine requirements from fintechs, banks, insurance companies, mobile operators, healthcare providers, universities and government agencies. Their workloads would not have to be identical. They would need to create sufficient collective demand to support the initial investment and maintain reasonable utilisation.

Industry associations could survey their members and quantify their current and expected requirements. Governments could aggregate demand across ministries and public agencies. Telecommunications companies could combine connectivity, cloud and computing services for business customers.

Universities and innovation hubs could pool research requirements, while development partners could underwrite access for early-stage companies working on socially valuable applications.

The outcome should be a demand pipeline showing:

• The organisations prepared to participate
• The applications they intend to run
• The computing capacity required
• When that capacity will be needed
• The duration of their commitments
• The security and data residency requirements involved

This information would allow infrastructure providers to design facilities around actual requirements rather than assumed future demand.

Anchor customers can make the market

Large infrastructure projects often begin with anchor customers. These are credible users willing to commit to a meaningful portion of the capacity before construction or expansion is completed.

An anchor customer reduces risk. Its contract gives lenders and investors confidence that the project will generate baseline revenue even before smaller customers join.

“In order to have a commercially sustainable operation, you need a predictable client base and some form of anchor customer,” Lars Johannisson said. “You build the wider retail customer base around those anchors.”

In Africa’s AI market, anchors could include governments, major banks, telecommunications companies, universities, global technology companies and large enterprises.

Government is particularly important because it generates substantial demand across health, education, taxation, identity, public safety and administration. Aggregating this demand could reduce duplicated technology spending while creating a stronger commercial foundation for local infrastructure. Nigeria’s National Digital Cloud Policy, championed by Kashifu Abdullahi, Director General, National Information Technology Development Agency (NITDA), for example, proposes using the government’s collective purchasing power to aggregate cloud demand across public institutions. The model is intended to create predictable anchor demand, improve procurement terms and support private investment in domestic infrastructure.

Similar models could be developed for AI computing capacity. However, government commitments must be credible, transparently procured and supported by budgets. A policy statement without an operational purchasing programme will not provide the assurance investors require.

The scalable purchasing model  

Demand aggregation will fail if computing capacity is sold only through contracts designed for hyperscalers and large multinational enterprises.

Many African startups and smaller businesses need flexible access. They may require a small amount of capacity today, a larger allocation during product testing and another increase when a service enters the market.

They may also be unable to make long-term commitments in dollars.

“We have built the brick and mortar, but we have virtually locked African enterprises out because the pricing and contracting models require a scale most of them do not yet have,” Abu said.

Providers will need products suited to the market. Options could include:

• Fractional access to specialised processors
• Reserved capacity for sector groups
• Shared inference platforms
• Local currency billing
• Managed AI infrastructure for organisations without specialised teams
• Secure environments for regulated industries
• Research and startup-access programmes

Infrastructure providers still need predictable revenue. Flexibility for smaller customers therefore can only sit on top of a stable commercial foundation created by anchor clients and aggregated demand.

For example, a bank or government contract could underwrite the baseline capacity of a facility, while unused capacity is sold more flexibly to startups, universities and smaller enterprises.

Regulation can make demand visible

Clear regulation can help convert potential demand into measurable demand.

Requirements concerning government cloud adoption, payment-data residency, cybersecurity or sector-specific data governance can give institutions a reason and timeline for moving workloads onto approved infrastructure.

This allows providers and investors to estimate the scale and timing of demand more confidently.

But regulation must be coordinated with available infrastructure. Rules that create demand without addressing power, connectivity, pricing and technical capacity can increase compliance costs without building a sustainable market.

Policy should also encourage competition. A bankable market should not depend on a single infrastructure provider or force customers into arrangements that create long-term lock-in.

The World Bank argues that successful cloud and data infrastructure markets require a combination of demand, reliable power, resilient broadband, skills and a clear regulatory environment. Hybrid and multicloud models also require portability, interoperability and competition between providers. Regulation is most useful when it gives investors confidence without prescribing a technology or protecting inefficient providers.

Supply, demand and rules must move together

Building computing capacity before customers are ready can create stranded infrastructure. Waiting for fully mature demand before investing can leave African businesses without the resources needed to develop their applications.

The two sides must develop together.

The World Bank describes AI readiness in emerging markets as coordinated market development involving supply, demand and rules. Supply includes data centres and connectivity; demand includes government migration and enterprise adoption; and rules include data protection, interoperability and cybersecurity. This coordination could take the form of an African AI infrastructure marketplace through which providers publish available capacity, prices and technical specifications while potential users register expected workloads.

Industry groups could then aggregate compatible requirements, negotiate shared terms and identify projects requiring additional investment.

Such a marketplace would also provide better evidence to policymakers. Instead of relying on broad estimates, governments could see where demand exists, which sectors are growing and what prevents organisations from purchasing capacity.

Measure use, not announced capacity

The success of an AI infrastructure project should not be judged solely by its installed power capacity, number of processors or construction cost.

IT load is good, but utilisation is better.

A smaller computing platform operating near capacity and supporting commercially useful services may contribute more to the economy than a much larger facility with few customers.

Useful measures would include:

• The percentage of available computing capacity being used
• The number of paying organisations
• The proportion of customers that are African companies
• Revenue generated by local AI applications
• The cost of computing access for smaller businesses
• Jobs and technical skills created
• Private investment mobilised
• Productivity gains achieved by customers

“Productivity is measured in megawatts used, not announcements made,” Abu said.

That standard would shift attention away from promises and towards the creation of functioning markets.

Build the customers alongside the infrastructure

Africa needs more computing capacity, but it also needs customers able to use and pay for it.

That requires more than financing data centres. Governments and industry must help businesses identify viable applications, prepare their data, build technical skills and move successful pilots into full commercial deployment.

Infrastructure providers must also design services around the realities of African customers rather than waiting for them to resemble global hyperscalers.

If workloads remain isolated, uncertain and unaffordable, capital will continue to hesitate. If they are aggregated into credible demand supported by anchor contracts and workable pricing, investors will respond.

The path to bankable AI infrastructure begins before any concrete is poured. It begins by identifying the customers, understanding what they need and securing their commitment to use what is built.