Nigeria's AI Infrastructure Ambition: What Has Actually Been Built?

Nigeria spent 2026 building the legal architecture for a domestic AI compute industry. Three regulatory instruments were signed in August. A governance committee was established. A coordinating taskforce followed in September. An international partnership was announced in May to develop a network of AI data centres. Measured by policy output, it was a productive year.

Measured by disclosed physical capacity, the public record is close to empty.

That is the finding of this analysis, and it needs stating carefully, because it is a narrower claim than it may first appear. It is not a finding that Nigeria has built nothing. It is a finding that on the public record, as of 18 September 2026, there is no disclosed figure against which anyone outside government could assess what has been built. Those are different statements, and the distinction matters for what follows.

The standard applied here

This analysis measures the programme against commitments the Nigerian government published itself. No external benchmark is imposed, and no comparison to the United States or China is drawn — a comparison whose result is known in advance carries no information.

Two documents set the standard. The first is the National Artificial Intelligence Strategy, published in September 2025, led by the Federal Ministry of Communications, Innovation and Digital Economy with the National Information Technology Development Agency (NITDA) as a key implementing body. It sets a five-year vision running to 2029, organised around three goals — economic growth and competitiveness, social development and inclusion, and technological advancement and leadership — and five operational pillars.

The second is the set of instruments issued during 2026 under the National Sovereign Cloud Initiative, which give the strategy its infrastructure limb.

Throughout, three states are recorded separately and never merged:

• Announced — a party has stated an intention.
• Contracted — a binding agreement, filing or procurement record exists.
• Operational — the capacity is in service and that service can be evidenced.

Most coverage of infrastructure programmes, in Nigeria and elsewhere, reports all three as one number. Keeping them apart is the entire analytical content of this piece.

What is documented

The policy layer has moved, and it has moved fast.

In August 2026 NITDA signed three instruments under the National Sovereign Cloud Initiative: the National Cloud Computing Guideline, the National Cloud Technical Guideline, and the National Digital Infrastructure Assurance Framework. A National Cloud Investment Strategy was presented alongside them. The agency stated that implementation would begin immediately on a phased basis, and established a Sovereign Cloud Governance Committee to provide strategic oversight.

NITDA also set a dated, checkable target: by October 2026, it plans to operationalise a national digital regulatory platform for the onboarding, assessment, certification and regulation of cloud and digital infrastructure providers operating in Nigeria.

That target deserves attention precisely because it is falsifiable. Much of what governments publish in this domain cannot be assessed — "positioning the country as a regional hub" has no test. A certification platform either accepts registrations by a stated month or it does not. It is the single clearest near-term indicator the programme has produced, and it resolves within weeks of this publication.

In September 2026 the agency introduced a broader package bringing together four national frameworks covering cloud governance, data protection, infrastructure standards and investment facilitation, and launched a taskforce to coordinate implementation of the sovereign cloud initiative. The package establishes a "Cloud First" policy direction for adoption.

On the physical side, NITDA announced a partnership with the International Data Centre Authority in May 2026, centred on what the parties called the Nigeria Digital Triangle — described as a network of large-scale, AI-powered data centres intended to support cloud computing, attract private investment and form a foundation for the digital economy.

Speaking at ITW Data Cloud Africa 2026 in Nairobi, NITDA Director-General Kashifu Inuwa Abdullahi framed the strategy as an attempt to convert policy and data-sovereignty requirements into investable demand — creating predictable demand for cloud, data centre and connectivity infrastructure so that private capital has a clearer path into the market.

All of the above is announced, and the instruments themselves are documented: they exist, they are signed, they are public. That is a real achievement and this analysis does not diminish it. A country cannot build regulated infrastructure without a regulatory framework, and Nigeria now has more of one than it did eighteen months ago.

What is not documented

Here the record thins sharply.

AAT Intelligence could not identify, anywhere in the public record, a disclosed figure for any of the following:

Accelerator capacity. No count of GPUs or other AI accelerators installed, contracted or on order in Nigeria — whether by NITDA, NCAIR, a federal facility or a named private operator. No SKU or hardware class is specified in any public document reviewed. This matters more than a raw number would: without a hardware class, it is not possible to say whether an announced cluster is intended for frontier training, for fine-tuning, or for inference serving only. Those are three different capabilities, three different cost structures and three different industrial outcomes.

Facility-level capacity for the Digital Triangle. The programme's flagship physical initiative has been announced without, as far as the public record shows, a site list, a floor area, an IT load figure, a phasing plan or a construction timeline. "A network of large-scale, AI-powered data centres" is a description of intent. It is not a specification, and it cannot be assessed as one.

Megawatt load. No projected electrical load figure for any announced AI facility.

Contracted generation. This is the most consequential gap, and it requires care to state accurately.

NITDA has addressed power directly, and its position is technically sound. The Director-General has stated that digital infrastructure operators will not be expected to rely solely on electricity from the national grid, and that existing regulatory provisions allow operators to pursue alternative arrangements including renewable energy, gas-fired generation and Independent Power Purchase Agreements.

That is the correct answer to the power question. Behind-the-meter and independently contracted generation is how large data centres are powered in many markets, including ones with far more reliable grids than Nigeria's. The agency is not ignoring the constraint; it has identified the standard solution to it.

But identifying a mechanism is not the same as executing one. AAT Intelligence could not identify a single executed power purchase agreement, gas supply contract or generation commitment tied to a named Nigerian AI or cloud facility in the public record. The regulatory pathway is announced. Whether anything has been contracted along it is not disclosed.

A data centre is, in engineering terms, a building that converts electricity into computation. Announced IT load without contracted generation behind it is a plan, not a facility. Until the second number is public, the first cannot be assessed — and at present neither number is public.

The governance gap

One further item bears on whether the programme's own timeline is holding.

An external policy assessment published in April 2026 by Tunani HQ found that, as of March 2026, the NITDA Code of Practice for AI had not been finalised, and that the independent AI governance body — which the assessment describes as the strategy's own capstone institution — had not been constituted.

The same assessment recommends that the governance body be constituted before the Code of Practice is finalised rather than after, on the reasoning that without an independent oversight body the Code becomes a self-regulatory instrument of the agency that also promotes AI development. It further recommends that Nigeria publish a binding implementation roadmap with quantified 12-, 36- and 60-month targets, a responsible agency for each of the strategy's 34 constituent strategies, and a budget — modelled on the roadmap Kenya published in December 2025.

AAT Intelligence has not independently verified the current status of either the Code of Practice or the governance body, and the assessment above reflects March 2026. Both may have advanced in the six months since. What can be said is that as of the most recent published external assessment, the strategy's capstone governance institution was not yet in place, roughly halfway through the second year of a five-year strategy — and that no subsequent public confirmation of its constitution was identified.

Kenya's published roadmap is the useful comparison here, and not for the reason that comparison usually gets drawn. The point is not that Kenya is ahead. It is that Kenya published a document that makes its own programme checkable: agency, target, date, budget. A programme that publishes quantified milestones can be held to them, including by its own government. A programme that does not cannot be assessed by anyone outside it — which is the condition Nigeria's is currently in.

What AAT Intelligence found

Three findings, stated at the confidence the evidence supports.

First, the policy layer is materially ahead of the disclosure layer. Nigeria has issued instruments, formed committees, launched a taskforce and set at least one dated, testable target. It has not published the figures that would let anyone determine what physical capacity exists. Both statements are true simultaneously, and reporting that presents only the first gives a misleading picture of what is known.

Second, the binding constraint on external assessment is disclosure, not delivery. This analysis cannot conclude that Nigeria has failed to build AI compute capacity. It concludes something narrower and more defensible: that on the public record, nobody outside the relevant agencies is in a position to say either way. Those two findings are frequently conflated, and the second is the one the evidence supports.

Third, the programme's most valuable near-term act would cost nothing to perform. Publishing announced load, contracted generation and operational capacity as three separate figures — per facility, with dates — would convert an unassessable programme into an assessable one overnight. It would also serve the government's own stated objective. The Director-General has framed the strategy as creating investable demand for private capital. Investors price disclosed, dated, contracted capacity. They discount announcements heavily, and they discount announcements from programmes that do not publish milestones more heavily still. On this reading, the disclosure gap is not merely an accountability problem; it works directly against the investment objective the programme was designed to serve.

Why it matters

Nigeria is attempting something genuinely difficult, and the difficulty is not primarily technical. Sovereign compute capacity requires regulatory clarity, contracted power, capital, skills and operational continuity to arrive roughly together. Most countries attempting it have assembled two or three of those and stalled.

The 2026 record suggests Nigeria has made real progress on the first. What the record does not show is whether the second has moved at all — and the second is the one that determines whether a data centre is a building or a press release.

What to watch

Four specific, checkable indicators over the next twelve months:

1. The October 2026 certification platform. Does it accept registrations by the stated month? This is the programme's own near-term deadline and it resolves imminently.
2. First disclosed power arrangement. An executed IPPA, gas supply agreement or generation commitment tied to a named facility would be the first hard evidence that the Digital Triangle has moved from announcement toward construction.
3. Constitution of the AI governance body. Its establishment, with a published work programme, would close the gap identified in the March 2026 assessment.
4. A published implementation roadmap with quantified milestones. If one appears, this analysis becomes straightforward to update against it. If none appears by the end of the strategy's second year, the programme will have passed its midpoint without a public instrument against which delivery can be measured — which is the outcome the external assessment above was written to warn against.

AAT Intelligence will revisit each of these as the record develops. Where a figure in this piece is superseded by a disclosure, it will be corrected and the change dated.