Ninety-eight per cent complete in 1994, and still not commissioned.
Construction began in 1979, with the Soviet state firm Tyazhpromexport as builder 2, and the contract scheduled completion for 1986 3. By 1994 the plant was reported 98% complete 2. Since then it has passed through a concession, the termination of that concession, and a $496m settlement 4. The rolling mills were built ahead of the steelworks and have run on imported billet. The blast furnace, which is what would make this a steel plant rather than a finishing shop, has never been turned on 3.
Published estimates put the spend at more than $8bn 1, and some at $10bn 2. The figure that matters is not any single year’s budget line. It is the total the sunk-cost fallacy has protected for forty-seven years.
This was always a make-or-buy question, priced as a nation-building one.
Strip the symbolism and the decision reduces to a single comparison: the delivered cost of a domestically produced tonne of steel against the delivered cost of an imported one, discounted by the probability the domestic tonne ever arrives. For forty-seven years, that third term, the probability of completion, has been treated as one. It is not.
The bear case the consensus is missing is rarely a new fact. It is the base rate the room has agreed to stop counting.
Ajaokuta’s own record is the base rate: scheduled for 1986, 98% complete in 1994, and not commissioned in 2026. The build case survives only when each new appropriation is underwritten as if that record did not exist, as if this tranche, unlike every tranche before it, closes the gap.
The disciplined move is to buy the steel and redirect the build.
Once completion probability is priced honestly, the domestic tonne is not marginally more expensive. It costs its nameplate price divided by its odds of ever arriving, before the opportunity cost of the capital held hostage to it. For an operator or a sovereign principal weighing exposure to the complex, the conclusion follows from the arithmetic, not from sentiment about industrial sovereignty.
Treat the sunk $8bn as sunk. Underwrite the next dollar on its own completion odds. Those odds say import the billet, redirect the capital, and revisit only against a financed, contracted commissioning date that has never yet existed.
The 2024 MoU with Tyazhpromexport 5 and the 2026 talks on a $2bn Chinese-backed plan 6 are the latest attempts. Neither has yet produced that date. None of this requires a view on steel demand, on the naira, or on the political will of any government. It requires only that the completion term be set to its observed value rather than to one. That is the whole of the analysis, and it is the part the story leaves out.