Eskom's 6 GW Renewables-Plus-Storage Tender: Inside the 1 GW/4 GWh BESS Mandate

On 9 October 2026, Eskom Holdings issued Request for Qualifications E3612GXR: a five-year panel of co-development and co-investment partners for up to 6 GW of utility-scale solar PV, wind and battery storage by 2030. Bids close 16 November 2026. Buried in the portfolio composition is the detail that matters most to a hybrid-systems engineer — Eskom is bringing at least 2 GW of solar PV and 1 GW / 4 GWh of battery energy storage to the table itself, and structuring every project as a ring-fenced SPV on project finance.

That 4 GWh is a four-hour system on a power basis. It is not a peaker. It is a mid-duration asset designed to move solar generation across the evening peak — which, in a system that has now run more than 350 days without load shedding, is a fundamentally different job than the one batteries were doing in South Africa between 2022 and 2024.

What Eskom is actually procuring

According to pv magazine, the RFQ establishes a pre-qualified panel; panel members are then invited to bid for specific projects through one or more subsequent RFPs. ESS News reports the framework covers solar PV, wind and BESS under a five-year agreement, with Eskom contributing the solar and battery blocks. Engineering News frames it as part of Eskom's transition toward a renewable-energy subsidiary, Eskom Green.

The structure is worth reading carefully because it inverts the usual independent power producer model. In South Africa's REIPPPP rounds, developers built, owned and operated assets against a 20-year power purchase agreement. Here, Eskom is supplying land, projects and a portion of the capital, and inviting partners into ring-fenced SPVs under project finance. The eligibility summary describes panel members bidding later for specific projects on both Eskom land and partner-sourced sites.

For the storage engineer, the practical consequence is that the battery specification arrives partly pre-decided. A developer who would have sized six hours of storage to capture a full evening peak now inherits a four-hour block from the utility and must decide whether to augment it, co-locate additional capacity, or accept the shorter dispatch window and rely on a different revenue stack.

Why four hours, and why right now

The context is the recovery of the Eskom generation fleet. Eskom reported it met the full 2026 winter demand without implementing load shedding, with the energy availability factor (EAF) at its highest level since 2020 — roughly 65% by mid-2026, against a low of 54.7% in the financial year ending March 2023. Diesel expenditure for open-cycle gas turbines fell by R4.84 billion year on year.

That changes the storage use case completely. When the grid is short on capacity, a battery arbitrages diesel: you charge in the afternoon and discharge into the evening peak to avoid burning OCGT fuel. That is a two-hour, high-C-rate play. When the grid has adequate capacity but a steep evening ramp and a coal fleet that is still cycling, the battery's job becomes energy shifting at lower C-rate over a longer window — four hours, not two. The 1 GW / 4 GWh ratio is a coherent answer to that second problem, and a much less coherent answer to the first.

The residual-load arithmetic

Take the 4 GWh at face value. If the system dispatches the full 1 GW into the evening peak, it delivers four hours of nameplate output — but only on paper. Real delivered energy is bounded by three derates that any hybrid simulation must model explicitly:

• State-of-health at year N. A 4 GWh nameplate block is not 4 GWh in year 8. LFP systems typically retain 70–80% of rated capacity at end of warranty depending on cycling intensity and temperature.
• Depth-of-discharge floor. If the operating window is held between 5% and 95% SOC to protect cycle life — a common EMS constraint — usable energy drops by 10% before any degradation is applied.
• Round-trip efficiency. AC-side RTE for a full charge-discharge cycle in a utility LFP system typically lands in the 85–88% band, meaning roughly 12–15% of the charging energy never reaches the grid.

Stack those and a nominal 4 GWh block delivers materially less than 4 GWh at the point of connection in later years. A degradation-aware model that treats the battery as a fixed 4 GWh block will overstate the project's evening energy delivery every single year after the first — and the error compounds.

What the co-investment structure means for modelling

Four things about this tender should change how a developer builds the simulation model.

1. Each project is its own SPV, so each project needs its own financial model

Project-finance structures do not tolerate portfolio-level averaging. Lenders size debt against a single asset's contracted cashflows, and each SPV carries its own offtake, its own degradation curve and its own augmentation schedule. If you are bidding multiple projects off the same panel position, you need per-asset 25-year projections — not one blended model divided by N.

2. Eskom contributes assets, so ownership boundaries need modelling

When the utility supplies the PV and battery blocks and the partner supplies capital and development capability, the energy flow crosses an ownership boundary inside a single physical plant. Simulation has to attribute generation, charging energy and discharge revenue between parties. This is where a shared model of dispatch (one EMS controlling the whole plant) diverges from a shared model of economics (two balance sheets).

3. The four-hour ratio is an input, not an optimum

The most valuable analytical work in this tender will not be building the base case — it will be testing whether four hours is the right duration for each specific site. Solar resource, evening peak timing, curtailment exposure and grid constraint all vary across South Africa's provinces. A site in the Northern Cape with a long, late afternoon production tail may want three hours; a site with constrained midday export may want six. The utility's 1 GW / 4 GWh is a portfolio-level commitment. Individual project economics should still be optimized.

4. Augmentation planning is now a procurement question

Under a 20-year REIPPPP PPA, augmentation was a developer's private decision about when to top up capacity. Under a co-investment SPV, augmentation is a capital call that requires partner consent. Any model presented during the RFP stage needs an explicit augmentation schedule with SOH trigger thresholds — not a single end-of-life capacity number.

How this compares to the REIPPPP storage rounds

South Africa has already run dedicated BESS procurement under REIPPPP's battery storage programme, which awarded projects on a bid-window basis with fixed technology assumptions. Eskom Green is structurally different: a continuous five-year panel rather than discrete windows, co-development rather than pure offtake, and a utility that is a capital partner rather than only a buyer.

What to watch between now and 16 November

Three things will determine whether this panel produces bankable projects or a list of pre-qualified names that never reach financial close.

First, the land question. Eskom-sourced sites come with existing grid connection or adjacent to it; partner-sourced sites may not. Grid capacity in the Cape provinces remains constrained, and a site with a short connection queue is worth substantially more than one that waits four years.

Second, the offtake structure inside each SPV. The RFQ is a qualification stage — it does not yet define the offtake. Whether projects are backed by bilateral corporate PPAs, a tariff-based mechanism, or merchant exposure will determine the discount rate and therefore the viable storage duration.

Third, how much of the 6 GW is actually storage. The committed BESS block is 1 GW / 4 GWh. The balance of the portfolio is solar, wind and "other renewable technologies" — a category that could absorb anything from more batteries to concentrated solar thermal to hybrid configurations. If awarded projects cluster at the four-hour mark, South Africa will have added 4 GWh of mid-duration storage by 2030 against a system whose evening ramp is considerably longer than four hours.

That gap is where the interesting engineering lives. The tender sets a floor, not a ceiling. The developers who win the subsequent RFPs will be the ones who can model, defend and finance a duration that is right for their site — rather than the four hours they were handed.

Sources