Hans van Linschoten: One grid, many tenants

Over the past three weeks, this series has made an argument in three steps.

Africa does not have to buy the AI bubble: open models are close enough to the frontier for many practical workloads, and previous-generation hardware is capable enough, that the continent can build AI capacity without betting national balance sheets on hyperscale campuses.

The right shape for that capacity is a federated grid: many independently owned locations, one platform, shared capacity, built on the same roaming logic that made African mobile work.

And that grid is the best preparation for 6G, because the next network will need exactly this kind of distributed, locally owned compute.

Arguments are cheap. This final article asks the practical question: what does it look like on an ordinary Monday morning?

Here are four organisations, in four sectors, using the same grid in different ways.

The four organisations below are illustrative examples, not named customers. They show how the grid could work in practice.

Healthcare: The clinic group that keeps its patients’ data at home

A private clinic group runs twelve facilities across central Uganda. It needs three things: an electronic patient record that works when the connection drops, a clinical decision-support tool that helps nurses in remote clinics, and an appointment system patients can use from a basic phone.

On the grid, the clinic group does not buy servers or sign with a foreign cloud. It contracts with a local health-IT provider, which operates as a Virtual Cloud Operator on capacity from ABQ Cloud, the Ugandan Cloud Enabler.

Patient records live in a Ugandan location, supporting the clinic group’s data residency requirements and compliance with the country’s Data Protection and Privacy Act.

A nightly backup goes to a second, separately owned location inside Uganda, not because anyone remembered to configure it, but because backup is a platform policy, provided an eligible second location is available.

The decision-support tool runs on a small open medical model, a few billion parameters, sharing a sliced GPU with other tenants. Ugandan teams have already field-tested offline clinical decision-support tools built on small open medical models with health workers in rural districts.

At the facilities themselves, a lightweight local copy keeps working offline and synchronises when the line returns.

Increasingly, the clinic group’s operations manager will not even need the IT provider for small applications. She describes the booking system she needs to an AI assistant; the assistant writes it and, through a controlled interface to the platform, deploys it into the clinic group’s own environment, within her permissions, with every action logged, and with backup and the second site inherited automatically under the configured platform policies.

The model decides what to build. The grid guarantees where and how it runs.

Government: A ministry that owns its cloud, not just its data

A revenue authority in an East African country operates under a cloud policy that requires government data to be stored and processed inside the country. Its tax filing system, its taxpayer registry and its internal analytics must stay sovereign. Its public website and published statistics need not.

On the grid, the authority runs a sovereign location under a Build-Operate-Transfer arrangement.

For the first years, the platform is operated with the authority’s own engineers embedded in the team. At the end, operations, governance and a copy of the platform’s source code, with the agreed development and usage rights, pass to the public institution.

Classified workloads never leave that location. Disaster recovery runs at a second national location operated by another federation member under the same rules.

Non-sensitive workloads, including the public portal, open data and a chatbot answering taxpayers’ questions in Kiswahili and English, can use federated capacity elsewhere in the country when filing deadlines create peaks.

The point is not only cost. It is that the state could keep running the system if any supplier, including us, disappeared tomorrow.

Fintech: One lender, three regulators, one operating model

A digital lender serves small traders in Uganda, Kenya and Rwanda.

It lives on mobile money rails, a market that processed more than $2 trillion in transactions globally in 2025, with sub-Saharan Africa accounting for the majority, according to the GSMA State of the Industry Report on Mobile Money 2026.

Its credit scoring and fraud detection depend on data subject to each country’s data protection and financial-sector requirements.

Today, that can mean a choice between three different hosting providers with three different platforms, or a hyperscaler region outside the continent and an uncomfortable conversation with three central banks.

On the grid, the lender keeps a single contract with its home provider in Kampala and deploys into federated locations in each of the three countries.

Each country’s customer data stays in that country. The applications, the security controls and the operations team are the same everywhere, because the platform is the same everywhere.

Its credit and fraud models are small, purpose-built and run next to the data, on shared accelerators. This is inference, not training, which is exactly the kind of workload previous-generation hardware can handle well.

When the lender expands to Tanzania, it does not start a new procurement. It deploys into the next federated location, if one exists and meets the applicable requirements. If it does not yet exist, the lender’s demand is precisely the signal a local operator needs to build one.

Education: The university that is both customer and supplier

A university in Kampala wants three things that rarely fit one budget: a reliable learning platform for twenty thousand students, research compute for a team fine-tuning language models for Luganda and Runyankole, and a way to teach students on the infrastructure they will actually work with when they graduate.

On the grid, the university can be a tenant and a member at once.

Its learning platform runs on the grid, scaling up in examination weeks and down in the holidays. Its research team uses shared accelerators for fine-tuning runs, paying for hours rather than buying a cluster that sits idle between projects.

And if the university runs a small location of its own, a server room it already has, it can offer spare capacity back to the federation during the holidays, provided it meets the federation’s technical and operational requirements, turning an idle cost into a modest income.

Most valuable is the third need.

Part 3 of this series argued that Africa’s options in 6G will depend on engineers who can run and extend open platforms. A university whose students learn on the same federated platform that operators, ministries and banks run in production is not just a customer. It is the pipeline that makes the whole system sovereign in practice.

Why the grid, and not something else

Put the four cases side by side and the same properties keep returning.

Data stays where the law says, workload by workload.

The clinic’s records stay in Uganda, the ministry’s in its country, and the lender’s in each of three. Residency is a rule the platform enforces according to the applicable requirements, not merely a promise in a contract.

Resilience comes from distribution.

None of these organisations built a second data centre. Each can get a second site because another member already has one, where suitable capacity is available.

Capacity follows demand.

Exam weeks, filing deadlines and month-end lending peaks draw on shared capacity instead of permanent over-provisioning. New locations appear where demand proves itself.

Ownership and jobs stay local.

The infrastructure belongs to Ugandan, Kenyan and Rwandan operators, ministries and universities. The margin, the skills and the operational knowledge can stay with them.

One contract, continental reach.

A provider in Kampala can serve a customer in Kigali or Dar es Salaam without building there, where federated locations are available. It is the practical, digital counterpart to the African Continental Free Trade Area.

And the grid is being prepared for its newest users.

AI agents will increasingly write and deploy software on behalf of people who are not engineers. That is a productivity gain Africa should welcome, but only if what the agents deploy lands somewhere with backup, isolation, monitoring, audit trails and a known jurisdiction.

A federated grid can give every agent a safe default destination inside the relevant border.

What is real, and what is still being built

Honesty matters more at the end of a series than at the start.

Parts of what I have described run today, according to whitesky.cloud: federated locations in Europe and Africa, including a live location in Uganda; one contract across locations; built-in backup with restore in another location; and GPU slicing for shared accelerators.

Other parts are being built. The interface that lets AI agents deploy onto the platform is in validation with partners, finer-grained permissions for agents are next, and replication between locations is in development.

Most of the African locations this vision needs do not exist yet. And the federation still needs the rulebook any roaming system depends on: admission criteria, service levels and dispute resolution, written and governed in a way members trust.

None of that is a reason to wait. It is a description of the work.

Closing the series

This concludes the four-part series.

The case, in one paragraph: Africa should not borrow to buy hardware at the top of a cycle; open models and right-sized, previous-generation compute already serve many of the continent’s practical AI needs.

That compute belongs in a federated grid: independently owned, commonly operated, sovereign by rule rather than by building.

That grid is also Africa’s best seat at the 6G table.

And, as these four examples show, it is not an abstraction. It is a clinic, a ministry, a lender and a university, each keeping control of what is theirs while sharing what makes sense to share.

African mobile became a success because independent operators agreed to work together under common rules. The AI era offers the same opportunity.

Governments, enterprises and service providers that want to help build it, as anchor tenants, as Cloud Enablers or as operators serving their own customers, are welcome to join.

Hans van Linschoten is the CEO of whitesky.cloud, a European sovereign cloud platform, and Co-founder of AfriQloud, a pan-African sovereign cloud and edge AI infrastructure initiative.

Go to TECHTRENDSKE.co.ke for more tech and business news from the African continent and across the world.

Follow us on WhatsApp, Telegram, Twitter, and Facebook, or subscribe to our weekly newsletter to ensure you don’t miss out on any future updates. Send tips to info@techtrendsmedia.co.ke