Global PPP Announcement Center — tracked projects and tenders

Our view: This is a textbook sovereign-to-sovereign PPP — two governments structuring the framework, with the private consortium to be tendered afterwards. The sovereign payment guarantee is the key credit term. For a buying government on either side, the negotiation is not about the tariff per kWh; it is about the change-in-law clause, the termination-for-convenience compensation, and the dispute resolution mechanism that sits between two sovereign legal systems. Those are the terms that will be tested over 25 years — and they are set now.

Our view: The local content commitments will dominate the bid evaluation — and they should, because they are the government's return on the concession. But with 22 years of availability payments, the real risk is the demand forecast: if the corridor underperforms against projections, the government is paying for infrastructure it does not fully use. The availability mechanism must include a volume collar — a floor below which the government's payment reduces and a ceiling above which the partner shares the upside. Without it, the risk sits entirely with the state.

Our view: Financial close on three power PPPs in one announcement cycle is a significant signal for the African PPP market. The take-or-pay offtake is the structure that makes these bankable — but it is also the structure that concentrates risk on the offtaker, which is the state. For a buying government watching this, the question is how the offtake agreement interacts with the gas supply agreement: if the gas supply fails, does the offtake obligation suspend or survive? That single clause can turn a bankable project into a contingent liability — and it is often buried in a schedule.

Our view: This is the template for sovereign AI compute procurement — and it raises questions every government contracting for compute should watch. First: who owns the data that trains on this infrastructure, and under which jurisdiction? Second: the availability model assumes predictable demand; AI compute demand is anything but. The payment mechanism must accommodate scale-up and scale-down without penalty, or the government will be locked into capacity it no longer needs. Third: technology refresh clauses. A 15-year contract for technology that doubles in capability every 18 months is a commercial risk that must be priced and governed — not left to goodwill.

Our view: Five concurrent port concessions in one call is an ambitious procurement. The key commercial question is whether the five concessions are independent or interdependent — if each operator's throughput depends on the others' performance, the government needs a coordination mechanism that no individual concession agreement will provide. The second question is the demand risk: the government must decide whether to guarantee a minimum throughput (taking volume risk) or leave it with the operator (which will price it into the tariff). Neither is wrong — but not deciding is.