Moment in Energy History-Part 2

South Africa has chosen to liberalise its electricity sector at the precise historical moment that the global energy order is being structurally rewritten—not by governments, not by regulators, but by the world’s most capitalised private corporations quietly building their own power stations and walking away from the grid entirely. The Electricity Regulation Amendment Act (ERAA) of 2024 is not merely a policy error of local dimension; it is a category mistake of global consequence, timed with the precision of a man who steps off a ship onto a departing pier.

I. THE TWIN TECTONIC FORCES RESHAPING GLOBAL ENERGY

Two concurrent and structurally unprecedented forces are converging on the world’s electricity infrastructure with a simultaneity that no grid was designed to absorb. The first is artificial intelligence. The second is transport electrification. Neither is hypothetical. Both are already loading the grid—and their combined effect constitutes the most consequential demand shock in the history of electrical engineering.

AI data centres do not consume power the way factories do. They require uninterrupted, high-density baseload power—24 hours a day, 365 days a year—with zero tolerance for voltage instability or supply interruption. A single hyperscale AI training cluster can demand 500 megawatts continuously. A cluster of such facilities demands the equivalent output of a mid-sized coal plant, indefinitely. The International Energy Agency (IEA) projects that global data centre electricity consumption will double by 2030, reaching between 945 and 1,050 terawatt-hours annually—equivalent to the entire electricity consumption of Japan[1]. Goldman Sachs Research estimates that data centre power demand will grow 160% by 2030, driven primarily by generative AI workloads[2].

Transport electrification adds a different but equally destabilising load profile: highly flexible, geographically distributed, time-sensitive, and politically non-negotiable. Electric vehicles (EVs), charging at scale, create sharp evening demand peaks that stress distribution infrastructure in ways that mid-century grid designs never anticipated. The IEA projects over 300 million EVs on global roads by 2030, each representing a 50–100 kilowatt-hour battery connected to the grid at irregular intervals[3].

What makes this convergence particularly dangerous for grid planners is its simultaneity. AI data centres demand stable baseload; EVs demand flexible peak capacity. These two load types are architecturally contradictory, and they are hitting most major grids within the same capital investment cycle. The result is not merely additional demand; it is a structural transformation of the demand curve itself.

The world’s most sophisticated energy consumers have concluded that the competitive electricity market cannot be trusted with their operational survival. They are not seeking better market access—they are leaving the market entirely.

II. THE GREAT DEFECTION: WHEN CAPITAL ABANDONS THE MARKET

South Africa’s policymakers would benefit from studying closely what the world’s most sophisticated private energy consumers are doing in response to this new reality—because their behaviour offers a devastating empirical verdict on the very model the ERAA is trying to import.

Amazon Web Services has concluded long-term Power Purchase Agreements (PPAs) with multiple nuclear operators across the United States, securing dedicated baseload capacity for its data centre infrastructure. Microsoft has gone a decisive step further: it signed a 20-year nuclear PPA with Constellation Energy to restart the Three Mile Island nuclear facility—a project that was economically unviable under market conditions until Microsoft’s contracted offtake made it bankable[4]—restoring approximately 835 megawatts of carbon-free baseload capacity to the grid[5]. Meta is developing dedicated gas generation capacity adjacent to its data centre campuses. Google has committed to procuring 24/7 carbon-free energy, constructing a dedicated energy supply chain that runs parallel to, and independent of, competitive wholesale markets.

These are not ideological statements about the inadequacy of markets. They are hard-nosed engineering and financial decisions made by organisations with trillion-dollar balance sheets and the deepest energy expertise in the private sector. Their conclusion is unambiguous: competitive electricity markets, as currently constituted, cannot guarantee the supply certainty, the baseload depth, or the transmission stability that 21st-century digital infrastructure demands. The American grid’s interconnection queue alone—which routinely takes 5 to 7 years to process new generation applications—has effectively rendered competitive market participation structurally impractical for entities that need power now, reliably, at scale.

The structural irony confronting South Africa is acute: the ERAA creates the formal apparatus of a competitive electricity market—trading platforms, market operators, balance responsible parties, a central purchasing agency—at the precise historical moment when the entities most capable of providing investment certainty in new generation capacity have decided they no longer need electricity markets. South Africa is building an electricity trading floor at the moment; the most powerful buyers are building their own power stations.

III. THE 2030 CAPACITY CLIFF: A MATHEMATICAL RECKONING

Against this global backdrop, South Africa faces a domestic crisis of extraordinary specificity. Approximately 8.4 gigawatts of coal-fired generation capacity is scheduled for retirement before 2030. This is not a projection or a modelling assumption; it is a contractual and technical reality governed by the physical limits of ageing plant infrastructure[6]. The country requires, simultaneously: replacing this dispatchable baseload, integrating an additional renewable energy pipeline to meet the 40% variable renewable target, and R390 billion in grid expansion to accommodate distributed generation—all within roughly five years.

The ERAA’s answer to this challenge is the competitive market mechanism. The architecture is elegant on paper: private capital, attracted by market signals, builds the required generation; the independent Transmission System Operator coordinates the grid; NERSA sets the regulatory framework; and the Minister of Electricity retains a backstop intervention power under Section 34 of the Electricity Regulation Act. The logic is internally coherent. The problem is that it is internally coherent, like a perfect clock—it tells the right time everywhere except where you are standing.

Section 34 is the mechanism most frequently cited by ERAA proponents as the ‘Sovereign Backstop’—the emergency brake that prevents market failure from becoming a national catastrophe. The argument deserves precise analytical treatment, because it is not wrong in theory; it is wrong in time. The Minister can only invoke Section 34 after a market failure has been formally identified. The minimum engineering lead time for a new coal or gas plant is 4 to 5 years from financial close to first power. New nuclear capacity requires 8 to 12 years. Even fast-track open-cycle gas turbines require 18 to 24 months under ideal conditions—conditions that do not presently exist in South Africa’s procurement environment. The arithmetic is simple and damning: a Section 34 determination triggered by a 2027 market failure cannot commission dispatchable capacity before 2032. The 2030 cliff does not wait for ministerial paperwork.

Section 34 is not a sovereign backstop. It is a sovereign postscript—a mechanism for explaining the collapse after it has occurred.

The transmission paradox compounds the problem geometrically. Private generation investors will not commit capital to new power plants without certainty about grid access and offtake. The TSO will not invest in transmission infrastructure without certainty about generation volumes and revenue streams. Each party rationally waits for the other. This is not a market failure in the theoretical sense; it is a structural coordination failure endemic to the fragmented architecture the ERAA creates. Eskom’s integrated ‘plan-build-operate’ model was, whatever its other deficiencies, immune to this failure mode by design: the same entity planned the generation, built the transmission, and operated the system. You cannot have a coordination failure when the coordinating entity is singular.

IV. INSTITUTIONAL DISTANCE AND THE ARCHITECTURE OF UNACCOUNTABILITY

There is a political economy dimension to the ERAA that the celebratory rhetoric of ‘reform’ consistently obscures. The fragmentation of Eskom into multiple entities—a generation company, a transmission SOC, a distribution function, and a market operator—does not merely create technical coordination problems. It creates what we might call an ‘architecture of unaccountability’: a governance structure so distributed that no single institutional throat can be seized when the lights go out.

Under the ERAA framework, the chain of accountability for electricity supply runs from the Minister’s determination powers through the TSO’s market operations, through the generation licensees’ plant performance, through NERSA’s regulatory environment, and back to the Minister’s backstop authority. When the system fails—and the 2030 mathematics suggest it will—the institutional reflex of each entity in this chain is to point to the next. This is not cynicism; it is the predictable behaviour of organisations responding rationally to their defined mandates. The ERAA does not merely fail to prevent this diffusion of accountability; it constitutionally enshrines it.

South Africa’s developmental state tradition—the constitutional mandate for the progressive realisation of access to energy services—requires an institutional architecture capable of discharging clear sovereign obligations. The ERAA transforms what was, in Eskom’s integrated model, a clear state duty into a complex contractual lattice between private parties, regulatory bodies, and state entities with overlapping and frequently conflicting mandates. When the grid fails at this crossroads, there will be no single responsible party; there will only be a thousand litigants.

V. THE TEXAS PRECEDENT SOUTH AFRICA IS CHOOSING TO IGNORE

The empirical literature on failures in deregulated electricity markets is substantial, but the most analytically relevant precedent for South Africa’s situation is the Texas ERCOT crisis in February 2021. Texas operates the most thoroughly deregulated electricity market in the United States—a market explicitly designed on the competitive efficiency principles that underpin the ERAA. During Winter Storm Uri in February 2021, the ERCOT grid came dangerously close to a catastrophic, uncontrolled collapse. The system frequency dropped below 59.4 Hz, triggering a nine-minute time delay on the generator underfrequency relays. If the frequency had remained below 59.4 Hz for nine minutes, approximately 17,000 MW of additional generation would have tripped offline, potentially causing a total blackout of the ERCOT Interconnection, which grid operators estimated could have left 12 to 14 million Texans without power for months. At least 210 people died during the Event, with most of the deaths connected to the power outages, of causes including hypothermia, carbon monoxide poisoning, and medical conditions exacerbated by freezing conditions[7]. Economic losses from output losses and damage were estimated at $130 billion in Texas alone and $155 billion nationwide [8]. ​

The post-mortem analysis of the Texas crisis identified the precise failure mode that critics of the ERAA warn awaits South Africa: a market optimised for efficiency under normal conditions but structurally incapable of maintaining the reserve margins required for systemic resilience under stress. Texas, like South Africa under the ERAA model, had no mandatory reserve margin requirements—because reserve margins represent idle capacity, and idle capacity is, by definition, economically inefficient in a competitive market. The market had rationally priced out exactly the redundancy that would have prevented the crisis.

South Africa’s proponents of the ERAA are aware of the Texas precedent. Their response is typically that South Africa’s market design is different, that the Section 34 backstop provides the reserve margin mechanism ERCOT lacked, and that the regulatory framework is more robust. These arguments are not without merit in the abstract. They collapse, however, when confronted with South Africa’s specific conditions: a reserve margin already negative, a coal fleet in accelerating structural decline, a transmission network requiring R390 billion in investment, and a 2030 deadline that Section 34 cannot, by the calculus of engineering, meet.

VI. WHAT SOUTH AFRICA ACTUALLY NEEDS

The argument advanced here is not that South Africa’s electricity sector requires no reform. The integrated Eskom model, at its operational peak in FY2017 when the Energy Availability Factor reached 78%—rising to 86% in July of that year—demonstrated that it was capable of high performance under capable leadership[9]. The argument is that the specific reform embodied in the ERAA is the wrong reform, designed for the wrong problem, at the wrong time.

A credible alternative framework would preserve Eskom’s integrated transmission and system operation functions while introducing disciplined private participation in generation through long-term, competitively tendered PPAs with state offtake guarantees — applied, with precision, to the capacity type the moment demands. For the 2030 cliff, that means firm dispatchable capacity: gas bridging at scale, with strategic life extension of existing assets as the contingency against gas infrastructure delays. The contractual architecture is the same one that the ERAA dismantles. The technology it must procure is categorically different from that delivered by the REIPPP.

What South Africa actually requires is precisely what Microsoft has demonstrated is necessary: long-term, bankable, contracted offtake certainty for new dispatchable generation capacity. The Constellation deal did not become financeable because of competition — it became financeable because Microsoft’s 20-year demand commitment provided the revenue certainty that no market signal, however well-designed, could replicate. That is the mechanism South Africa needs for its 2030 capacity replacement challenge.

The ERAA’s defenders will correctly note that the Act is not a pure market for merchants. It contains a Central Purchasing Agency empowered to conclude long-term power purchase agreements, and a Section 34 mechanism under which the Minister may act not only in the event of market failure, but also proactively, to ensure the security of energy supply in the national interest. These are not trivial provisions. In principle, they are precisely the instruments this analysis identifies as essential.

They are also insufficient for 2030. The Section 34 mechanism — however triggered — cannot alter the engineering reality that follows. The minimum lead time for a new gas plant from financial close to first power is four to five years under ideal conditions; conditions that do not presently exist in South Africa’s procurement environment. A Section 34 determination made today cannot commission dispatchable baseload capacity before 2029 at the earliest. South Africa does not have a mechanism problem. It has a timing problem — and at this stage of the 2030 countdown, a timing problem is a capacity problem.

Consider what 2.64 gigawatts per year of grid-ready VRE integration actually represents[10]. It is not a policy target that better governance could substantially exceed. It is a structural ceiling derived from the physical limits of South Africa’s transmission infrastructure[10]. Achieving it requires sustaining 1,500 kilometres of new transmission construction per year — already a fivefold increase over the baseline rate and already straining every available supply chain [10]. Doubling that rate to 3,000 kilometres per year — physically implausible according to the Transmission Development Plan’s own supply chain assessment — would raise integration velocity only to 4.4 gigawatts per year, which still falls short of the 4.75 gigawatts per year being retired[10]. No single intervention, under any realistic parameter, closes the gap before 2030[10].

The paradox runs deeper still. South Africa’s solar irradiation is 60–70% higher than Germany’s, with capacity factors among the highest in the world[10]. Germany, with six times the transmission density and none of South Africa’s greenfield construction burden, integrates renewables at 15 gigawatts per year — 5.7 times faster[10]. Superior renewable resources, contractually secured through any procurement model, cannot evacuate power that the transmission network cannot carry.

But there is a further dimension that the nameplate comparison obscures, and it is the most consequential of all. The 1.8-to-1 ratio — 4.75 gigawatts per year retiring against 2.64 gigawatts per year integrating — compares nameplate capacity to nameplate capacity[10]. It treats a gigawatt of coal and a gigawatt of solar as equivalent units. They are not. Coal generates power on demand, at full rated output, whenever the system operator requires it. Solar generates when the sun shines — at South Africa’s documented capacity factors of 25–28% for solar PV and 35–40% for wind, a blended portfolio of 2.64 gigawatts of VRE nameplate delivers approximately 810 megawatts of effective firm contribution — power that can actually be counted on when demand peaks[10]. The 4.75 gigawatts per year being retired does not carry such a discount. It is synchronous, dispatchable baseload — the kind that anchors system frequency, provides rotational inertia, and generates regardless of weather.

When the comparison is made in firm capacity terms rather than nameplate terms, the ratio is not 1.8-to-1. It is closer to 5-to-1[10]. For every gigawatt of firm dispatchable capacity that retires, South Africa would need to build, connect, and firm up approximately 5 gigawatts of VRE nameplate capacity to achieve an equivalent adequacy contribution — within a 24-month window, with a total nameplate integration ceiling of 5.28 gigawatts[10]. The arithmetic is not close. It is not a matter of optimising procurement models, institutional reform, or market design. Replacing dispatchable baseload with variable renewables at the velocity the 2030 cliff demands is a physical impossibility — and no market architecture, however well-designed, repeals the physics.

This is not an argument against the REIPPP or against renewable energy. The REIPPP demonstrated, conclusively, that sovereign offtake certainty unlocks private investment in generation: 6.4 gigawatts of new capacity between 2011 and 2020, delivered because contracted revenue certainty provided what no competitive market signal could replicate. That model works. It must form the backbone of South Africa’s post-2030 generation build. But it cannot be retrofitted onto the 2030 cliff because the cliff is not a procurement-model problem — it is a dispatchability problem operating within physics constraints that no contractual innovation resolves [10].

What remains must be stated with equal candour. The preceding analysis has identified sovereign offtake certainty as the essential mechanism — the instrument that makes dispatchable capacity bankable. That prescription confronts a fourth constraint that this analysis cannot dissolve: the sovereign is fiscally compromised. The Transmission Development Plan flags Eskom’s liquidity constraints and tightening fiscal consolidation as material execution risks, noting that peak transmission investment of R112 billion over the first five-year horizon coincides with the period of greatest fiscal pressure [11]. These are not background conditions — they are active constraints on the state’s capacity to issue and honour the sovereign offtake commitments the 2030 cliff demands. A sovereign guarantee issued from a position of documented liquidity stress is not costless. It is priced by lenders as elevated risk, raising the cost of capital for the very projects it is meant to make affordable. The REIPPP itself illustrated this dynamic: the contractual mechanism worked when the offtaker’s covenant was credible; by 2018 that covenant was under stress, and the sovereign backstop behind it was being tested in precisely the circumstances the mechanism was designed to avoid[10].

South Africa, therefore, does not face a trilemma. It faces a quadrilemma: accepting adequacy deficits exceeding 4 TWh annually that breach the constitutional right of access to basic services[13]; extending coal operations beyond environmental compliance deadlines and forfeiting the $13.7 billion JETP partnership[12]; pursuing gas bridging at scales that exceed current LNG infrastructure and supply security[10]; or issuing sovereign offtake guarantees that stress an already fiscally constrained balance sheet at the moment transmission investment demands are at their peak[11]. No pathway through this quadrilemma is clean. The honest policy choice is not between a good option and a bad one — it is between managed costs that are visible, bounded, and sequenced, and unmanaged costs that are diffuse, compounding, and constitutionally indefensible. Four terawatt-hours of annual unserved energy is not an abstraction. It is hospitals, schools, and police stations operating in darkness — a condition the Constitution explicitly prohibits and that no fiscal constraint can retrospectively justify [13].

VII. CONCLUSION: THE CLOCK CANNOT BE UNINVENTED

The global energy equation has changed. The world’s largest and most sophisticated energy consumers—operating at the frontier of the AI economy South Africa aspires to participate in—have rendered their empirical verdict on competitive electricity markets: they are insufficient for the demands of the 21st-century digital economy. They are building their own power plants. South Africa, by contrast, is dismantling its.

The ERAA will proceed. President Ramaphosa’s insistence on this point is not in doubt, and the Act is law. What remains in doubt is whether South Africa will have sufficient dispatchable generation capacity by 2030 to power the industrial economy, the data infrastructure, and the electrified transport network its citizens require. The mathematical evidence suggests it will not—not because the ERAA’s architects lacked intelligence or sincerity, but because the model they imported was designed for a world that no longer exists.

The 2030 Capacity Cliff will not be argued away by ministerial rhetoric or regulatory ingenuity. It is a physical reality, governed by the thermodynamics of ageing turbines and the calendar of contracted retirements. When that cliff arrives, the architecture of unaccountability the ERAA has constructed will ensure that no single institution bears responsibility for the collapse—only the South African economy and the citizens whose livelihoods depend on it will bear the consequences.

The fatal flaw is not a bug in the system. It is the system itself. And in 2030, the calculus will finally settle the argument that the politicians refuse to hear.

ABOUT THE AUTHOR

Matshela Koko served as Eskom’s Chief Technology Officer (2009-2014), Chief Generation Officer (2015–2016) and Interim Group CEO (2016–2017). He is currently the Managing Director of Matshela Energy and a doctoral candidate at the UNISA Graduate School of Business Leadership, where his research examines Eskom’s operational performance, infrastructure governance, and the political economy of electricity-market reform in South Africa.

NOTES AND REFERENCES

1. International Energy Agency (IEA). Electricity 2024: Analysis and Forecast to 2026. Paris: IEA, 2024. Data centres and AI chapter, pp. 89–104.
2. Goldman Sachs Research. “AI Is Poised to Drive 160% Increase in Data Centre Power Demand.” GS Briefings, May 2024.
3. International Energy Agency (IEA). Global EV Outlook 2024. Paris: IEA, 2024, p. 12.
4. Constellation Energy. “Microsoft and Constellation Establish Partnership to Advance Nuclear Energy, Reopen Three Mile Island.” Press release, 20 September 2024.
5. Ibid. The Microsoft-Constellation agreement represents the first long-term nuclear PPA of its kind in the United States at this scale.
6. Eskom, National Transmission Company of South Africa. Medium-Term System Adequacy Outlook 2026-2030, p.32.
7. Federal Energy Regulatory Commission (FERC) & North American Electric Reliability Corporation (NERC). The February 2021 Cold Weather Outages in Texas and the South Central United States. Joint report, November 2021, p. 14.
8. Busby, J.W., et al. “Cascading risks: Understanding the 2021 Texas winter blackout.” Energy Research & Social Science 77 (2021): 102106.
9. CSIR Energy Centre. Statistics of utility-scale power generation in South Africa 2022 (1 Jan 2022 – 31 Dec 2022). v1.0, slide 118
10. Koko, M. (2026). South Africa’s 2030 Electricity Capacity Cliff: Institutional Frictions, Sociotechnical Inertia, and the Political Economy of Accelerated Coal Phase-Out. UNISA Graduate School of Business Leadership, Midrand. Available at: https://dx.doi.org/10.2139/ssrn.5794522
11. National Transmission Company South Africa (2024). Transmission Development Plan 2024. NTCSA, Johannesburg, p.10. Available at: https://www.ntcsa.co.za/transmission-development-plan/
12. UK Government (2026). Leaders mark progress on policy reforms, investment for SA’s JETP. Available at: https://www.gov.uk/government/news/leaders-mark-progress-on-policy-reforms-investment-for-sas-jetp
13. High Court of South Africa (Gauteng Division, Pretoria), 2023. United Democratic Movement and Others v Eskom Holdings SOC Ltd and Others. Case Nos 005779/2023; 003615/2023; 022464/2023, [2023] ZAGPPHC 1949. Available at: https://www.saflii.org/za/cases/ZAGPPHC/2023/1949.html.