The sharper deterioration was in nuclear generation. After three months near 650 MW, average output fell to 523 MW, with a sustained run of negative reported readings at the end of August. Demand also fell, reaching the lowest August average in the bulk record.

August 2026 saw:

• EAF fall 4.6 percentage points, while remaining the strongest August since 2018
• Planned outages rise by 2.6 GW, offsetting another 0.5 GW improvement in unplanned outages
• Nuclear output average 523 MW, with 98 consecutive negative hourly readings late in the month
• Residual demand average 20,920 MW, down 6.6% year on year
• OCGT output peak at 2,647 MW, more than three times July's peak
• PV and CSP set August hourly records, although total renewable generation remained slightly below August 2025

Maintenance pulled EAF back down​

August's 71.0% EAF was 4.6 points below July, but 4.8 points above August 2025. The last stronger August in this record was 2018, at 73.7%.

Availability eased from July's high but remained ahead of the previous three Augusts shown here.

The outage breakdown explains the change. Planned losses rose from 4.09 GW to 6.69 GW, an increase of 2.60 GW. Unplanned losses moved the other way, falling from 7.41 GW to 6.90 GW. Other losses increased from 39 MW to 135 MW.

Together, outages averaged 13.73 GW, up from 11.55 GW in July. Planned maintenance accounted for more than the entire net increase; the improvement in unplanned outages offset part of it.

The planned-maintenance increase outweighed the further reduction in unplanned losses.

Unplanned losses were also 3.08 GW below August 2025. As a share of installed capacity, UCLF fell to 14.6%, from July's 15.7%. That matters when interpreting the lower EAF: August still showed progress in the unplanned-outage measure.

Nuclear generation dropped sharply at month-end​

Nuclear output averaged 523 MW, down 20% from July's 652 MW and 44% from August 2025. It was the lowest August nuclear average in the record, which starts in 2017.

The hourly detail is more revealing than the monthly average. Output stayed close to 650 MW through 19 August, then fell into a lower, more variable range. It briefly returned to around 650 MW on 26 August before dropping sharply again.

From 01:00 on 27 August through 02:00 on 31 August, the export reports 98 consecutive negative hourly readings, generally around −35 MW. Positive readings resumed at 03:00 on 31 August, and output reached about 640 MW that evening.

The late-month decline reduced the monthly average even though output had stayed near 650 MW for most of August.

Eskom explained the initial reduction in its 21 August update: marine material had built up at the seawater intake, causing a cooling pump to switch off automatically. Unit 1 was reduced to 50% power as a precaution, while Unit 2 was already offline for planned maintenance.

In its 31 August update, Eskom said Unit 1 had subsequently been taken offline following a turbine trip and returned to service at 03:00 that morning. That return time matches the first positive reading in the export. Eskom said nuclear safety was maintained throughout.

The chart and monthly average preserve the signed values as published. Nuclear output remained below its early-August level at month-end, despite the return to service.

Demand set another August low​

Average residual demand fell to 20,920 MW, 3.7% below July and 6.6% below August 2025. This was the lowest August in the 2017–2026 bulk record, and 14.4% below August 2020.

August demand has fallen from 26.1 GW in 2017 to 20.9 GW in 2026.

Lower grid demand gives the system more room to absorb lost generation. It does not itself increase EAF, which measures the availability of Eskom's installed fleet. Nor does this residual-demand series alone tell us how much of the decline comes from private generation, changing electricity use or other factors.

Diesel peaks returned​

Combined Eskom and IPP open-cycle gas turbine output averaged 152 MW, up from July's 116 MW but still slightly below August 2025's 160 MW.

The larger change was in peak use. OCGT output reached 2,647 MW at 19:00 on 3 August, compared with July's maximum of 790 MW. Another pair of hours exceeded 2 GW on 31 August. The ten largest OCGT readings of the month all fell between 18:00 and 20:00.

August's average remained modest, but the evening peaks returned to multi-gigawatt levels.

Interruptible load use remained limited: ILS peaked at 447 MW, below July's isolated 820 MW event, and averaged less than 1 MW. The export's manual load reduction field remained at zero throughout August.

Solar peaks improved; monthly renewable output barely changed​

PV reached 2,117 MW at 11:00 on 28 August, beating the previous August hourly record of 1,943 MW by 8.9%. CSP also set an August peak, reaching 539 MW at 16:00 on 23 August, compared with the previous high of 500 MW.

Those peaks did not translate into a record month for total renewable generation. PV's monthly average was 455 MW, down 12.2% year on year. Wind averaged 1,211 MW, up 2.8%, while CSP averaged 158 MW, up 12.9%.

Combined wind, PV, CSP and other renewable output averaged 1,863 MW. That recovered 13.4% from July's weak result but remained 0.9% below August 2025 and below the stronger August totals of 2022–2024.

New PV and CSP hourly peaks coexisted with a slightly lower combined monthly average than a year earlier.

Electricity trade also remained below its earlier export levels. Exports averaged 822 MW, up from July's 722 MW but still 54% below August 2025. Imports averaged 1,223 MW, leaving net imports of about 402 MW.

What August tells us​

The lower EAF needs to be read alongside the continued improvement in unplanned outages. Planned maintenance increased substantially, but August still delivered the best availability for that calendar month since 2018. That is a stronger result than the month-on-month headline alone suggests.

Nuclear generation is the clearest concern in this month's hourly data. Its contribution was already much lower than in the first quarter, and August brought a further late-month decline. Lower demand helped the system accommodate that loss, while diesel turbines supplied several large evening peaks. The rest of the fleet's availability improved year on year, but the nuclear recovery remained incomplete at month-end.

Method: monthly means calculated from the Eskom bulk export ESK19715.csv, merged with the historical base covering April 2017 onward. All 744 August hours have values for the metrics discussed. EAF is derived hourly as 100 × (1 − (PCLF + UCLF + OCLF) / installed Eskom capacity), then averaged; it is not Eskom's separately published official monthly EAF. Times follow the export's hour labels. Historical comparisons use the latest merged export, which can revise figures in earlier posts. “August records” refer only to August observations in this dataset. Charts and analysis by unofficialeskom.com.

The improvement came with two large qualifications. Nuclear generation remained stuck near 650 MW for a third full month, and average demand fell to the lowest July in the data. Solar also set a pair of seasonal records, although weak wind meant renewables as a whole did not break a record.

July 2026 saw:

• EAF jump to 75.6%, its highest monthly average since July 2018
• Total outages fall by 4.2 GW from June, to 11.6 GW
• Nuclear remain near 652 MW, after the station's output first dropped on 10 April and fell again on 25 April
• Demand fall to 21,726 MW, the lowest July on record
• PV and CSP set new July records, while weak wind pulled total renewable generation below July 2025
• OCGT output peak at only 790 MW, down from 2,322 MW in June

EAF reached its highest level since 2018​

The derived Energy Availability Factor rose from 66.7% in June to 75.6% in July, an 8.8-point improvement in one month. July 2025 averaged 63.7%, putting this July almost 12 points ahead year on year. The last stronger month in the bulk record was July 2018 at 78.0%.

Both planned and unplanned outages improved. Planned losses fell from 5.2 GW to 4.1 GW, while unplanned losses dropped from 10.4 GW to 7.4 GW. Other capability losses were negligible. Together, the three categories averaged 11.6 GW, down from 15.7 GW in June and their lowest monthly level since July 2018.

EAF rose sharply to 75.6% in July, its best monthly result since July 2018.

Total capability losses fell by 4.2 GW in one month as both planned and unplanned outages eased.

This is more than the normal winter maintenance wind-down. Planned maintenance did fall, as expected, but the larger contribution came from unplanned outages. UCLF moved from 22.0% of installed capacity in June to 15.7% in July, also its best result since July 2018.

Koeberg remained near 650 MW for a third full month​

The recovery in the rest of the fleet did not include Koeberg. Nuclear output averaged 652 MW in July, barely changed from 644 MW in June and 645 MW in May. January through March averaged roughly 1,875 MW.

The hourly bulk data shows two distinct steps in the decline. Nuclear generation was still near 1.8 GW on 9 April. It fell to roughly 900 MW on 10 April, consistent with losing one unit's output, then dropped again to about 620 MW on 25 April. Output recovered slightly into the 650 MW range during May but stayed there through the end of July.

By 31 July, the initial unit-sized loss had lasted 112 days. The lower-output plateau that began on 25 April had lasted 97 days. July was the third complete month with Koeberg producing only about one-third of the output seen in the first quarter.

Nuclear output collapsed in April and remained near 650 MW through May, June, and July.

The contrast with the fleet-wide numbers is striking: Eskom produced its best EAF since 2018 while roughly 1.2 GW of nuclear output remained missing. Coal availability had to improve enough to offset that loss.

Demand set another record low​

Average residual demand was 21,726 MW, the lowest July in the record. That is 7.1% below July 2025 and 13.8% below July 2020, when Covid restrictions were still affecting the economy.

July 2026 demand fell below every previous July, continuing the record-low monthly demand seen in May and June.

This makes the supply-side improvement easier to sustain. Lower demand does not create a higher EAF, because EAF measures fleet availability rather than electricity consumption, but it reduces the consequences of an outage and the need to run expensive peaking generation. July combined both: the fleet was genuinely more available, and it had less load to serve than in any previous July.

Diesel use fell despite the winter peak​

Open-cycle gas turbine output averaged 116 MW in July, down from 134 MW in June. The more surprising number was the monthly peak: only 790 MW, reached at 18:00 on 13 July. June peaked at 2,322 MW, and May at 2,879 MW.

July's OCGT peak stayed below 800 MW, a sharp break from the multi-gigawatt evening peaks in May and June.

Interruptible load shedding did briefly return. ILS reached 820 MW at 18:00 on 7 July, after peaking at zero in June. It was an isolated event: the next-highest July hour was only 67 MW, and the monthly average was 1.2 MW. Manual load reduction remained at zero.

International trade remained weak. Exports recovered slightly from 652 MW in June to 722 MW, but still sat below imports of 1,090 MW and were 60% lower than July 2025.

Solar set July records, but renewables overall did not​

The renewable-record hunch is correct for solar, with an important limit: these are July records, not all-time records.

PV output reached 2,051 MW at 13:00 on 24 July, beating the previous July high of 1,741 MW from 2025 by 18%. Concentrated solar power averaged 145 MW across the month, ahead of the previous July record of 131 MW in 2021. CSP also reached a new July hourly peak of 535 MW on 31 July.

PV set a new July hourly peak, while CSP recorded its highest July monthly average.

Total renewable output did not set a record. Wind averaged 1,030 MW, well below the 1,532 MW recorded in July 2025. Combined wind, PV, CSP, and other renewable generation averaged 1,642 MW, down 22% year on year. The new solar highs partly offset a poor wind month, but not enough to lift the combined total.

Outlook​

July is the strongest month in the recent recovery. A near-12-point year-on-year EAF improvement, 3 GW less unplanned capacity offline, and an OCGT peak below 800 MW are meaningful gains. Unlike some previous good months, this was not only planned maintenance moving around.

The low-demand caveat remains, and Koeberg is now the clearest unresolved weakness in the data. The grid carried winter comfortably while nuclear output stayed near 650 MW, but doing so required better performance from the rest of the fleet and another record-low month for demand. Solar's July records helped during daylight hours; weak wind and the persistent nuclear shortfall kept the generation picture less complete than the headline EAF suggests.

Data and charts from the Eskom bulk export, analyzed by unofficialeskom.com. Corrections and questions welcome.

June 2026 saw:

• EAF edge up to 66.8% (from 65.5% in May), and ~6.5 points better than June 2025
• Nuclear stuck at 645 MW — a second consecutive month with a Koeberg unit crippled, still unexplained
• Demand fall to 21,795 MW — the lowest June on record, below even Covid-2020
• Exports stuck at the floor (652 MW), still less than half of imports
• Demand-reduction tools stood down — ILS peaked at zero for the first time in months

Outages: the seasonal easing continues​

The Energy Availability Factor rose again, from May's 65.5% to 66.8% — the third straight month of improvement off April's 60.1% low. As in May, most of this is seasonal: planned maintenance (PCLF) fell from 12.7% to 11.0% as Eskom finishes prepping the fleet for winter peak. Total outages eased from 16,311 MW to 15,697 MW of capacity.

The year-on-year picture is the real good news. June 2025 sat at 60.3% EAF with unplanned outages (UCLF) at a brutal 31.2%; this June, UCLF is down to 22.0% — roughly 4.4 GW less unplanned breakage than a year ago. The fleet is meaningfully less broken than it was, even if it's still a long way from healthy.

EAF climbs for a third month to 66.8%, tracking the seasonal maintenance wind-down.

Total outages (black) eased to ~15.7 GW as planned maintenance (yellow) came off; unplanned (red) is well below its June-2025 level.

Koeberg still broken — now for a second month​

The one number that hasn't moved is the one Eskom won't discuss. Nuclear output averaged 644.5 MW in June — statistically identical to May's 645 MW, and roughly a third below the ~1,870 MW the station ran in January–March. This is now a second full month of a Koeberg unit running at heavily reduced capacity, the same fault first visible in April (1,129 MW), with still no public explanation.

A flagship nuclear station losing a unit's worth of output for two months running, silently, is not a footnote. It's ~1.2 GW of the cheapest baseload on the grid quietly missing while Eskom talks up its winter readiness.

Nuclear falls off a cliff in April and stays pinned at ~645 MW through May and June — one unit well below its rating, two months and counting.

OCGT: cheap on average, but the evening peaks are still expensive​

On a monthly average OCGT (diesel turbine) use looks trivial — 138 MW, flat on May and a third of June 2025's 394 MW. But the average buries the story. The monthly peak was 2,322 MW: down from May's 2,879 MW, but the second-highest of 2026 and well above April's 2,180 MW. Diesel-peak use has climbed steadily through autumn into winter (March 1.7 GW → April 2.2 GW → May 2.9 GW → June 2.3 GW), not fallen away.

That peak has a very specific shape. Averaged across the month, OCGT output is near-zero overnight and through the middle of the day, then jumps to a mean of 586 MW at 18:00 — the single busiest hour, more than 3× any daytime hour. The fleet is being fired almost entirely to get through the 6pm evening peak, when demand spikes and solar has dropped off. The monthly maximum of 2,322 MW landed at exactly 18:00 on 21 June, and every one of the ten biggest OCGT hours in June fell between 17:00 and 19:00.

It wasn't an everyday event: 12 of June's 30 days saw an evening OCGT burst above 500 MW, 7 days above 1 GW, and just 3 days above 2 GW — clustered in a cold snap around 21–23 June. So the picture is a fleet that coasts most of the month, then leans hard on the most expensive plant it owns for a handful of winter evenings.

Average OCGT use (red bars) stays minimal; the monthly peak (yellow line) eases from May but holds well above the summer months.

Every hour of June. Output is near-zero most of the time, spiking into sharp evening peaks — biggest during the 21–23 June cold snap, topping out at 2,322 MW at 6pm on the 21st.

Demand at a record low — again​

The recovery keeps being underwritten by shrinking demand. Average residual demand in June was 21,795 MW — the lowest June figure in the record, below June 2020 (24,718 MW) at the height of Covid, and down ~7.7% on June 2025 (23,606 MW).

June demand is normally higher than May's as winter heating load comes on, and it was (21,795 vs 21,249). But even with that seasonal lift, it still undercut every prior June on the board. Rooftop solar, industrial load curtailment and a flat economy keep eating grid demand — which is a big part of why the "surplus" looks as comfortable as it does.

June 2026 (red) sits below every prior year — the lowest June demand on record.

Exports still stuck at the floor​

The trade-side story from April and May didn't change: exports stayed depressed at 652 MW, less than half the 1,728 MW of June 2025 and still below imports (962 MW). The collapse in exports that first appeared in April has now held for three months straight — no longer a blip, but a settled lower level, and a sign that demand across the region remains weak.

Exports (orange) remain pinned near the bottom of their range and below imports (green) — the depressed level first seen in April held through June.

Demand-reduction tools stood down​

For the first time in months, Eskom barely touched its demand-reduction levers. Interruptible load (ILS) peaked at zero in June — down from 507 MW in May and 1,124 MW in June 2025 — and Instantaneous Operating reserves (IOS) peaked at just 116 MW. Manual load reduction (MLR) has been dormant since 2024. On the demand-reduction side, at least, June genuinely looks like a comfortable month.

Monthly peak demand reduction. ILS (teal) drops to zero for the first time in months; MLR (orange) remains near zero.

Outlook​

June continued the trend of the last two months: a slowly-healing fleet, a quiet winter so far, and demand-reduction tools that Eskom didn't need to pull. That's real progress on unplanned outages year-on-year, and the low diesel burn is a genuine positive.

But the caveats haven't gone anywhere. The improvement is still half a demand-collapse story — the lowest June demand ever recorded — and Koeberg is now two months into an unexplained reduced-output run. The grid is more comfortable than it was a year ago, but "comfortable because demand keeps falling and nobody's asking about the nuclear unit" is not the same as "fixed."

Data and charts from unofficialeskom.com. Corrections and questions welcome.

May 2026 saw:

• EAF recover to 65.8% (from 60.1% in April), but still below January–March
• Nuclear output get worse — averaging just 645 MW, with Koeberg running a unit at reduced capacity for a whole month
• A one-day OCGT peak of 2.9 GW, the highest in months — though average use stayed minimal
• Demand fall to 21,249 MW — the lowest May on record, below even Covid-2020

Outages eased — mostly seasonally​

The headline number recovered: the Energy Availability Factor rose from April's 60.1% to 65.8%, and total outages fell from 39.9% to 34.2% of fleet capacity. Most of that is the expected seasonal swing — planned maintenance (PCLF) was pulled back from 16.0% to 12.6% as Eskom clears the decks for winter peak demand. Unplanned outages (UCLF) also eased, from 23.7% to 21.4%, though they remain elevated.

Year-on-year it's genuinely better: May 2025 sat at 59.4% EAF, so the fleet is ~6 points more available than a year ago.

EAF recovered to 65.8% in May after April's dip, but remains below the 70%+ of January.

Total outages (black) eased back to ~15 GW as planned maintenance (yellow) wound down for winter; unplanned (red) remains elevated.

Koeberg still broken​

The one metric that got worse in May is the one Eskom has said least about. Nuclear output averaged just 645 MW for the month — down from 1,132 MW in April and 946 MW in May 2025. To be clear, this isn't a unit being taken fully offline: Koeberg has been running one unit at reduced capacity for the entire month, the same problem flagged in April, and Eskom still hasn't said a word about it. A whole month of a flagship nuclear station limping along with no public explanation is not normal.

Nuclear output drops off a cliff at the right edge and stays there through May — one unit running well below its ~900 MW rating, all month.

OCGT: a one-day spike, not a trend​

Average OCGT (diesel turbine) use stayed genuinely low for May at 137 MW. The headline is the monthly peak of 2,879 MW — the highest in months and up from April's 2,180 MW — but that was essentially a single day's burst, not a sustained ramp. It's worth flagging because it's the most expensive plant in the fleet and diesel is in short supply, but on its own one peak day isn't yet a winter trend; it's a reminder of how quickly the fleet leans on diesel when something slips.

Average OCGT use (red bars) is barely visible, while the monthly peak (yellow line) jumps to ~2.9 GW — driven by one day.

Demand at a record low​

The recovery is only possible because demand keeps falling. Average residual demand in May was 21,249 MW — the lowest May figure in the record, below even May 2020 (22,437 MW) when Covid shut the country down, and down ~8% on May 2025.

The "surplus" remains as much a demand-collapse story as a generation one: rooftop solar growth, industrial load curtailment, and a weak economy are doing a lot of the work that the generation fleet is being credited for.

May 2026 (red) sits below every prior year on the chart — the lowest May demand on record.

Exports stuck at the floor​

April's standout on the trade side was exports collapsing to the lowest level on record and falling below imports for the first time since May 2024 — a sign that neighbouring grids were feeling the strain before South Africa did. That didn't bounce back: exports stayed at those record-low levels for the whole of May, not just a bad week. The April collapse wasn't a one-off; it's now looking like the new normal.

Exports (orange) sit at the bottom of their range and below imports (green) — the depressed level first seen in April held through May.

Demand-reduction tools still in use​

Eskom kept pulling demand-reduction levers through May. Interruptible load (ILS) peaked around 507 MW — small, but not zero, and it keeps undercutting the claim of a comfortable power surplus. The monthly-peak view makes the contrast clear: manual load reduction (MLR), which spiked to multiple gigawatts during the worst loadshedding years, has been essentially dormant since 2024, while ILS and IOS tick along at a few hundred megawatts.

Monthly peak demand reduction. MLR (orange) — once multiple GW — is now near zero; ILS (teal) still peaks around 500 MW.

Outlook​

May was a step back in the right direction after April, but for the wrong reasons: the improvement is seasonal maintenance plus collapsing demand, not a healthier fleet. Koeberg has now run a unit at reduced capacity for a full month with no public explanation, exports are stuck at record lows, and demand keeps falling. None of that is a healthy "surplus" — and the gap between Eskom's public confidence and what the data shows is still there.

Data and charts from unofficialeskom.com. Corrections and questions welcome.

April 2026 saw:

• Worst EAF since May 2025
• Highest unplanned outages since July 2025
• Highest total outages since June 2025

Capability loss factors (top) show total outages climbing back to 18.8 GW, while the Energy Availability Factor (bottom) dropped to 60.15% — the worst since May 2025.

Koeberg nuclear troubles​

One unit at Koeberg shut down about a week before it was planned to go off for maintenance. The second unit is currently operating at 600 MW instead of 900 MW. Eskom has said nothing about either issue.

Nuclear generation (top) shows the sharp drop in output. OCGT generation (bottom) is climbing again, with maximum use hitting 2.6 GW.

OCGT reliance returns​

Open Cycle Gas Turbines (OCGTs) are being used to generate up to 2.6 GW again. The average is still okay, but things are clearly breaking. With diesel in short supply, it is not a great time to be dependent on OCGTs.

Imports and exports​

Exports dropped to the lowest level on record, falling below imports for the first time since May 2024. Neighbouring countries appear to be suffering as a result before South Africa does.

International exports (orange) dropped below imports (green) for the first time since May 2024.

Interruptible load shedding​

Eskom is also starting to use the interruptible load shed lever again to reduce demand when needed. This is still small compared to even January, but it undermines the claim that South Africa has a power surplus.

Interruptible load shedding is ticking up again after months of near-zero use.

The "surplus" narrative​

Despite these issues, Eskom continues to push out statements about how good things are.

An FT article from 2 May 2026 quotes Eskom chief executive Dan Marokane claiming South Africa now has a power surplus.

Demand collapse masks the problem​

For now, South Africa is still very much relying on historically low demand — almost exactly the same as April 2020 when everything shut down for Covid.

Demand (top) and generation (bottom) for April 2026 are at levels comparable to April 2020, driven by rooftop solar growth, industrial load curtailment, and possibly economic weakness.

Sonnet 4.6 analysis​

Here's an AI analysis from Sonnet 4.6 also focusing on the nuclear wobbles and the demand collapse:

Sonnet 4.6 identified two distinct Koeberg failure events in April 2026 and noted that the claimed "surplus" is largely a demand collapse rather than a generation triumph — average demand fell from 24,226 MW in April 2024 to 21,686 MW in April 2026, a drop of 2,540 MW (-10.5%).

Outlook​

Hopefully this is just a temporary setback before getting back on the improvement train again. But with diesel in short supply, rising dependence on OCGTs, and a lack of transparency around Koeberg's status, the gap between Eskom's public statements and the underlying data is widening.

Data and charts from unofficialeskom.com. Corrections and questions welcome.

Small reversal in the headline metrics - unplanned losses up a bit, EAF down a bit to below 70% again, but nothing too concerning here

More use of the emergency generators too but it seems this was because Eskom is under 'use it or lose it' contracts so they choose to burn private diesel at higher cost rather than cheaper coal as they would have to pay for the diesel whether or not they used it. Weird world.

More interesting, IOS and ILS are both at nearly zero for the first time since like 2018?

I always get them mixed up so here's what they mean(?)

ChatGPT battles with the distinction too but I think broadly ILS is for big customers (smelters?) who Eskom can call and ask them to reduce and IOS is more broad and includes transmission faults etc too.

Otherwise not the most interesting month. Demand and generation continue with the pattern of a big year-on-year reduction. Still to be seen how much autumn/winter changes that.

See this data and more on my site unofficialeskom.com and feel free to shout with any questions or requests.

We skipped December as Eskom stopped publishing data in November and didn't fix it for several weeks.

Starting with a highlight: best EAF in January since June 2020!

Lowest use of emergency diesel on record? I have one lower value in Nov 2020 but I think that's a data glitch. Last month we never generated more than 411MW with the OCGTs.

demand and generation still at very record lows though, so we're still playing on easy mode in terms of high availability factor and low backup generator use. I'm interested to see what happens in July when the private solar drops off.

also nice to see Koeberg performing at near full capacity for the last 3 months. A small dip in jan from a brief outage but seems to have been quickly resolved.

Hopefully they figure out the long-term plan for the waste in one of the coming decades..

Otherwise not much to say. Things look OK for now. Yay for getting back into the green zone on my dial for the first time since I added it.

This data and more at my site unofficialeskom.com - feel free to tag me with any questions.

End of Diesel? According to the data, we never generated more than 350MW with the OCGT backup generators, which is the lowest I have on record except for nov 2020 which might also be a glitch.

New installed renewable record of 7.5GW - looks like all those announced installations are slowly coming online. I'm not sure exactly when they count as 'installed' from.

Thermal is dropping, but that's a usual trend for this time of year. Hopefully it starts increasing from Jan->July again if we can keep squeezing the coal plants.

Koeberg second unit is back, so we have 900MW of extra baseload to play with which is nice (and maybe related to not needing the OCGTs?)

Interruption of Supply and Manual Load Reduction looking good. I always forget exactly what these are an how they relate with each other

IOS – Interruption of Supply: all contracted as well as mandatory demand reduction resources utilised by Eskom National Control. Includes IOS due to Transmission network faults.

MLR – Manual Load Reduction. An estimation of demand that has been reduced due to load shedding and/or curtailment.

so super clear? Anyway, nice that they're low.

wait there are three

ILS – Interruptible Load Shed. This is consumer load(s) that can be contractually interrupted without notice or reduced by remote control or on instruction from Eskom National Control. Individual contracts place limitations on usage.

still playing on easy mode because there's no demand, so let's see how that changes when the sun goes away and private solar turns off.

All data from my site unofficialeskom.com - feel free to lmk any questions or feedback

Not anything too interesting this month - EAF down again due to an increase in unplanned outages, but still 'ok' overall.

Using those diesel generators a bit more again as a result

Lowest imports, maybe ever?

The big drop in supply and demand from last month was not a one-off.

not visible in the monthly graphs yet, but the other Koeberg unit is busy powering up again. AFAIK this is the end of the cycling/swopping of the two units for refurbishment and refuelling, so we should have both running simultaneously for the first time in years.

I learned that we store all high-level waste on the Koeberg site itself, and that it's full, so we're buying expensive 'dry storage casks' from the US in order to store more, also on-site, but those are also full now?

www.businessday.co.za/bd/national/...

Maybe we'll build something by 2030

As always, more graphs on my site unofficialeskom.com and feel free to ask me any questions and I'll try answer.

Another dip in unplanned outages takes us close to 2020 levels!

I think the best EAF in september since 2018, top of the chart for my curve graphs as they don't go back that far.

We're hardly using the OCGT backup generators any more, with very low average and max use (relatively speaking).

Very low imports too. Is everything down.

Massive demand drop.

I need to go double check that I'm not missing data that causing this, because it looks sus.

also weird, IOS (interruption of supply, a mechanism that lets Eskom cut off business users) went up a bit compared to last month.

but ILS (interruptible load shed) went down 🤔

I'm not sure I totally understand each of these, but maybe it's because iOS includes transmission network faults.

The transmission network seems to be the bottleneck now instead of generation which everyone has been focusing on, as many people predicted?

As always all data and graphs from unofficialeskom.com. Feel free to let me know any questions or answers.

I finally got a reply to my data request for August data (usually takes a few hours / days, this time several weeks)

Metrics looking good compared to the last few months, but not that far off where we were a year ago.

Very nice dip in unplanned outages.

EAF is up to best since July last year where it was slightly higher, but slightly better ratio of planned:unplanned since then so still a win.

similar for diesel use, big improvement for the last few months, but similar to last winter.

I'm guessing that another bigger than (I) expected dip in demand is an important factor here. Generation is only slightly up year on year, but demand is way down.

Overall Eskom deserves some of the crowing they've been doing recently (unplanned outages down = good), but in context things seem to be

a) similar to a year ago b) a lot less pressure due to reduced demand

So I'm maintaining my cautiously optimistic stance about future stability.

Sorry I missed June, I was travelling.

July is looking positive! Big decrease in total loss, and this time it's in the good way (decrease in unplanned, increase in planned)

Diesel use also looking better - lowest since November 2024, but still twice as high as July 2024.

And we're still using these things up to 3GW at peaks, so the max is still very high.

Thermal (coal) is looking really flat, so I'm not sure where all that extra availability is coming from. Maybe we have coal plants that are working but switched off? Not sure exactly how they calculate that re planned maintenance.

Nuclear is also flat so it's not that

Kinda interesting pattern in Demand vs Generation here - demand is still dropping, so either SA is having a sunny winter or private solar isn't as big a factor as I hypothesized before in reducing demand (maybe the economy is just that dead?)

Hopefully we can get back at least to 2024 levels soon

Overall I'm surprised that things held together pretty well for July, usually the worst of winter I think. Koeberg unit 2 return got delayed again (not that unexpected, but was meant ot be July) and Medupi is only scheduled to get to full commercial operation in September.

I'm still 'cautiously optimistic about progress with a side dose of something could blow up at any time'

• Unplanned outages still looking bad
• Small bump from new wind installations
• Diesel use trending down again

🧵

Planned maintenance is dropping (as expected as we head into colder months), which increases EAF, but not in the good way - unplanned outages still at 13.4GW - slightly higher than last month.

Thermal generation is up a bit, slightly higher than a year ago, and best so far this year, but looking a bit doubtful we will get back to a regularly >22GW peak from last winter.

New wind installations are helping a bit - highest peak of 3.2GW (from an installed base of 3.7GW), and ~200MW more on average than this time last year is nice to see

Still using those diesel generators for generating up to 3GW during peaks, but on average looking better than the last two months.

Exports up a bit - higher than May 2024, but still far below the November peak

Overall generation and demand is flat and the same as last year, so take your pick on whether the glass is currently half full or half empty I guess 🙃

As always - this data and more available on my site, unofficialeskom.com.

Ping me with any feedback or questions.

Note: the sqlite databases that all the metabase queries are based on are now available for direct download at unofficialeskom.com/dashboards/#.... I still have a TODO to create a data dictionary and explain the columns etc, but it's pretty straight forward.

🧵

EAF is not looking good with total loss at the highest level since March 2024.

Silver lining is that April is actually probably better last month - reduction in unplanned outages and an increase in planned outages, leads to higher total outages, but still better than more unplanned outages.

coal slightly down again.

Last month (bsky.app/profile/sixh...) I wrote that we installed a small amount of new wind capacity for the first time in a long time.

The newer data that they released shows it coming online in April instead of March

I wasted time coming through the CSV data downloads to make sure it wasn't a bug

They've taken their foot off the diesel a bit, but we're still using a lot. Peaking at 3GW (more than 3 Koebergs) and averaging 600 MW (close to one unit of Kusile, 800 MW x 6 units, en.wikipedia.org/wiki/Kusile_...).

I also added Interruptible Load Shed as a tracked graph for the first time after @c4talyst.bsky.social mentioned it was peaking.

This is a weird mechanism where Eskom can create virtual capacity by shutting off demand separate from loadshedding.

I haven't figured out what the numbers mean yet.

The other graphs are averages of the data I get from Eskom, but ILS seems very short lived, so all the 0s created weirdly small numbers (averaged by hour). This graph shows the total instead but that also doesn't seem right.

Whatever it is, they're using more of it than they have in a while.

I'm also tracking IOS (interruption of supply) now, which is similar but where they can turn off specific big business customers if they need. They haven't been squeezing more out of this than usual recently though.

Exports are continuing to decline from their weird peak a few months ago

Our hydro plant is working again. I don't understand this cycle either, but it seems to vary quite a lot. Either way, it's not much power so not that important, but I'm curious what leads to these swings.

See unofficialeskom.com for more and let me know any questions or comments.

The update this month was late because I didn't get the normal file after asking for it. Shout out to Yegz for sharing it with me. Maybe they blocked my email or something

There have been a lot of articles and press releases in the last few days about Eskom's diesel use and the upcoming elections.

One one side there is the DA, ex CEO De Ruyter, and some others claiming broadly that

• The situation at Eskom is as bad as ever
• Our record-setting 50 days without loadshedding is due to excessive diesel use
• After the elections, we'll have used up the diesel budget and things will go back to being bad
• This is due to political pressure that the ANC puts on Eskom

On the other is current Eskom leadership and the ANC who claim that

• Eskom has "turned the corner" and things are long-term improving
• Diesel is being burned responsibly and at a slower pace than before

I think the truth is somewhere in between, but I'm just some guy on the internet with an unhealthy obsession at trawling through data and news about Eskom, so take everything below with a healthy dose of skepticism, and only quote the bits that support whichever side of the debate you were already on.

Articles and claims​

https://businesstech.co.za/news/energy/771174/the-truth-behind-no-load-shedding-according-to-andre-de-ruyter-2/

• He mentioned that during his tenure at Eskom, he and former COO Jan Oberholzer only had a budget of approximately R6 billion per year to spend on diesel to fuel the utility's open-cycle gas turbines to generate electricity.
• the current budget for diesel this year under new management is R24 billion, which is four times more than what they had at their disposal during his time.
• Public Enterprises Minister Pravin Gordhan recently revealed that Eskom spent R65 billion on diesel over the past five years to fuel its open-cycle gas turbines (OCGTs).

https://www.da.org.za/2024/05/confirmed-anc-government-is-putting-political-pressure-on-eskom-to-burn-more-diesel-da-wants-answers

• According to NERSA, Eskom has already blown more than half of its quarterly budget of diesel for the first quarter of the 2024/2025 financial year in just one month.
• NERSA has confirmed beyond any doubt that the use of diesel-powered OCGT's was ramped up during the April-June 2024 financial quarter to coincides with the national elections on 29 May.

https://www.moneyweb.co.za/news/south-africa/eskom-nersa-set-the-facts-straight-on-diesel-usage/

• Unplanned outages have been reduced by 4 400MW since 26 April 2024 due to extensive maintenance and the success of the Generation Operational Recovery Plan initiated in March 2023.

Some concepts​

Part of what makes the discussion confusing and convoluted is that everyone quotes different metrics. A simplified version is that Eskom needs diesel to run emergency generators. These were originally designed to be run for very short 'peak' periods where demand exceeded supply at 6pm when everyone came home from work, or when a big coal plant had to be shut down for emergency repairs. Now they are run a lot more than that as demand has exceeded supply for many years.

But this diesel needs to be

1. Budgeted for
2. Bought
3. Used

And this leaves a lot of room for pointing fingers with statements like "Eskom's diesel budget is twice as big as last year" while leaving out inconvenient facts like how much was actually bought (budgets can be overrun) or used.

The main metric that Eskom tracks is "EAF" or energy availability factor, which is a percentage of what percentage of the power generation fleet is currently operational. Big number = good. Small number = bad. But again, it's more complicated because the inverse of the EAF (generators that are not operational) is broken into planned maintenance (good) and unplanned maintenance (bad), so it's possible to increase the EAF (good) by reducing the planned maintenacne (bad).

In reality we want a world where we have

• High EAF
• High planned maintenance (means we are not sacrificing future performance of the plants to make them work now)
• Low unplanned maintenance (things aren't breaking unexpectedly)
• Low diesel usage (diesel is expensive, and bad for the environment)

Further complications​

To make things even more complicated, reporters talking about this stuff are well known for getting facts and numbers very wrong. For example from a Moneyweb article.

Eskom's budget for diesel in the current financial year (April to June 2024) is R5.8 billion, and R1.16 billion has been spent as of 09 May 2024 (19.7% of the total budget). [emphasis added]

And Eskom doesn't really publish its actual diesel budget or purchases in a neat way (at least one that I've found) in their annual reports, so we have to rely on media statements and long NERSA documents to try to figure those out.

Looking at the OCGT usage data​

The easiest data to look at is the OCGT use. This includes all the emergency power generators that Eskom controls, including its own diesel generators, and those owned by independent power producers like Peakers.

This is daily data, aggregated monthly, so you can see we regularly use these to generate up to 500MW of power and can go up to just over 3GW if necessary (for context, our only Nuclear power station Koeberg generates a maximum of 1.8GW of power, and Kusile our in-progress coal power plant and fourth largest coal power plant in the world if it ever gets completed does 4.8GW).

The highlighted blocks show February, March, and April months, with the arrows showing pre-election months (South African general elections in May 2019 and May 2024). Looking at the 2019 period, it's easy to see where this idea of "we are burning diesel at unsustainable rates to keep the lights on and keep voters happy in the run-up months to the election" came from. Generally we'd expect to need emergency power more in winter, when demand is higher, but in reality we need emergency power when the coal power plants break down.

Whatever is the case for the increased usage in early 2019, the data doesn't really support the theory of using diesel to support this election. We had no loadshedding throughout the whole of April, and yet OCGT use was significantly lower than March and also below April last year. So it's not as simple as "more diesel, less loadshedding".

What about planned maintenance​

While everyone is focused on the diesel use, a 'sneakier' way that Eskom could reduce loadshedding if instructed by the powers that be is to reduce planned maintenance. Remember from earlier that we can increase EAF by reducing unplanned or planned maintenance, but we don't want to reduce planned.

In the 'retaliation' statements, Eskom has been quick to point out its progress in reducing unplanned maintenance, but has failed to mention its reduction in planned maintenance too.

If you look at the black line coming down in the last four months, you'll see it matches the yellow line coming down (bad) far better than it matches the red line coming down (good). The downward trend of the red line in the last 12 months is the strongest argument for a bit of optimism.

Interestingly enough for the conspiracy theorists, a similar pattern was seen before the 2019 elections, though not in the February-April period in any of the in between years.

On the other hand, the current planned maintenance is still not low. Generally most maintenance is done in December when demand is lower, and then ramps down in winter, so I would put this one down as something to watch, but not strong evidence that this is due to political pressure related to the elections.

Summary of claims​

• Eskom's diesel budget is 4x higher than before. Hard to verify. Probably it is true or close to true but also irrelevant as the previous budget was simply overrun, and diesel use was similar
• Eskom is burning diesel at a rate of knots. Weird way to measure, but probably true. Also irrelevant as this has been the case for years and not much has changed recently.
• Eskom has turned a corner and reduced unplanned breakdowns. Partly true, but lying by omission if you don't also look at the reduction in planned maintenance.
• Eskom is under political pressure to reduce loadshedding. Hard to say. I guess we'll see what happens in June.

Summary of my impression of the actors​

• Eskom: largely incompetent still as a whole, but there seems to be some smaller parts that are trying to improve things. Their media people seem to make the fewest outright false claims, although they do highlight certain parts that make them look good while ignoring some uncomfortable facts.

• De Ruyter: looking for media attention (good for his book sales?) and spouting nonsense without any regard to facts or data. Always happy to claim credit for anything good and deny responsibility for anything bad.

• DA: similar to De Ruyter, ready to jump on anything that riles up the conspiracy theorists.

• NERSA: probably the most competent of the lot, but also the quietest. Most of the 'NERSA said' claims turned out to be false.

• Conspiracy theorists: always happy to have a logical answer to the chaos that they live in, whether or not the data supports that answer.

It's controversial but probably because of state capture and general corruption in South Africa and especially at South Africa's State Owned Enterprises.

Because Eskom failed to build new generation capacity, it ran its aging power plants far harder than it should have. (Eskom benchmarks the performance of its coal-fired power stations against those of the members of Vereinigung der Großkesselbesitzer, a European-based technical association for electricity and heat generation industries.)

Here's a graph showing the utilization factor of Eskom's plants compared to VGB benchmarks.

Why do you trust the data?​

I don't particularly. It is erratically updated, and often unavailable. My current email thread with mediadesk@eskom runs to hundreds of emails. Each time I notice something wrong with the data (out of date, broken format, missing, unlikely) I ask them. Sometimes they respond. Sometimes I follow up a few times and then they respond. The shorter term data that they usually update every day (the graphs and CSVs on their 'data portal') often doesn't perfectly match the longer term data that you can request though their data request form. The data in the PowerBI charts often does not match the data in the CSV exports that should be hard-linked to those charts. Shrug emoji.

But from what I've seen, this can all be ascribed to incompetence, not malice. I've heard every conspiracy theory imagineable, including but not limited to

• ANC controls eskom and creates loadshedding for political gain
• ANC controls eskom and stops loadshedding for political gain
• Corrupt contractors control eskom and sabotage power plants to milk maintenance contracts (I think Eskom supports this directly so probably not a conspiracy theory)
• Corrupt IPPs control eskom and sabotage coal plants in order to sell more OCGT (diesel) energy at high rates

I'm sure some mix of these are true, but I don't believe that Eskom is a cohesive and well-run organization. This means that it's as hard for any evil person to gain control over it as it is for anyone competent and well meaning.

The data published through the Eskom.com website is barely ever referred to or used in any Eskom media statements or government media statements. It seems to be run by some smaller function within Eskom and to have at least some people who care enough to fix their CSV files when they break and are bothered about it, but not care enough to actually make it reliable or easy to understand (hence this website). Often individuals from Eskom, from the government, or related to either makes statements that contradicts the data.

I think most people don't know that the 'eskom data portal' exists, and most don't care.

So tl;dr I'm open to any new information or well-reasoned arguments anyone wants to share about why they think this data is bullshit and why I'm wasting my time collecting and analysing it, but overall I will apply Occam's Razor and Hanlon's Razor, and assume that the published data at least broadly gives some kind of indication of what is happening inside Eskom.