The Department of Electricity and Energy announced on October 28th the closure of the first window for applications in the Independent Gas Generation Procurement Program without receiving a single preferred application. Four submitted proposals, offering capacity of 2830 MW against the target of 2000 MW, failed to meet mandatory criteria, mainly due to participants' inability to guarantee sufficient gas supply volumes.
The situation is complicated by the fact that South Africa does not have an LNG import terminal, and its pipeline supplies from Mozambique are in decline. However, this was not unexpected for sector specialists.
Earlier that month, the minister issued a decision on an additional 5000 MW of gas generation, as well as 4600 MW of battery storage. Furthermore, a separate decision was made allowing eKurhuleni to purchase 2400 MW from independent producers, of which 780 MW is natural gas. Notably, gas is considered first, while wind and solar energy must await a subsequent decision. The technology, which recently failed due to fuel shortages, is now a priority.
The procurement decision creates a legal authorization, but it does not secure fuel, terminals, or reliable contracts. The minister is aware of this, stating that distribution must occur 'through a coordinated fuel infrastructure and generation program,' and emphasizing that a plant without a reliable fuel source cannot ensure stable power supply.
Following the failure of the First Application Window, the department stated that it is not backing down and will now focus on infrastructural, commercial, and institutional conditions. Nevertheless, the October decisions are based on an unverified assumption: repeating the demand instrument while maintaining the supply problem will not lead to a different outcome.
Official statements acknowledge the fuel problem but do not demonstrate that it has been solved. A more complex issue is that the fuel problem that halted the First Application Window is physical. Its resolution will inevitably create a financial problem.
The guarantees that bidders could not provide will eventually be secured by long-term LNG supply contracts, terminal agreements, and 'take-or-pay' obligations, whose prices will be set in dollars and linked to global markets. This is where the price risk arises, and once these contracts are concluded, it will be fixed for a generation of electricity consumers.
The question arises as to why a rational politician would prioritize gas generation under such circumstances. What is happening in energy policy bears the hallmarks of tragedy. This is not due to the bad faith of anyone, but because every participant acts rationally within a system that leaves the biggest risk ownerless.
Between January and October 19, 2026, the price of gas at Europe's main trading hub, the Dutch TTF, rose from approximately 28 to 88.38 euros per megawatt-hour, equating to about 9.20 to 29.00 US dollars per million British thermal units (MMBtu). Using the thermal efficiency coefficient employed in the author's study, 7 gigajoules (about 6.6 MMBtu) per megawatt-hour, a combined cycle gas turbine would incur fuel costs of approximately 192 US dollars per MWh at this price. At an exchange rate of 16.53 Rand to the dollar, this amounts to about 3.18 Rand per kilowatt-hour for fuel alone. This figure is a lower bound, not an estimate, as it excludes costs for transportation, regasification, capacity charges, and efficiency losses when operating a medium-sized station, such as the Richards Bay Eskom station proposed, in a demand-following mode.
On the scale of the entire park, the Integrated Resource Plan 2025 foresees 6000 MW of gas generation by 2030. According to the author's estimates, a park of this size will require about 3.4 million tons of LNG annually. At a price of 20 US dollars per MMBtu on the Asian benchmark JKM, the fuel cost alone will add approximately 14% to the average electricity tariff, which is not accounted for in the current definition of prices by the South African National Energy Regulator (NERSA). At a price of 40 US dollars and a 20% weakening of the Rand, this share will amount to about 32%.
It is also unclear how this bill will be covered. Today, Eskom passes on excess fuel costs to consumers through the Nersa Regulatory Settlement Account 18–24 months after the event. However, LNG supply contracts of this magnitude run for 15–20 years, far beyond 2031, when the Electricity Amendment Act transitions the system to market pricing. The draft Nersa Trading Rules for 2026 lists six categories of non-bypassable levies, and variable fuel costs are not among them; subsequent arrangements remain at the discretion of the wholesale tariff methodology, which has yet to be determined. Contracts will be signed under one reimbursement regime but will exceed its term.
Financial institutions will not lend to an unviable project. This is the key point. A lender's credit analysis focuses on whether the plant will be built, whether its revenues will cover the debt, and whether purchase and support agreements can be enforced. Risks related to gas prices and exchange rates are not ignored; they are distributed. Fuel costs and capacity charges pegged to the dollar, 'take-or-pay' obligations, and state buyer support move these risks outside the lender's control. Within this boundary, the project is considered financially viable. Outside of it, the same contracts transfer global gas price shocks and currency shocks into the consumer's expenses, the treasury, and ultimately, the regulated tariff paid by every household and business. External exposure goes beyond the scope of the lender's assessment, not due to oversight, but by design.
The Karpowership licensing demonstrated how this looks in practice. When opponents raised the issue of dollar-indexed fuel costs for three floating power stations licensed by Nersa in 2021, the regulator responded by considering the licensee's right to reimbursement of its costs, rather than who would pay if gas prices rose. The question was asked, and the institution best suited to answer answered a different one.
This is where the tragedy lies. The developer optimizes the project. The lender protects its loan. The buyer signs contracts they are obligated to sign. The regulator assesses the application before submission. Every account is correct within its own system. Risk accumulates in the space between these systems, and no institution is formally assigned to measure it before it becomes fixed. Households will not face this risk as a condition in the electricity purchase agreement. They will face it as a tariff increase years after the contracts are signed.
Three steps could eliminate this gap, regardless of the conclusion drawn about gas. Firstly, before the financial closing of any LNG-to-power project, Nersa should require and publish a tariff stress test: what happens to the regulated tariff if hub prices replicate this year's movement with a weaker Rand. Mathematically, this takes minutes; only the obligation to publish it is missing.
Secondly, the National Treasury must report on contingent liabilities created by state support for gas procurement as clearly as it reports on guarantees to state-owned enterprises, before they arise, not after.
Thirdly, the department must make the terms of procurement risk allocation public: how fuel costs are indexed, who pays 'take-or-pay' obligations when the plant is not needed, and how currency fluctuations are handled. In the Karpowership case, the High Court in 2024 ruled that Nersa submitted a heavily edited record according to the company's wishes. Terms determining what citizens will pay cannot be treated as a trade secret.
A rational politician can still choose gas generation. But only one who has seen the numbers outside the boundaries of lenders and is willing to publish them can claim to make a rational choice.
Before the next application window opens, one question must be resolved first: who owns the risk? The Karpowership contracts expired before anyone was forced to take responsibility for the risks they incurred. The five-fold larger program may not expire. Its risks need an owner before the ink dries.