The national data centre is not enough

Over two years on air, I have covered this same announcement about a dozen times. The set changes, the press release does not. A head of state cuts a ribbon in front of a new building, and we hear about 'end of dependence', 'repatriated data', 'sovereignty restored'. Gabon inaugurated its first national data centre. Burkina Faso is working on its own infrastructure and on BurkinOS. Guinea reclaimed control of its top-level domain, .gn. Algeria is displaying an AI ambition backed by a technological sovereignty doctrine.

These decisions are good. They are also vastly insufficient, and I believe it must be said plainly, because confusing owning an infrastructure with controlling one will cost the continent dearly.

Six layers, only one built

A sovereign information system requires mastery of six layers. The building is only one of them.

There is the physical layer: the land, the walls, the power supply, the cooling. This is the layer that gets inaugurated, and the only one that photographs well.

There is the network: where packets enter and leave, who operates the transit, where the exchange points are, what the real topology of routes looks like.

There is the hardware: servers, switches, disks, and above all their firmware. A server purchased from a foreign integrator arrives with a BIOS, a remote administration controller, and an update chain that the buyer does not control.

There is the infrastructure software: hypervisor, orchestrator, storage system, monitoring. Almost always under foreign licence, often with telemetry dependencies.

There is operations: the teams that keep production running at three in the morning. This is the hardest layer to build and the least funded.

There is finally the law: under which jurisdiction the data falls, what contract binds the state to the integrator, who holds the encryption keys, and what happens in case of dispute.

A country that builds the first layer and buys the other five has not won its sovereignty. It has moved its dependency from abroad to its own territory — which is real progress in terms of latency, employment and resilience, but changes almost nothing about the balance of power.

The routing problem nobody talks about

If only one thing were to be retained, it would be this.

For a long time, a considerable share of internet traffic exchanged between two neighbouring African countries physically transited through Europe. An email sent from one operator to another in the same sub-region would travel up to Marseille, London or Amsterdam, then come back down. The industry has a name for this: tromboning.

The consequences are concrete. Latency explodes, degrading all interactive services. Costs increase, because international transit is paid for. And above all, strictly intra-African communications become observable from third-party jurisdictions.

The technical answer exists and is well known: multiply national and regional internet exchange points, attract operators and content providers there, develop local peering. It is inexpensive compared to a data centre, it produces immediate and measurable gains, and it does not warrant a presidential inauguration — because an exchange point, visually, is a rack in a room.

The real indicator of sovereignty is not where my servers are, but where my packets go.

The national domain, an example of low-cost sovereignty

The repatriation of Guinea's .gn domain deserves more attention than it received, because it illustrates sovereignty that costs almost nothing and produces a great deal.

A national top-level domain is a country's name on the internet. Historically, several African domains were delegated to entities or individuals located outside the continent, under conditions negotiated at a time when nobody measured the implications. Whoever administers the registry decides who gets a name, at what price, under what conditions, and holds the DNS infrastructure that makes those names exist.

Reclaiming control of one's registry means reclaiming control of the digital identity of one's institutions, businesses and government. This requires technical skills, a legal framework and governance — not hundreds of millions.

No photograph accompanies this kind of decision. Yet it is one of the most structuring a state can make.

What should be measured

If one wanted to seriously evaluate the digital sovereignty of an African country, I propose looking at this rather than the number of data centres.

The share of national traffic that remains national. The number of active exchange points and the number of operators actually connected to them. The proportion of the country's most-visited content cached locally. The number of system engineers capable of running critical infrastructure without external assistance. Control of the national domain registry. The existence of a data protection authority with real resources. And the contractual clause that determines, in outsourcing agreements, who holds the encryption keys.

None of these indicators lends itself to a ceremony. All lend themselves to verification.

What I am not saying

I am not saying national data centres are useless. They reduce latency, create skilled jobs, open a local hosting market and give the state a negotiation lever. Gabon was right. Burkina Faso is right.

I am saying a ribbon cut is not sovereignty won, and that a country that stops at the first layer will have spent a great deal for a result it will believe to be larger than it is.

Digital sovereignty is not inaugurated. It is built layer by layer, over ten years, with engineers who must be trained, contracts that must be renegotiated and routes that must be brought back home. It is less photogenic. It is what counts.