Australia’s electricity debate is usually conducted in percentages: renewable penetration, emissions reductions, storage targets and dates by which coal should disappear. Those numbers are useful, but they do not give much intuition for the physical system underneath them.

So I tried a different approach. I asked a simple question:

If South Australia had to cover its electricity demand with solar, its existing wind fleet and batteries, what would we actually need to build—and what might it cost?

This was never intended as a forecast or a proposed electricity plan. It was a thought experiment designed to expose the scale of the problem. The point was not to prove that renewables “work” or “don’t work.” It was to understand what changes as we progressively replace convenient assumptions with more demanding ones.

We began with an ordinary day. How much electricity does South Australia consume? How much of it is consumed after the sun goes down? How much extra solar would be required to meet daytime demand while also charging a battery for the night?

That produced a manageable-looking answer—but an ordinary day is the easy problem.

We then replaced flat demand with a realistic daily demand curve. We replayed actual solar and wind conditions from 2024 and 2025. We extended that to ten years of weather, from 2016 through 2025, while holding demand and today’s generating fleet constant. That allowed us to ask what the present system would have experienced under ten different years of weather.

The result for South Australia, population 1.9M, rose to approximately 15.6 GW of grid solar and 5.1 GW/54 GWh of batteries, costing about A$45.5 billion in incremental capital. That system survived every half-hour in our ten historical weather years—but that is still not the same thing as guaranteeing future reliability.

We then repeated the process for Victoria, population 7.2M, whose demand is much larger and whose winter conditions are less forgiving. Its ten-year no-gas result was approximately 53 GW of grid solar and 383 GWh of battery storage, at about A$213 billion.

The next step was to stop treating the states as islands.

We combined South Australia, Victoria and NSW, combined population 18M, allowing them to exchange electricity through Heywood, Project EnergyConnect and the major NSW–Victoria links. We used their end-2025 wind, grid-solar and battery fleets, fixed 2025 operational demand, and the same ten weather years.

The first three-state no-gas result required approximately:

• 131 GW of additional grid solar

• 42.5 GW of battery power

• 1,020 GWh of battery energy

• A$556 billion of incremental capital

Annualised over the lives of the assets, including financing, battery augmentation and operating allowances, that was about A$47 billion a year—roughly 17% of the three state governments’ combined annual expenditure.

This is where the exercise became especially interesting.

Tripling the major cross-border interconnectors barely changed the answer. The modeled generation-and-storage bill fell from A$556 billion to about A$538 billion, before paying for the additional transmission. Interconnectors can move electricity, but they cannot create it. When several states experience the same broad shortage, a larger wire has little surplus electricity to carry.

Tripling the wind fleet made a much greater difference.

That scenario added 22 GW of wind, taking the three states to 33 GW of wind nameplate, and still required approximately 85 GW of additional solar and 628 GWh of battery storage. Including the cost of constructing the additional wind farms, the total fell to approximately A$448 billion, or A$38 billion a year.

To put the physical scale in perspective, the resulting system had about 126 GW of wind and solar nameplate capacity serving average demand of only 13.7 GW and a coincident peak of approximately 23.3 GW. Renewable nameplate was therefore more than nine times average demand and more than five times peak demand.

That is not evidence of a simple calculation error. Solar is unavailable every night, wind output varies, batteries lose energy while charging and discharging, and the system must survive prolonged periods when renewable production is poor. Much of the generation produced during favorable weather must therefore be curtailed.

Finally, we relaxed the most severe assumption and allowed the existing gas fleet to operate for a limited number of hours.

This produced perhaps the most important intuition of the entire exercise. The four cases below show the capital staircase. Tripling the major interconnectors alone saved only A$17 billion—and that is before paying for the extra wires. Tripling wind as well saved a further A$90 billion, including the wind construction. Then, in that same tripled-wind system, allowing freely dispatched gas up to 0.25% of annual demand reduced required capital by another A$70 billion, to A$378 billion.

Four cases in the three-state thought experiment. Tripled-interconnector construction is unpriced; the additional wind fleet is included.

The gas allowance was 0.25% of annual demand, but the optimized system actually used gas for only about 0.06%—roughly 73 GWh a year. Gas was valuable because it covered the rare, controlling intervals that otherwise set the size of the entire renewable and storage fleet. The modeled fuel-and-variable operating cost was about A$11 million a year: tiny beside the A$70 billion of avoided capital, although keeping gas plants available has a fixed cost that this thought experiment does not price.

None of these figures should be read as a detailed plan for the National Electricity Market. The model simplifies transmission, ignores many local network constraints, assumes perfect coordination between states and does not constitute a formal reliability study. Some regional construction results—particularly the enormous concentration of new solar in NSW—are just optimizer artifacts rather than realistic siting proposals.

But the exercise does provide something that is often missing from the public discussion: a sense of scale.

It shows why an ordinary night is not the real storage problem; why ten years of weather matter more than a representative day; why transmission is valuable but cannot cure a shared energy shortage; why additional wind can be more valuable than still more solar; and why eliminating the final fraction of dispatchable generation can be vastly more expensive than eliminating most of it.

For anyone who wants to inspect the machinery, I include the detailed report.

Scienceofdoom Sa Vic Nsw Network Extensions And Gas Report

1.45MB ∙ PDF file

Download

Download

In a few subsequent articles I tried to tease out some others trends and didn’t really succeed.

There were a two main points in Klotzbach’s paper that made me want to dig:

1. We worked from 1990 because the global satellites prior to 1990 didn’t have proper visibility of the Indian Ocean

2. Observing changes between 1980-1990 made trend data before 1990 challenging

When someone screens out data it makes me want to dig a little deeper. Is that screening really justified? If we include that data what do we find?

My question was simple:

Are there more very intense tropical cyclones now than in the 1980s?

By “very intense” I mostly mean Category 4 and 5 storms. Cat 5s are rare, so Cat 4+ is a better metric if we want enough storms to say anything useful.

Here’s the plain English version:

Weak, short-lived tropical cyclones are a major problem for trend analysis because modern instruments are much better at finding them.

Cat 4+ storms are different. They were not invisible in 1980. But their exact intensity estimates still changed with satellite capability, the Dvorak technique, agency practice, and aircraft reconnaissance.

If we exclude the Indian Ocean and start in 1980, Cat 4+ storms appear to increase.

But if we allow a step change around 1990, the continuing trend disappears.

That last point is the main finding.

The key distinction is this: weak storms have a detection problem; Cat 4+ storms have an intensity-classification problem.

A weak, short-lived storm might not have been noticed or named in 1980. A Cat 4+ storm almost certainly was noticed. But the question is whether its peak wind would have been analysed the same way: 105 kt, 110 kt, 115 kt, 125 kt? That matters because Cat 4 begins at about 115 kt in the best-track data. A storm does not need to vanish from the record to change the trend; it only needs to move across a category boundary.

Klotzbach’s 1990 boundary

Klotzbach et al. used 1990–2021. Their reason was that the global observing system before 1990 is less clean.

They write that before 1990, global TC intensity estimates are likely compromised by data issues. In particular, the North and South Indian Oceans lacked direct geostationary satellite data until Meteosat-5 in 1989. They also note a possible artificial jump in eastern North Pacific intensities after forecast responsibility shifted to NHC following the 1987 season.

That made me wonder:

What if the Indian Ocean is the main problem?
What if we simply remove the Indian Ocean and start in 1980?

This seems reasonable at first. The Indian Ocean is not the largest contributor to global TC activity. If its pre-1990 record is the problem, perhaps excluding it gives us a longer usable record.

Here’s the graph from 1980, by category of TC, excluding the Indian Ocean:

Klotzbach et al. found that short-lived named storms increased significantly from 1990–2021. They explicitly say this is likely due to improved sensors. Over the same period, global hurricane counts and ACE decreased, mainly from lower western North Pacific activity.

The key distinction is this:

For weak storms, the question is often: did we even know the storm existed?

For strong storms, the question is: did we estimate its peak intensity the same way?

Those are different problems.

A detour into ADT-HURSAT

I spent some time looking at ADT-HURSAT because it looks like the kind of dataset we might want.

IBTrACS is best track. ADT-HURSAT is more like a homogeneous satellite-only intensity reconstruction. That sounds attractive. If the issue is changing agency practice, maybe ADT-HURSAT solves the problem.

Mission failed.

That’s a little unfair. ADT-HURSAT is useful for some questions. But it is not useful for the question I had in mind: absolute counts of Cat 1, Cat 2, Cat 3, Cat 4, and Cat 5 storms comparable to IBTrACS.

HURSAT is a storm-centered satellite dataset. ADT-HURSAT applies the Advanced Dvorak Technique to HURSAT infrared imagery. Importantly, the HURSAT history files use IBTrACS to initially identify storms. So ADT-HURSAT is not independently finding missing historical storms. It is estimating intensities for storms already in the storm-centered archive. NOAA’s documentation also says that ADT-HURSAT should not be used to determine actual storm intensities, because individual peaks and valleys are not accurately represented with the reduced spatiotemporal sampling.

In my comparison, ADT-HURSAT produced far fewer Cat 1–5 storms than IBTrACS.

The explanation is probably simple. IBTrACS best tracks can use satellite imagery, Dvorak analysis, aircraft reconnaissance, scatterometers, microwave data, radar, buoys, ships, and post-storm judgment. ADT-HURSAT is a homogeneous satellite-IR algorithm. It is more consistent, but it throws away a lot of information. It also smooths and undersamples short-lived peaks.

So I discarded ADT-HURSAT for this article. Not because it is bad, but because it is not the right tool for this question.

The ≤2-day storm problem

Klotzbach et al. split named storms into storms lasting ≤2 days and >2 days. This is a very useful test.

A short-lived weak storm is exactly the kind of thing that modern instruments help detect. Before scatterometers and better microwave data, an analyst might see a cloud swirl over the ocean but not have enough evidence for a named tropical storm. With modern instruments, one good pass can show a closed circulation and tropical-storm-force winds.

QuikSCAT launched in 1999 and provided daily ocean surface wind coverage over most ice-free oceans for more than a decade. ASCAT on Metop-A launched in 2006 and continued the ocean surface wind vector capability. These instruments do not measure wind directly in the way an anemometer does; they infer near-surface winds from radar backscatter off the roughened ocean surface. But they are incredibly useful for detecting 34-kt winds in weak and marginal systems.

This does not mean those storms are fake. It means similar storms in the past may not have made it into the archive.

Landsea et al. made this point for the Atlantic. They found that the increase in Atlantic TC counts was dominated by very short-lived storms, while medium- to long-lived storms increased little, if at all. They also note that recent very short-lived systems entered the record partly because of new tools such as QuikSCAT, ASCAT, AMSU, ADT, GPS dropsondes, buoys, and cyclone-phase-space analysis.

So far, so good.

But this short-lived-storm problem is mostly a weak-storm problem. Here’s the “2 day and above” graph:

And here’s the “less than 2 day TC” graph:

IBTrACS category counts for storms with hurricane-strength duration ≤2 days, excluding the Indian Ocean. Short duration matters greatly for Cat 1 and Cat 2, somewhat for Cat 3, and hardly at all for Cat 4+. Cat 5 is essentially absent.

This graph shows what we would expect. The ≤2-day category is dominated by Cat 1 storms, with some Cat 2 and a few Cat 3. Cat 4 is rare. Cat 5 is basically absent.

That means we should not use the ≤2-day exclusion to dismiss Cat 4+ counts. Cat 4+ storms were not invisible. A Category 4 or Category 5 tropical cyclone in 1982 was going to be seen.

But that doesn’t mean the Cat 4+ record is problem-free.

The strong-storm problem is different

For Cat 4+ storms, the issue is not whether the storm existed. The issue is the assigned maximum wind speed.

A storm estimated at 105 kt is Cat 3.
A storm estimated at 115 kt is Cat 4.

Same storm, different category.

Most non-Atlantic basin intensity estimates depend heavily on satellite-based Dvorak analysis. The Dvorak technique is extremely important, but it does not directly measure wind or pressure. It estimates intensity from satellite-observed cloud patterns, eye temperature, cloud-top temperature, and related features. Velden et al. note that Dvorak estimates are often good, but they are not perfect: in one Atlantic validation, half the maximum-wind estimates were within 5 kt, 75% within 12 kt, and 90% within 18 kt of reconnaissance-aided best tracks. That is easily enough uncertainty to move a storm across the Cat 3/Cat 4 boundary.

This is especially important around 1990.

Klotzbach and Landsea reviewed the old claim that global Cat 4–5 storms had roughly doubled from 1970–2004. They argued that much of that increase was due to observational improvements in the 1970s and 1980s. Satellite imagery improved dramatically. The Dvorak technique became more widely used. The eastern Pacific had an agency transition after 1987. Western North Pacific aircraft reconnaissance ended in 1987. And there was no standardized global intensity database.

Their recommendation was very direct: global studies of extreme hurricanes should begin around 1990 because before that the record is incomplete and can give a distorted view of activity.

This is the context for the next graph.

Cat 4+ storms excluding the Indian Ocean

First, I removed the North and South Indian Ocean basins. This was the test:

If the Indian Ocean is the big pre-1990 problem, what happens if we take it out and start in 1980?

IBTrACS Cat4+ annual counts excluding the Indian Ocean. Starting in 1980 gives a weakly significant positive trend. Starting in 1990 gives no trend.

The result is interesting.

Excluding the Indian Ocean, 1980–2025 gives:

+1.12 Cat4+ storms per decade
95% CI: +0.03 to +2.20
p = 0.043

That looks like a positive trend.

But using the same basin exclusion and starting in 1990 gives:

−0.08 Cat4+ storms per decade
95% CI: −1.72 to +1.55
p = 0.920

That is nothing. No trend at all.

This is where a straight-line trend can mislead. If the 1980s are low and the post-1990 period is higher, a straight line from 1980 to 2025 will slope upward. But that doesn’t mean there has been a steady rise.

Maybe the data are telling us this:

Low before 1990.
Higher after 1990.
Flat after that.

That is a step, not a trend.

Step change or continuing trend?

To test that idea, I used two simple models.

The first model is just a straight line: do counts go up or down over time?

The second model allows a jump around 1990. Then it asks: after allowing the jump, is there still a continuing trend?

I also used a Poisson model, because these are counts. Counts are not quite like temperature or pressure measurements. They are whole numbers: 7 storms, 12 storms, 21 storms. The Poisson model is a standard way to check this kind of data.

Step-versus-trend models for Cat4+ counts excluding the Indian Ocean. A simple trend line slopes upward, but the step-plus-trend model shows a pre/post-1990 level change with essentially no continuing trend.]

The result is clear.

The straight-line trend says there is an increase. But once the model is allowed to include a pre/post-1990 step, the continuing trend disappears.

The step is large:

OLS step: +4.75 Cat4+ storms per year after 1990
Poisson step ratio: 1.51
In plain English: the post-1990 level is about 51% higher than the pre-1990 level.

But the continuing trend is basically zero:

OLS continuing trend: +0.063 Cat4+ storms per decade, p = 0.933
Poisson continuing trend ratio: 1.004 per decade, p = 0.916

That is not a trend. That is noise.

A simple analogy: imagine your salary is flat for ten years, then jumps once, then stays flat. If you draw one straight line through the whole period, it slopes upward. But the real story is not “steady salary growth.” The real story is “one raise, then flat.”

That appears to be what is happening here.

What caused the 1990 step?

The graph and the statistics do not prove the cause.

But the timing is suspicious.

The late 1980s and early 1990s are exactly when several known changes matter for TC intensity analysis:

• the eastern North Pacific responsibility shift after 1987;

• the end of western North Pacific aircraft reconnaissance in 1987;

• wider and more mature use of Dvorak analysis;

• better satellite imagery and interpretation tools.

Even after excluding the Indian Ocean, the eastern North Pacific and western North Pacific remain in the dataset. So removing the Indian Ocean solves one problem, but not all of them.

The western North Pacific is especially important because it contributes a large share of global intense storms. Kamahori et al. compared JMA and JTWC data and found that after the end of US Air Force reconnaissance in 1987, the two datasets diverged in their treatment of extremely intense tropical cyclones. JTWC showed a large increase in extremely intense TC days, while JMA showed a decrease. They connected this divergence to the post-reconnaissance reliance on Dvorak-based estimates.

Again, this does not prove that every post-1990 Cat 4+ storm is overestimated. It says something more modest:

The 1980s and post-1990 records are probably not homogeneous enough to treat the 1980–2025 slope as a clean climate trend.

Where I ended up

I began with the thought that Klotzbach et al. might have been cherry picking by starting in 1990. Maybe the Indian Ocean was the main reason, and removing it would let us go back to 1980.

But the data argue against that.

The apparent 1980–2025 increase in Cat 4+ storms, excluding the Indian Ocean, is mostly a low-1980s versus higher-post-1990 step. Once we allow that step, there is no continuing trend.

This is the cleanest summary I can make:

For weak named storms, modern observing systems create a serious detection problem, especially for storms lasting ≤2 days.

For Cat 4+ storms, the detection problem is much smaller. The storms were seen. But the intensity classification problem remains.

In IBTrACS USA_WIND, excluding the Indian Ocean, Cat 4+ counts are higher after 1990 than before 1990. But since 1990, there is no trend in Cat 4+ counts in this construction.

That does not mean intense TCs cannot be affected by climate change. It does not address future projections. It does not address rainfall, storm surge, rapid intensification, or landfall risk.

It only says that this particular observed-count question gives a more cautious answer than the simple 1980–2025 trend line suggests.

Conclusion

The 1990 boundary is annoying. But it seems justified.

Short-lived weak storms have a clear observing-system problem. Modern scatterometers and microwave instruments make weak, short-lived storms easier to detect and classify. That explains why raw named-storm counts are dangerous.

Cat 4+ storms are different. They are not being newly discovered by QuikSCAT. But their peak intensities are still sensitive to satellite resolution, Dvorak interpretation, agency practice, and the loss or presence of aircraft reconnaissance. A storm does not need to disappear from the archive to change the trend; it only needs to move from 105 kt to 115 kt.

So the answer to “are there more Cat 4+ storms than in the 1980s?” is:

In this dataset, yes, the post-1990 level is higher than the 1980s.

But the better interpretation is a pre/post-1990 step change, not a continuing increase.

Since 1990, the Cat 4+ count trend is essentially zero.

That’s not as exciting as a headline. But it is probably closer to the data.

References

Klotzbach, P. J., Wood, K. M., Schreck III, C. J., Bowen, S. G., Patricola, C. M., & Bell, M. M. (2022), Trends in Global Tropical Cyclone Activity: 1990–2021, Geophysical Research Letters.

Klotzbach, P. J. & Landsea, C. W. (2015), Extremely Intense Hurricanes: Revisiting Webster et al. (2005) after 10 Years, Journal of Climate.

Landsea, C. W. et al. (2010), Impact of Duration Thresholds on Atlantic Tropical Cyclone Counts, Journal of Climate.

Velden, C. et al. (2006), The Dvorak Tropical Cyclone Intensity Estimation Technique: A Satellite-Based Method that Has Endured for over 30 Years, Bulletin of the American Meteorological Society.

Kamahori, H., Yamazaki, N., Mannoji, N., & Takahashi, K. (2006), Variability in Intense Tropical Cyclone Days in the Western North Pacific, SOLA.

NOAA/NCEI, International Best Track Archive for Climate Stewardship and ADT-HURSAT dataset documentation.

Notes

The data used here are from IBTrACS — the International Best Track Archive for Climate Stewardship. IBTrACS combines tropical cyclone best-track data from many agencies into one archive. It is probably the best place to start for global historical TC work, but it is not one uniform observing system. It is a compilation of agency best tracks. Different agencies, different basins, different observing systems, and changing methods all matter. NOAA describes IBTrACS as a merged global best-track archive, while UCAR’s Climate Data Guide notes that changing operational procedures and observing systems have produced significant heterogeneities in the record.

For these graphs I used the US-agency winds, USA_WIND. In the Atlantic and eastern Pacific this is mostly the NHC/CPHC side of the record; elsewhere it is generally the JTWC side. Each storm is counted once, by its peak category. Cat 4+ means Cat 4 plus Cat 5.

That raises the obvious next question. If South Australia can produce so much very cheap electricity for part of the day, what does that actually mean for the price people pay? And since the rest of the National Electricity Market is moving in much the same direction — more rooftop solar, more utility solar, more wind, more batteries, and more stress on evening supply and on the transmission system — what does the South Australian experience tell us about electricity prices more broadly across eastern Australia?

The answer is that wholesale prices and household bills are connected, but they are not the same thing. Wholesale prices are what electricity trades for in the market, region by region, every five minutes. Retail prices are the bills households receive after wholesale costs are combined with network charges, environmental scheme costs, hedging, retail operating costs and margin. So the fact that wholesale prices can be extraordinarily low at noon does not automatically mean household bills will be low as well.

That is why it helps to step back and look at the longer sweep of the data. The wholesale charts show a market that has become more volatile, more seasonal and more shaped by renewable output. The retail charts tell a different but equally important story: what households actually experience after all the other pieces of the electricity system have been added in.

Wholesale: lower on average, harsher by hour

Retail bills fold in far more than the five-minute spot price. In the AER’s 2025–26 default-offer framework, wholesale costs make up roughly 33% to 41% of the bill, network charges about 33% to 48%, and environmental costs about 3% to 4%. Retailers also hedge heavily, because households are billed on relatively stable tariffs even though wholesale prices can swing violently from one interval to the next.

The wholesale story over the past 25 years is dramatic - nominal prices:

And taking inflation into account:

Prices were volatile in the 2000s, surged at times in the late 2010s, then blew out in the energy crisis of 2022. The AER says prices in the NEM fell from those unprecedented highs as fuel prices eased and renewable generation increased, but remained above pre-2022 levels and still showed significant volatility.

Then came a sharp easing late in 2025. AEMO says wholesale electricity prices across the NEM averaged $50/MWh in Q4 2025, down $39/MWh, or 44%, from Q4 2024 and down $37/MWh, or 43%, from Q3 2025. It also says the contribution from prices above $300/MWh dropped to just $3/MWh of the quarterly NEM average, down from $17/MWh a year earlier.

Why did wholesale fall so hard? Because the grid got more renewable and more flexible. AEMO says renewables including storage supplied 51.0% of NEM energy in Q4 2025, the first quarter above 50%. Wind output was up, solar was up, and batteries mattered most when prices usually bite: during the 4 pm to 8 pm evening peak, battery discharge was marginal in 25% of pricing intervals, displacing gas and hydro. That combination lowered both ordinary energy prices and the number of extreme price spikes.

But that is only half the story, and South Australia sits right in the middle of the other half. AEMO says transmission constraints and interconnector limits helped drive record occurrences of zero and negative prices in South Australia (48% of intervals) and Victoria (43%) in Q4 2025. In other words, Australia is increasingly producing very cheap electricity at some times of day, especially when solar is abundant, but not always moving it to where it is needed, when it is needed. Cheap generation does not automatically mean cheap firmed, delivered power.

Retail: why bills can stay high

And adjusted for inflation:

That is why retail bills can keep rising even when spot prices cool. The most useful official example is South Australia’s 2025–26 default offer. The AER says a South Australian residential customer without controlled load saw the safety-net price rise by $71, or 3.2%, to $2,301. But the cost stack underneath that total tells the real story: network costs fell by about $25, environmental costs fell by about $25, while wholesale costs rose by about $53 and retail costs rose by about $64.

The AER says the wholesale increase in South Australia was driven by higher contract prices and a load profile that had become peakier in the evening, raising hedging costs. It says retail costs rose because of higher operating costs and increases in bad and doubtful debt and smart-meter costs. That is the key point: bills do not follow the spot market quarter by quarter; they follow hedged costs, network charges and retail overheads as well.

The retail charts tell the public story more directly than the wholesale charts alone. What households feel is not the elegance of a falling midday wholesale price, but the lived cost of the full electricity system: energy, networks, risk management, retail overheads, policy charges, and the cost of keeping supply available when the easy solar hours are over.

What the AER’s 3–4% environmental figure really means

This also matters for one of the most politically loaded claims in the energy debate: that “green certificate” costs are the main reason household bills are high. The AER’s 3% to 4% figure is real, but it is narrower than many readers may assume. In the regulator’s accounting, “environmental costs” mainly means compliance costs for schemes such as the federal Large-scale Renewable Energy Target (LRET), the Small-scale Renewable Energy Scheme (SRES), and jurisdictional schemes. The AER says most of that bucket relates to complying with the LRET and SRES. It is not a measure of every cost associated with the energy transition.

For South Australia’s 2025–26 default offer, that environmental line is $72 on a $2,301 annual residential bill — about 3.1% of the total. By contrast, the wholesale line is $859 and the network line is $897. So the AER is not saying that renewables or the transition only affect bills by 3% to 4%; it is saying that this specific policy-compliance bucket is 3% to 4% within the DMO cost stack.

That distinction matters, because broader transition-related costs often sit elsewhere. The AER says network charges cover the efficient costs of building and operating electricity networks, and those charges make up roughly a third to almost half of DMO 7 prices. So if readers are thinking about shared transmission, interconnectors, resilience spending, or the cost of adapting the grid to a different generation mix, much of that is more likely to show up under network costs rather than under the environmental line. In its South Australian DMO summary, the AER explicitly points to resilience spending, consumer energy resource integration, safety, cybersecurity, and network revenue decisions as important drivers inside that broader network bucket.

And in South Australia specifically, the latest bill increase was not driven by that environmental bucket anyway. The AER says environmental costs fell across all South Australian customer classes in 2025–26, mainly because federal and state renewable scheme costs fell. For residential customers without controlled load, that environmental component fell by 26% year on year.

There is one caveat to the retail proxy. ABS electricity CPI in 2024–26 was heavily distorted by government rebates. The ABS says electricity prices were up 37% over the year to February 2026 including rebate effects, but only 5% excluding those rebates. So the proxy is best read as a long-run picture of trend, scale and relative movement, not as a precise measure of what happened to actual out-of-pocket bills in each recent quarter.

The bottom line

The practical conclusion is not that electricity has become cheap. The longer-run charts suggest the opposite: even after the recent easing, wholesale prices in real terms remain well above where they stood two decades ago. What seems to be changing is not the long-run fact of higher prices, but the shape of the market: more hours of extraordinarily cheap electricity, especially around midday, coexisting with periods of intense scarcity and very high prices in the evening.

That is why South Australia matters so much. It is showing, earlier and more clearly than most other regions, what a solar-heavy system looks like in practice. It can produce astonishingly low prices when renewable output is strong, but that does not automatically translate into cheap electricity across the whole day, let alone across the whole household bill.

So the hopeful case is narrower, and more conditional, than a simple claim that renewables are making electricity cheap. It is that wholesale prices may finally be starting to soften at the right times and in the right conditions — when wind, solar and batteries are plentiful — even though the long-run trend remains higher than it used to be. Whether that becomes a durable fall in average delivered prices is still an open question.

Residential bills will only come down clearly and persistently if those wholesale savings are large enough, frequent enough and stable enough to feed through hedge books, if network costs stop adding upward pressure, and if retail competition passes more of the benefit through to households. And even then, some of the costs of a changing system will continue to sit elsewhere in the bill, especially in networks, resilience and system integration.

So the real question is no longer whether solar can make electricity very cheap in the middle of the day. We already know it can. The question is whether the whole system can turn that into lower and more stable prices over the full day, and ultimately into lower bills for households.

Notes

Wholesale price series
Wholesale electricity prices in this chart are based on the official National Electricity Market regional spot prices published by AEMO. I use AEMO’s regional price-and-demand data to calculate demand-weighted annual averages for each NEM region; the same kind of annual and quarterly volume-weighted regional price series is also published by the AER using AEMO data. In that sense the wholesale series is market-observed and directly verifiable, not a constructed estimate.

Retail price series
Retail electricity prices are harder to show as one long-run series because the official bill data are recent and area-specific rather than a single published 2000–2025 household-bill series. AER retail-bill estimates are built from offers observed in different pricing areas and are not directly comparable across all areas because they depend on local average consumption; ACCC bill estimates also depend on benchmark consumption and offer data. For the long-run chart, the retail line is therefore a constructed proxy: recent AER bill estimates are used as anchors and earlier history is extended with ABS capital-city electricity CPI (Sydney, Melbourne, Brisbane, Adelaide and Hobart as proxies). It is useful for trend and scale, but not as an exact measure of any one household’s bill.

First: how far can wind and interconnectors carry South Australia once solar is gone? Then: when the system is short, how many hours in a row does that shortfall last? Those pieces found something real and encouraging. Wind plus interconnectors can make a very large dent in South Australia’s night problem. But they also found the tail: long difficult runs still appear, especially in bad wind years.

The obvious next question is whether South Australia is a template.

Can you tell the same story for Victoria?

That matters because South Australia is a small state attached to larger neighbours. Victoria is not. Victoria is itself a large load centre. If the South Australian result is partly “renewables plus interconnectors work”, that is interesting. But if it is partly “small region leaning on bigger regions”, that is a different story.

So this time I ran the same kind of thought experiment for Victoria.

And the result is: South Australia is not a template.

First, make the comparison fair

If I simply compare Victoria and South Australia as they are today, Victoria looks much worse. But that is not quite a fair test, because Victoria’s wind fleet is not nearly as large relative to its own demand.

That is easy to miss because Victoria’s wind fleet is bigger in absolute terms.

But absolute terms are not the point.

In this 2025 night analysis, Victoria’s selected-hour demand averages about 5.4 GW and peaks around 9.8 GW, while the Victorian wind fleet in the model is about 5.5 GW. So Victoria’s current wind fleet is roughly the same size as average selected-night demand, but only a bit more than half of selected-night peak demand. By comparison, South Australia’s wind fleet is much larger relative to the size of its own load.

So before asking whether Victoria “works like South Australia”, I wanted to put Victoria on something closer to a South Australian footing.

The simplest way to do that was to scale Victoria’s wind fleet up until its wind-to-demand ratio matched South Australia’s on a peak-demand basis.

That produced a factor of about 1.44 times today’s Victorian wind fleet.

So there are really four Victoria cases in this article:

• today’s Victoria wind plus today’s practical network

• a Victoria wind fleet scaled up to South Australia’s relative footing

• twice that SA-relative Victorian fleet

• three times that SA-relative Victorian fleet

And then I asked what happens if Victoria also gets a stronger NSW link via a stylised VNI West future.

Today’s Victoria: almost every night is still short

The first graph is the blunt one.

With today’s Victorian wind fleet and today’s practical network, Victoria is still short on almost every selected night. In the practical-2025 run I did earlier, the model found zero selected hours with no shortfall in the capped SA+NSW case, with 365 events lasting at least 8 hours and 362 lasting at least 12 hours. That earlier summary also showed a remaining night shortfall of 16,500 GWh in the practical-2025 capped SA+NSW case.

That is not a small tail problem.

That is a system that, once solar is gone, is still structurally short on wind most of the time.

And that by itself is an important result. It tells us that “South Australia works, therefore Victoria should too” is not a serious argument.

So what if Victoria had an SA-sized wind fleet?

This is where the fairer comparison starts.

To make Victoria a fairer counterpoint to South Australia, I first scaled Victoria’s wind fleet up by about 1.44x. Then I asked what happens if Victoria also gets VNI West — the proposed new 500 kV NSW link that AEMO says would add about 1.7 GW of extra import capability into Victoria, with current planning for service around 2030–31. In the model I represented that with a simplified future NSW→Victoria cap of 3.57 GW. That is not a dispatch limit; it is a planning-style envelope for asking what a much stronger NSW connection might do.

And here is the first surprise:

• even after doing that, Victoria still does not look remotely like South Australia.

Under the “SA-relative Victorian wind plus VNI West” case, the model still finds about 252 shortfall events lasting 8 hours or more, and about 202 lasting 12 hours or more.

That is much better than where Victoria is now.

But it is still a lot of long bad nights.

Not a narrow tail. Not an almost-solved problem.

Double it again — and the problem is still large

So then I pushed it further.

What if Victoria gets not just an SA-like wind fleet, but twice that? In other words, not just “put Victoria on the same relative footing as South Australia”, but then double again from there.

That case does improve things a lot.

[repeated graph from above]

The count of 8-hour-plus shortfall events falls to about 138, and the 12-hour-plus count falls to about 90.

That is a serious improvement. It means a lot of nights that were bad before are no longer bad, and many others are less bad.

But it still does not turn Victoria into an SA-style story.

Even after giving Victoria an SA-like wind fleet relative to demand, and then doubling it again, you are still left with roughly a hundred and forty long shortfall runs of eight hours or more.

That is not the edge of the distribution. That is still a large part of the year.

Triple it — and the tail is smaller, but still not small

Then I tried three times the SA-relative Victorian wind fleet.

Again, the curve improves materially. The 8-hour-plus shortfall count falls to about 92, and the 12-hour-plus count to about 51.

That is real progress.

But it is also the clearest possible demonstration that Victoria is a different kind of problem.

With South Australia, stronger eastern support plus more wind pushed the issue into a much narrower tail. With Victoria, even after scaling up the wind fleet hard and assuming a stronger NSW link, the system still shows a lot of long difficult nights.

What if NSW grows too?

That naturally raises the next question.

So far, I had scaled up Victoria’s wind fleet and given it a stronger NSW link. But if Victoria is building much more wind through the 2030s, NSW probably is too.

So I reran the exercise with a more optimistic eastern-grid buildout: not just more Victorian wind, but NSW wind scaled up in step with Victoria wind.

That makes a real difference.

In the earlier Victoria-only scaling case, the SA-relative Victorian fleet still left about 252 shortfall events of 8 hours or more. Once NSW scales too, that falls to about 224.

At 2x SA-relative Victorian wind, the improvement is much larger: the count of 8-hour-plus shortfall events falls from about 138 to about 79.

At 3x SA-relative Victorian wind, it falls again, from about 92 to about 39.

The same thing shows up in energy terms. Remaining night shortfall falls from about 9.5 TWh to about 8.2 TWh at the SA-relative step, from about 5.4 TWh to about 3.1 TWh at 2x, and from about 3.7 TWh to about 1.6 TWh at 3x.

So this clearly helps. A lot.

But it also reveals something else. Once NSW scales too, the 3.6 GW NSW→Victoria cap starts to matter. It binds in only 0.1% of selected hours at the SA-relative step, but in 33% of hours at 2x and 56% at 3x.

That is useful, because it shows the problem changing shape again.

At first, Victoria’s problem is not enough local wind.
Then, even after Victoria builds much more wind, the wider system still does not have enough wind often enough.
But once NSW grows too, the bottleneck starts to shift back to transmission.

Even then, Victoria still does not become South Australia. The numbers are much better, but the tail remains real.

Where this leaves the debate

If you point to South Australia and say, “see, a high-renewables system can work”, it’s true — for the case where that system has big interconnects to larger neighbours.

If you’re the big neighbour, the problem is much harder.

More wind helps. Bigger interconnectors help. But they do not turn Victoria into South Australia.

South Australia after sunset is a difficult but tractable tail problem.
Victoria after sunset is still a much bigger system problem.

And that is why South Australia is not a template.

That was useful, but it left out the question engineers really care about.

A one-hour shortfall is one problem.
An eight-hour shortfall is another.
A 12- or 15-hour shortfall is another again.

So this time I am asking a simpler question:

When South Australia is short at night, how many hours in a row does that shortfall last?

What is a “shortfall event”?

I kept the same setup as before: only 5pm to 8am Adelaide time, so the focus is on the hours when solar is effectively gone.

A shortfall event is then a run of consecutive night-time hours when South Australian demand is still greater than the supply available from the wind-and-interconnector scenario being tested.

That does not mean “no renewable energy for eight hours.” It means that even after counting the wind and interconnector support in that scenario, there is still demand left over that has to be met by something else.

That “something else” might be batteries, gas, hydro imported from elsewhere, demand response, or some combination.

The 2025 picture: better, but not gone

The first chart asks a very direct question:

How many shortfall events last at least 1 hour, 2 hours, 3 hours, and so on?

Lower is better.

And the result is useful.

With South Australian wind alone, long shortfall runs are common. That is exactly the sort of thing critics of renewable-heavy systems worry about, and rightly so.

But once I add the stronger eastern-support case — South Australian wind plus capped support from Victoria and NSW — the curve drops sharply. Double South Australian wind, and it drops further. Triple it, and further again.

So the good news from the previous article was real. Wind plus stronger interconnection really does make a big dent.

But this chart also adds the realism that the earlier energy totals can blur.

Even in 2025, under the stronger eastern-support case, there are still roughly thirty nights with eight or more consecutive hours of remaining shortfall. Double South Australian wind and that falls to roughly twenty. Triple it and it gets down toward ten.

That is much better.

But it is not “problem solved.”

And these are not tiny misses. Spread across those 8+ hour runs, the remaining gap is still on the order of hundreds of megawatts, and in the base case it is closer to the scale of about a gigawatt than to the scale of one small emergency peaker. In other words: this is backup-plant territory, not something you wave away.

Then the obvious question: what if 2025 was a good year?

So I reran the exercise using wind conditions from each year back to 2015, while holding 2025 night demand and the 2025 wind fleet assumptions fixed.

That gives a cleaner stress test:

If today’s or planned system had existed earlier in the decade, which wind year would have been toughest?

For the eastern-support cases, the answer is 2017.

And the bad-year picture is noticeably rougher than the 2025 chart.

In the stronger eastern-support case, the number of 8-hour-plus shortfall runs rises from roughly thirty in 2025 to roughly sixty in the worst replay year. Double South Australian wind and the bad-year number is still in the thirties. Even with 2x SA wind, last year’s result was around twenty such runs.

That is a pretty useful way to picture the problem.

A decent wind year can make the system look manageable.
A bad wind year can roughly double the number of long ugly nights.

And again, the residual gap in those long runs is not trivial. In the bad-year replay, you are looking at something closer to a gigawatt-scale backup problem than a few loose odds and ends.

So the story is mixed.

The earlier encouraging result survives: stronger eastern support helps a lot, and more South Australian wind helps a lot.

But the weather-year replay shows that the tail can come roaring back in a bad year.

This is where gas re-enters the picture

This is also where the debate gets less ideological and more practical.

If the problem were mostly one-hour dents, batteries would look like the overwhelmingly obvious answer. And batteries are indeed part of the answer.

But once you start seeing long consecutive shortfall runs, especially in bad wind years, you are back in the world of firm backup.

That can be gas. It can be hydro if you have enough interconnection to somewhere with dispatchable hydro. It can be some combination of gas, batteries and demand response. But it has to be something that can sit there and carry the system through the tail.

This is why I think the political presentation of the transition can be misleading.

A state can turn off its own gas plants. It can declare progress. It can issue the press release.

But if it is still depending on interconnectors to a wider grid that contains gas and coal, the underlying engineering problem has not disappeared. It has been partly exported.

That may still be a perfectly sensible strategy. South Australia is a small state attached to a much larger system. That is an easier problem than asking Victoria to do the same thing.

But it is not the same as saying the problem no longer exists.

So what should we take from this?

The previous article’s encouraging result was real: wind plus interconnectors can dramatically shrink South Australia’s night-time problem.

But duration matters. The issue is not only how much energy is missing in total. It is how often you get long consecutive runs where wind plus interconnection still do not cover demand.

And the weather-year replay matters even more. In a bad wind year, those long runs become much more common.

So my conclusion is pretty simple:

South Australia’s night-time problem looks solvable, but not by wishful thinking.

It looks solvable by some mixture of:

• a lot of wind

• stronger interconnection

• storage

• and firm backup for the ugly tail

That is less romantic than saying “the batteries have won” or “the gas era is over.”

But it is more believable.

One final caveat: these are still simplified thought experiments, not a full dispatch model, and the historical replay is cleaner for South Australia than for Victoria or NSW because the eastern wind fleets were smaller in the earlier years.

Still, the broad lesson is hard to miss:

A decent year can make the problem look manageable.
A bad year can roughly double the number of long bad nights.
And that is where firm backup earns its keep.

The next question is Victoria — and that is where South Australia’s status as a small state attached to larger neighbours may turn out to matter a great deal.

Solar disappears in the evening. Wind is unreliable. And when South Australian wind is weak, Victorian wind is probably weak too. So interconnectors will not save you. They may help at the margin, but once the sun goes down you are still in trouble.

That sounds plausible. But it is also vague. So I wanted to test it more concretely.

I started with actual 2025 wind and demand data, and then narrowed the problem to the hours that matter most for this question: 5pm to 8am Adelaide time, when solar is effectively absent and the system has to lean on some mix of wind, interconnectors, storage, gas and demand response.

This is not a dispatch model, and it is not a formal market simulation. It is a set of structured thought experiments using real data. The question is not “what would the market do in every 5-minute interval?” The question is simpler and more physical:

How much can wind and interconnection shrink South Australia’s night-time supply problem?

First: does Victorian wind actually help when South Australian wind is weak?

Before getting to the night-time charts, it is worth starting with a simpler question.

The calendar-2025 correlation plot is a useful first check.

This uses all 24 hours, not just the evening and overnight period. It shows that South Australian and Victorian wind are positively correlated, which is what you would expect from large shared weather systems. But it also shows that they are far from perfectly locked together.

That matters.

The 2025 results say that when South Australian wind falls into its bottom decile, Victorian wind still averages about 660 MW. Victoria is also in its own bottom decile only about 31% of those times. So the two regions are clearly not independent — joint weak-wind periods happen much more often than they would if the two systems were unrelated — but nor are they simple clones of each other.

That is the first important point.

Victoria helps, but not perfectly.
It is neither a magic insurance policy nor a useless one.

That is also why I would not dwell too long on the correlation charts. They tell us the meteorological relationship is helpful but incomplete. What readers really want to know is what that means once the sun has gone and actual demand still has to be met.

Plot A: what does South Australia actually need at night?

So I restricted the data to 17:00 to 08:00 Adelaide local time across calendar 2025 and looked first at South Australian demand on its own.

That chart does something important: it grounds the discussion in scale.

Average demand in those selected hours is about 1.56 GW. It moves around through the year, with winter evenings and mornings generally higher, and with some sharper spikes at other times. Once solar is gone, that load still has to be covered. The problem does not disappear just because rooftop PV did a lot of work in the middle of the day.

This is the load that the rest of the exercise is trying to meet.

Plot B: South Australian wind and available Victorian wind

The next obvious question is: what does the wind look like against that demand?

In Plot B I put South Australian demand alongside South Australian wind and Victorian wind. The eastern line here is best thought of as an available weather envelope, not as something South Australia can necessarily import in full.

Two things jump out.

The first is that South Australian wind alone often does a meaningful amount of work, but it also has plenty of weak periods. That is no surprise.

The second is more interesting: Victorian wind is often substantial even when South Australian wind is unimpressive. Not always. But often enough to matter.

This is the visual version of the earlier correlation result. The weather diversity is real. But at this stage it is still just potential help, not actual deliverable help.

Plot C: cap Victorian support and the transmission problem appears

So the next step is to stop pretending that all eastern wind is freely available.

In Plot C, I cap the Victorian contribution. Now you can see the bottleneck much more clearly. The eastern line is suddenly flattened at the cap in many intervals. The problem is no longer just meteorology. It is transmission.

This is a crucial transition in the story.

When people say “Victoria can help South Australia,” that is true in a weather sense. But what the grid actually cares about is not how windy Victoria is in the abstract. It cares about how much of that wind can get through when South Australia needs it.

That is why interconnectors matter. Not because they create energy, but because they convert geographic diversity into usable support.

Plot D: add NSW and the eastern weather envelope gets much bigger

The next thought experiment is to stop looking only at Victoria and instead look east more broadly.

In Plot D I add Victoria plus NSW wind as an unconstrained eastern envelope.

This is where the picture becomes much more encouraging.

Once NSW is included, the eastern weather resource becomes much larger. On many nights there is a lot of wind somewhere to the east, even when South Australian wind is not doing much. The supply potential is clearly there.

But again, this is still only a weather envelope. It is not yet a deliverable supply line.

Plot E: cap eastern support and you see what the network is throwing away

Plot E then applies a cap to the combined eastern contribution.

And this is where the tension in the system becomes very easy to see. There are many hours when the eastern wind resource is plainly there, but the deliverable contribution is pushed flat against the cap. In those moments the limiting factor is not the existence of wind. It is the size of the pipe.

That is a big deal.

It means that, in the hours where solar is gone, part of the problem is not “there is no renewable energy available.” Part of the problem is “there is renewable energy available, but the network cannot move enough of it into South Australia.”

That is not the whole problem. But it is very much part of it.

A more ambitious thought experiment: strong post-PEC eastern support

I then pushed the exercise further with a stylised stronger-eastern-support case.

To be clear, this is not a fantasy number pulled from thin air. Current plans are for Project EnergyConnect Stage 2 to lift the South Australia–southern NSW interconnector from today’s 150 MW Stage 1 to 800 MW in both directions, and AEMO says that also lifts nominal Heywood capability to 750 MW. AEMO’s latest timetable has PEC Stage 2 entering service in late 2026 and reaching full capacity in 2027; HumeLink, the new 500 kV NSW backbone linking Wagga Wagga, Bannaby and Maragle, is also scheduled for 2027. So this stronger-eastern-support case should be read not as today’s grid, but as a stylised version of the network planners are trying to build over the next couple of years.

That is what the separate post-PEC-style plot is for.

Here the picture changes markedly. South Australian wind plus strong capped eastern support gets much closer to demand for a very large share of the selected hours. The gap does not disappear, but it starts to look much more like a tail problem — a problem of difficult residual hours — rather than a blanket impossibility across most of the night.

That is already more encouraging than the standard rhetoric would lead you to expect.

The most useful summary: sort the hours from worst shortfall to zero

Chronological time series are good for intuition, but they are not the easiest way to compare cases.

So the next step was to rank the selected night hours from largest remaining shortfall down to zero shortfall, and draw duration curves. In these charts, lower is better.

This turns out to be a very clean way to see what each assumption is actually buying.

SA wind only

With South Australian wind on its own, the average remaining shortfall is about 648 MW, and there is no shortfall at all in only 19% of the selected hours.

That is a useful baseline. South Australian wind does a lot, but on its own it leaves a substantial night-time gap.

A very tight eastern cap

A tightly capped eastern-support case helps, but not dramatically. Average shortfall falls to about 532 MW, and the no-shortfall share rises only to about 26%.

So a small amount of extra eastern transfer is nice, but it does not transform the problem.

Double or triple South Australian wind

I then tried simple overbuild thought experiments: SA wind x2 and SA wind x3. These are not physical build-out simulations. They are intuition-building scaling exercises.

Doubling South Australian wind cuts the average shortfall to about 320 MW, with 59% of selected hours showing no shortfall.

Tripling it gets the average shortfall down to about 202 MW, with 72.5% of hours showing no shortfall.

So local overbuild clearly helps — and helps a lot.

But the really striking comparison is this:

Strong eastern support does even more

Under the stylised strong-eastern-support case (purple line in figure above), the average shortfall falls to just 67.9 MW, and 84.7% of selected hours show no shortfall at all.

That is the result that surprised me most.

On this particular set of assumptions, stronger interconnection does more to shrink South Australia’s night-time gap than simply doubling or even tripling South Australian wind on its own.

That does not mean local wind build-out is unimportant. It means geography plus transmission is doing an enormous amount of work.

Final thought experiment: what extra does South Australian wind overbuild buy you once strong eastern support exists?

The last chart is, to me, the most encouraging one.

It asks a more refined question. Suppose strong eastern support exists. Suppose South Australia can already pull heavily on Victorian and NSW wind in difficult hours. What extra does South Australian wind overbuild still buy you on top of that?

The answer is: still quite a lot, but in a more marginal way.

With strong capped eastern support alone, the average remaining shortfall is about 67.9 MW, and 84.7% of selected hours have no shortfall.

Add 2x South Australian wind, and the average shortfall falls to about 37.2 MW, with the no-shortfall share rising to 91.4%.

Add 3x South Australian wind, and the average shortfall falls further to about 23.6 MW, with the no-shortfall share rising to 94.2%.

That is a nice result because it shows the roles are complementary.

Interconnection does not make local wind unimportant. And local wind overbuild does not make interconnection unimportant. The two work together.

The eastern connection gets you a very long way. Additional local wind then makes the remaining difficult tail smaller still.

So what is the actual takeaway?

I think the basic message is now fairly clear.

At night, South Australia is not saved by any single magic answer. Not by local wind alone. Not by interconnectors alone. Not by a fantasy that some other region is always windy when South Australia is not.

But neither is it trapped in the simplistic story that “once solar is gone, renewables are basically useless.”

What the data suggests is something much more practical:

• Wind diversity across SA, Victoria and NSW is real, but incomplete

• Transmission turns that diversity into usable support

• Stronger eastern links can make a very large dent in South Australia’s night-time gap

• Further South Australian wind overbuild can then push the remaining tail down further still

That does not eliminate the need for storage, gas, demand response, or other firming tools. It changes the size of the remaining problem.

And that is the key point.

The issue after sunset is not necessarily a giant all-night cliff. Under reasonably strong wind and transmission assumptions, it becomes a much smaller set of difficult hours at the edge of the distribution.

That is still an engineering problem. But it is a far more tractable engineering problem than the usual caricature suggests.

A final note on what this is — and what it is not

These are simple, transparent thought experiments. They are not a full dispatch model. The strong-eastern-support case is stylised, not an official simultaneous transfer limit. The South Australian wind x2 and x3 cases are scaling exercises, not construction schedules.

But those caveats do not weaken the central result.

They sharpen it.

Because the first-order point is robust: wind plus interconnection can make a very large dent in South Australia’s solar-free hours. And when you combine stronger eastern support with more South Australian wind, the residual night-time problem becomes smaller again.

That does not prove the whole system is solved.

It does show that, after sunset, the problem is much less hopeless — and much more interesting — than either boosters or skeptics usually admit.

Two caveats matter. First, Victoria and NSW may need some or all of that wind themselves; these charts show availability and transfer envelopes, not a full market-clearing dispatch outcome. Second, South Australia is a relatively small region connected to much larger neighbours. That is a much easier problem than asking whether the same trick could carry Victoria. That is a separate question — and probably a subsequent article.

The next question is not just how often shortfalls appear, but how they cluster. One isolated short hour is a very different engineering problem from ten or twelve consecutive hours of low residual supply.

Sources: AEMO NEMWeb public data. Demand is South Australia’s Five-minute Actual Operational Demand series; wind output is aggregated from unit-level Actual Generation and Load SCADA data for wind DUIDs in SA, Victoria and NSW. Overnight charts use Adelaide local time and include only the 5pm–8am window in calendar 2025.

Background - I was trying to figure out why Matlab wouldn’t let me add an important Mapping addition so I could rerun some tests on global Tropical Cyclone data (ADV-HURSAT). At the same time, a friend said that China was really making huge advances in renewables, and this prompted a diversion. I thought it would be a short diversion, but I was wrong. I do plan to be back writing about TCs soon.

In the meantime - I’ve written a cluster of posts on renewables and electricity markets — starting with a broader look (including China and Europe), then drilling into South Australia as the most useful real-world laboratory for “high renewables.”

If you’re new to this series, the best way to read it is in time order. Each post stands alone, but together they build a clear picture of why “cheap renewables” doesn’t automatically produce cheap electricity, and why the system ends up paying for flexibility, backup, and wiring.

Below is an index of the recent posts, in earliest → latest order, with a short summary for each.

1) China Renewables and PR

A look at how easy it is to tell very different stories using the same data. China’s renewables growth is huge — but so is coal. This post is about absolute numbers vs percentages, and why PR charts can mislead.

2) Germany and Renewables

Germany’s domestic generation mix: coal down hard, nuclear to zero, wind/solar up massively — but the interesting question is what fills the gaps. This post is the “big picture” before following the power across borders.

3) Germany, Renewables & Keeping the Lights On

Interconnectors as “extended grid storage”: Germany’s imports, who supplies them, and why cross-border wiring quietly does three jobs at once — sharing cheap generation, sharing backup, and sharing weather luck.

4) The 1,500km Relay Race: Why Spanish Wind Struggles to Reach Germany

A deep dive into the physical bottleneck: transmission. Spain can be “anti-Germany” meteorologically, but moving power across borders is a multi-decade, multi-billion euro relay race.

5) UK and Renewables

UK generation from 2015–2024: coal collapses, wind surges, nuclear declines, and gas stays central. A simple data-first view that sets up the follow-up question: what does “renewables success” mean in system terms?

6) UK and Renewables: How the Magic is Done

A continuation that digs into the “how”: what the UK is leaning on (imports, gas, biomass accounting, etc.), and how a high-renewables narrative changes once you include full system constraints.

7) Germany, UK, and France: The Price of Energy Transition

A comparative look at what consumers experience — the cost stack, the trade-offs, and why “more renewables” doesn’t map neatly to “lower bills” without flexibility and wiring.

South Australia: the daily price rollercoaster and what it really means

8) The 100% Experiment: How a Small State Became the World’s Renewable Laboratory

South Australia is the cleanest real-world lab for high renewables because it has huge rooftop solar, lots of wind, and tight interconnection. This post sets the stage: what changed from 2015–2025, and why the grid behaves differently from fossil-era intuition.

9) The Pricing Paradox: Why ‘Free’ Solar Leads to $15,000 Energy Rollercoasters

Midday prices go negative; evening prices spike to extreme levels. This post explains the “canyon curve” and why wholesale volatility doesn’t translate into cheap household bills. Best entry point if you only read one SA piece.

10) Batteries in SA: The Market in Operation

Batteries aren’t being built to store “a day” of electricity — they’re being built because the market pays for fast response in the shoulder periods. This post shows what batteries actually contribute on the grid, by time of day.

11) Batteries and the Market Machinery — How the Grid’s ‘Invisible Hand’ Rewards Speed

How the NEM rewards speed: 5-minute settlement, FCAS “twitch” markets, negative price behaviour, and why grid-forming inverters matter.

12) The Negative Price Puzzle — Why Generators Pay to Produce Electricity

Negative prices aren’t a glitch — they’re how the market clears surplus supply when some generators can’t or won’t stop. Coal/gas cycling economics, wind/solar incentives, and why negative bids can still be rational.

13) The Invisible Renewables Market: From LGCs to CIS to REGO

The “second market” that helped fund renewables: LGCs. Who pays, how the cost shows up in bills, why LGC values collapsed (because the scheme worked), and what newer mechanisms are trying to do instead.

14) The Solar Lunch Break That Coal Can’t Take (and Gas Only Sort of Can)

A more technical post: why thermal plants can’t simply shut down at 8am and restart at 4pm without cost and risk. The physical side of the negative-price story — ramps, starts, cycling, and why batteries are a useful bridge.

Suggested reading paths

• If you only read one: The Pricing Paradox

• If you want the SA market story: Pricing Paradox → Batteries in SA → Market Machinery → Negative Price Puzzle → Invisible Renewables Market → Solar Lunch Break

• If you like infrastructure + geography: Germany Keeping the Lights On → 1,500km Relay Race

• If you want “costs to consumers”: Germany/UK/France: Price of Energy Transition

• Midday: solar floods the system → prices crash (often below zero).

• Early evening: solar vanishes on schedule → prices spike.

A natural question follows, and one of our commentators asked it:

Why don’t thermal generators just turn off at 8am when solar takes over, then turn back on at 4pm when the shoulder ramp starts?

If you’ve never been inside a power plant (or a power-plant P&L), it sounds reasonable.

It usually isn’t.

1) The 8am–4pm thought experiment

Coal (circa 2010): the “battleship boiler” problem

Take a representative coal unit: not ancient, not brand new — something like a few-hundred-MW subcritical or early supercritical unit.

The key constraint isn’t ideology. It’s thermal mass.

A coal unit is a big boiler, big steam lines, big turbine casings — thick metal that expands and contracts. Heat it up fast and cool it down fast and it fatigues. Do it daily and you don’t just pay in fuel. You pay in maintenance, forced outages, and shortened life.

So coal plants have three awkward characteristics:

1. They have a minimum stable operating level (often around 25–40% of nameplate for “average” coal units).

2. They often have long minimum up/down times (order tens of hours, in generic “average plant” summaries).

3. Warm starts take hours, not minutes.

A useful set of indicative start timings comes from the Parsons Brinckerhoff assessment (prepared for UK DECC). For a coal unit, they break start-up into (i) “notice to synchronise” and (ii) “synchronisation to full load.” In their table, a warm start for coal is roughly 300 minutes notice-to-synch plus 85+ minutes synch-to-full-load — i.e. ~6+ hours to full load.

That creates the core problem with the “8am–4pm switch-off” plan:

If you want coal back for the shoulder, you often have to start the restart around mid-morning — which means you never really switched off.

So coal plants tend to idle low, export if they can, or accept negative prices if they must.

A simple cash intuition (very approximate):
Consider a 500 MW coal unit forced to sit at ~35% minimum load for 8 hours (175 MW). That’s 1,400 MWh generated. If the price is –$20/MWh, it pays $28k to stay online. Add fuel burn, and the “midday pain” can be on the order of tens of thousands of dollars for that one unit.

Could it be cheaper to shut down and restart? Sometimes on paper, yes — but the long start time and minimum downtime often make the “intra-day off” option non-existent, even before you count the reliability risk and wear.

Gas (circa 2010): much more flexible, but still not a light switch

Now take a typical gas plant of that era. The important distinction is:

• CCGT (combined-cycle): efficient, but slower and more constrained.

• OCGT (open-cycle “peaker”): fast and flexible, but expensive to run.

For a representative existing CCGT, Parsons Brinckerhoff reports indicative hot start timing around 15 minutes notice-to-synch plus 40–80 minutes synch-to-full-load (so roughly ~1 hour to full load). Warm and cold starts take longer, but still generally in the “hours, not half-days” range.

So gas can actually take the lunch break.

Economically, gas has the opposite problem from coal:

• Gas fuel is expensive compared with coal.

• Idling a CCGT at minimum load is very costly fuel-wise.

• Starting costs exist (maintenance, stress, lifecycle), but they’re usually far smaller than running through long negative-price periods.

This is why batteries show up as the perfect bridge: they don’t replace gas, they buy time — covering the first part of the evening shoulder while the gas fleet ramps and synchronises.

Most readers will stop here, but for those more interested..

2) What newer plants (or retrofits) can do — and a simple comparison table

The range of “what’s possible” is much wider than many people assume, because flexibility is partly design and partly retrofits/control strategy.

IRENA’s flexibility summary gives a simple “average plant vs post-flexibilisation” comparison:

• Coal minimum load can be reduced (e.g., 25–40% → 10–20% in some post-flex cases).

• Coal start-up time can shorten (e.g., 2–10 h → 80 min–6 h), and minimum down/up times can shrink in some configurations.

• CCGT can also be improved (minimum load lower, ramps faster, start time shorter).

• OCGT is inherently fast (minutes to full load), which is why it exists.

A “solar glut” economics table (illustrative, not precise)

Assumptions (for intuition only):

• Midday “glut” period = 8 hours (08:00–16:00)

• Midday price = –$20/MWh

• Coal fuel = $2–$3/GJ, gas = $10/GJ (you can swap in your own)

• Plants forced to operate at minimum load (i.e., they can’t simply opt out)

Startup costs are based on NREL’s synthesis of cycling/start costs (Intertek APTECH / NREL) and include startup fuel + representative cycling O&M/capital components.

A few points on the the table:

• The “idle cost” depends massively on the actual price, contract position, and whether the unit is genuinely forced to run. The point is the scale and the incentives.

• For coal, feasibility is dominated by start time and minimum down/up constraints, not by whether the spreadsheet says restart might be cheaper.

• “Newer coal” is largely academic in Australia (because they aren’t building it), but the flexibility ranges matter for international readers and for retrofits.

3) The boring details section: ramp rates, start types, minimum load, and why cycling hurts

Start types: hot, warm, cold

A common definition is:

• Hot start: down for < ~8 hours

• Warm start: 8–48 hours

• Cold start: >48 hours

IRENA uses this style of definition and notes explicitly that cold starts take longer than hot starts.

Parsons Brinckerhoff’s table makes the same point in a way traders care about: start-up consists of (i) preparation to synchronise and (ii) ramp from synchronisation to full load — and both get worse the colder the plant is.

Ramp rates and minimum loads (typical ranges)

From IRENA’s summary table:

• Hard coal (average): min load 25–40%, ramp 1.5–4%/min

• CCGT (average): min load 40–50%, ramp 2–4%/min

• OCGT (average): ramp 8–12%/min, start minutes, but lower efficiency at part-load

The important nuance: ramp rates only matter after you’re synchronised. Coal’s killer is time-to-synchronise.

Efficiency penalties at part load

Part-load operation isn’t just annoying — it burns more fuel per MWh.

IRENA’s table gives representative efficiencies:

• Hard coal: ~43% at full load vs 40% at 50% load

• CCGT: ~52–57% at full load vs 47–51% at 50% load

• OCGT: ~35–39% at full load vs 27–32% at 50% load

So “keeping the unit warm at minimum load” comes with a very real fuel penalty.

Cycling costs: it’s not just startup fuel

NREL’s cycling cost synthesis is useful because it turns “wear and tear” into ballpark numbers.

Two relevant tables:

• Startup fuel inputs (MMBtu per MW of capacity): supercritical coal has much higher startup fuel requirements than other types; gas combined cycle startup fuel is small in comparison.

• Maintenance & capital start costs (per MW of capacity, by hot/warm/cold starts): even the median “lower bound” costs imply tens of thousands of dollars per start for large units — and there’s a wide range between plants.

NREL also notes that long-term heat-rate degradation attributable to cycling is plausibly ~1–5%, and that a ~1% heat-rate increase over 4–5 years is “not uncommon.”

So the intuition is:

Cycling doesn’t just change today’s operating cost.
It changes the plant’s future cost and reliability profile.

Which is why market participants price flexibility — and why “just ramp harder / shut down more” isn’t free.

Wrap-up: this is why batteries make sense

The shoulder problem isn’t “we need 24 hours of batteries.”

It’s “we need fast power for the first hour or two while slower assets wake up.”

• Coal usually can’t take an 8-hour lunch break and still be ready for the shoulder.

• Gas often can — but it’s expensive to idle, and it still dislikes being treated like a yo-yo.

• Batteries thrive precisely in that gap.

That’s the practical reason the evening price spike exists — and why the spike has been a profitable invitation for batteries.

Notes

NREL (2012) Power Plant Cycling Costs (startup fuel inputs, startup costs, cycling impacts):

https://docs.nrel.gov/docs/fy12osti/55433.pdf

IRENA (2019) Flexibility in Conventional Power Plants (min load, ramp rates, start times, efficiencies):

https://www.irena.org/-/media/Files/IRENA/Agency/Publication/2019/Sep/IRENA_Flexibility_in_CPPs_2019.pdf

Parsons Brinckerhoff for UK DECC (2014) Technical Assessment of the Operation of Coal & Gas Fired Plants (start timings):

https://assets.publishing.service.gov.uk/media/5a7dfc11ed915d74e33ef4be/Technical_Assessment_of_the_Operation_of_Coal_and_Gas_Plant_PB_Power_FIN....pdf

We saw that the price crashes below zero in the middle of the day - because of solar. And that the price soars high in the early evening - because of solar.

But behind that volatile, five-minute auction there has always been a second market running in parallel. One that helped build the renewable fleet in the first place.

That market is the market for Large-scale Generation Certificates (LGCs).

If the spot market is the fast-paced auction for energy, LGCs were the slow-burning policy engine that got the renewables sector off the ground.

To understand what’s happening now — and why new schemes like CIS and REGO exist — we need to understand that invisible layer.

The Two-Check System

A wind farm in South Australia does not receive just one payment per megawatt-hour (MWh) generated.

It receives two:

• The wholesale electricity price - that’s the price we’ve been analysing — sometimes $150/MWh, sometimes $15,000/MWh, sometimes negative.

• An LGC.

For every MWh of eligible renewable electricity generated, the project receives one certificate.

Think of an LGC as a green receipt. It contains no energy. It is simply proof that one MWh of renewable electricity was generated.

But that proof has value because electricity retailers are legally required to surrender a certain number of these certificates each year.

That obligation creates demand. Demand creates a price.

So the total revenue of a wind farm historically looked like:

Total revenue ≈ Spot price + LGC price

And for many years, that second term was enormous.

The Scarcity Machine

The Renewable Energy Target was designed as a scarcity engine:

• Set a fixed annual renewable target.

• Force retailers to buy certificates.

• Let scarcity drive price.

• Let price drive investment.

In the mid-2010s, that worked exactly as intended. LGC prices rose toward $80–$90.

At those levels, certificate revenue could rival electricity revenue. Wind farms were financed on the back of that policy-created scarcity.

And then something interesting happened - the capacity got built.

By 2021, the 33,000 GWh annual large-scale target had been met.

The machine had done its job, but the machine didn’t switch off. It kept issuing certificates — only now they weren’t scarce.

The Price Collapse

Once supply exceeds a fixed target, certificates stop being rare. Over 2024–2025, LGC prices slid hard.

From the $40s, to the $20s. And by early 2026, they were trading around $4–$6.

And that tells you something important:

• The policy worked.

• The investment signal did its job.

But once the target was met and capacity abundant, the extra revenue stream largely evaporated.

Which leads to the an obvious question - why new federal government schemes (CIS & REGO)? But first, who pays for the current system..

The Bill: Who Pays?

The government doesn’t pay for LGCs. Grid operators don’t pay for LGCs. Consumers and businesses pay.

The money flow:

1. The Government sets a Renewable Power Percentage (RPP)—the fraction of total energy a retailer must “cover” with certificates (currently ~19%).

2. The Retailer (AGL, Origin, etc.) must buy these LGCs on the open market.

3. The Consumer: The retailer adds a “Green Charge” to the bill to recover this cost.

In 2016, this “Green Tax” was a significant 15% of the SA wholesale electricity cost. Today, at less than $1 per MWh, it is a rounding error.

If Renewables Are Cheap… Why Another Scheme?

A lot of commentary talks as if “wind and solar are cheap now” means “problem solved.”

But that’s mixing up two different things:

• Cheap energy, sometimes (midday solar / windy hours)

• versus cheap electricity, always (the thing households and factories actually require)

Solar is cheap when it shows up — and then it disappears right on schedule. Wind is cheap when it’s windy — and then it may not be. The grid doesn’t care about annual averages; it cares about the tight hours, especially the shoulder period we’ve been looking at in South Australia.

That’s the punchline from the earlier articles:

• Midday prices go negative because solar overwhelms demand.

• Evening prices spike because solar vanishes and wind often can’t fully cover the gap.

• The scarcity signal has pulled in batteries — not to “run the state overnight,” but to act as a fast bridge while slower plant and imports ramp.

So the real shift isn’t “renewables are cheap.” The real shift is:

We built a lot of cheap intermittent energy, and now the expensive part is making the system work when that energy isn’t there.

LGCs were designed for the first phase: get renewable energy built. They created a second revenue stream that made projects bankable, and—by their own success—they eventually destroyed their own scarcity value.

Once LGC prices collapsed, the question stopped being “how do we incentivise wind and solar?” and became more like:

How do we incentivise the missing half of the system — the stuff that makes renewables usable at the hours that matter?

That’s where CIS seems to fit.

Not as “incentivising a new renewable technology that doesn’t exist,” but as a way to underwrite investment in the system-level complements we now need: dispatchable storage, firming, transmission-linked renewables in the right places, and capacity that can be relied upon through the shoulder.

Whether CIS is the best way to do that (or whether the market would have solved it anyway, given enough time and pain) is a bigger question.

For this article, the point is narrower:

• LGCs helped build the renewable fleet.

• The fleet changed the shape of prices (cheap/negative at noon, expensive at sunset).

• And the next wave of policy is mostly about integration and reliability, not “inventing renewables again.”

I’ll leave CIS and REGO as open questions — because the goal of this article is to explain the LGC mechanism — but now we can at least see why governments feel pressure to add new tools even after “renewables went mainstream.”

Note - What CIS Actually Does

The Capacity Investment Scheme (CIS) isn’t a certificate like LGCs, and it isn’t a simple “$X per MWh” subsidy.

It’s revenue underwriting, awarded through competitive tenders. Projects bid in, and winners sign a long-term contract with a revenue collar:

• Below a floor: government tops up most of the revenue shortfall.

• Above a ceiling: the project shares part of the upside back.

So: a seatbelt on the downside, and profit-sharing on the upside. It stabilises financing without turning projects into guaranteed-return utilities.

It’s also highly competitive (not central planning):

• CIS Tender 1 — NEM generation (target 6 GW): 119 registrations → 84 bids → 19 winners, 6.4 GW awarded.

• CIS Tender 3 — NEM dispatchable (target 16 GWh / 4 GW 4-hour equivalent): 166 registrations → 124 bids → 16 winners, 4.13 GW / 15.37 GWh awarded.

Example: the Goulburn River Standalone BESS (Lightsource bp) — 450 MW / 1,370 MWh — was one of the successful Tender 3 projects.

How can something be so abundant that the seller pays the buyer?

But negative prices in the NEM aren’t a glitch. They’re a feature: a way to clear a market where supply and demand must match every instant, and where some generators can’t (or rationally won’t) turn off just because lunchtime is sunny.

In this article I want to build a simple intuition:

• Why prices go negative

• Why coal and gas sometimes prefer paying to keep running

• Why wind and solar can bid negative and still come out ahead

• Why this can be “beneficial” (or at least rational) for everyone involved

1) First principles: electricity is perishable

Most commodities have storage. If the price collapses, you pile your widgets in a warehouse and wait for a better day.

Electricity doesn’t do that. The grid is not a warehouse. It’s a tightrope.

When the system has too much generation (relative to demand + export capability + storage), the market needs a signal that says:

“Please stop producing. Now.”

In the NEM, that signal can go all the way to the market price floor of -$1,000/MWh. Negative prices exist precisely to create a strong disincentive to generate when the system is oversupplied.

And the mechanism is brutally simple: AEMO dispatches the lowest-priced offers first. If you bid lower than everyone else (including into negative territory), you’re more likely to be dispatched.

That’s why you see negative bidding at all: bidding negative is often a way to avoid being turned down.

2) Why negative prices have exploded (especially at midday)

AEMO’s own summary of the modern pattern is basically the entire story in one sentence: negative prices are highest during daytime hours when operational demand is low because distributed (rooftop) PV is high, large-scale wind/solar output is strong, and coal plants maintain minimum stable output.

In Q4 2025, South Australia recorded negative prices in 48.4% of all dispatch intervals (the highest of any region).

This is the “negative price canyon”:

• Rooftop PV suppresses grid demand hard at midday

• Utility-scale solar and wind may also be strong

• Interconnectors and system constraints limit how much surplus can be exported

• Some thermal generation elsewhere can’t or won’t fully back off

• The market clears by pushing price negative to force someone to blink

3) The three families of “why would you pay to generate?”

A) Coal: “pay a little now, or pay a lot later”

Coal generators typically have high startup costs. Turning a large unit off and back on isn’t like flicking a light switch; it’s a process with fuel, wear, and risk.

So a coal plant can face a choice:

• Stay online at minimum stable load through the solar glut, even if prices are negative

• Or shut down and pay the startup penalty (plus higher forced-outage risk, maintenance, etc.)

When negative pricing lasts a limited number of hours, it can be cheaper to “ride it out” and pay the market than to stop and restart. The AER makes this point directly: paying to continue generating for a limited period can be more cost-effective than being switched off and restarting.

South Australia no longer has coal plants, but SA is still coupled to the broader NEM through interconnectors, and a large fraction of imported energy can be coal-fired (in my NEMWeb analysis for 2025 it was roughly two-thirds). When those coal units in Victoria and NSW “choose” to stay online, that behaviour leaks into regional pricing—especially when SA can’t export surplus and local demand is low.

B) Gas: more flexible than coal… but still not free

Gas generators are usually more flexible than coal, but they still face:

• startup and minimum-run constraints (especially combined-cycle plant)

• efficiency penalties at low output

• cycling and maintenance costs

• the operational need to be “ready” for the evening shoulder

The AER also notes that some gas generators increasingly avoid short on/off cycles by offering at very high prices when they do not plan to run, because fluctuations in wind/solar can otherwise require them to come on for short periods. That’s the mirror image of negative bidding: the market is pricing the value of not being jerked around.

So gas can behave in two ways during oversupply:

• Some units back off (or don’t start).

• Some choose to stay on at low output and accept some negative pricing if the alternative is worse.

C) Wind and solar: the “side-payment” that makes negative bids rational

Wind and solar have marginal costs near zero, which means they’re happy to generate whenever they can—unless the price becomes negative enough to overwhelm their other revenue.

Two big “other revenues” matter:

1. Renewable Energy Certificates (LGCs)
   Under Australia’s Renewable Energy Target, eligible renewable generators can create one LGC for each MWh produced and sell those certificates.
   Liable entities (mostly retailers and some large users) must acquire and surrender LGCs each year, which is what creates demand for them.

2. Contracts / PPAs
   Many wind/solar projects also have contracts that reduce their exposure to spot prices. The AER notes that some wind and solar generators source revenue from PPAs and the sale of renewable certificates, and may operate profitably even when wholesale prices are negative.

Put those together and you get a very practical rule-of-thumb:

If your “side revenue” is worth +$X/MWh, you can bid down to roughly -$X/MWh and still be whole (before other costs/clauses).

That’s the heart of “why would a solar farm pay to generate?”
Because the spot price is not the only thing they’re being paid.

4) Why negative prices got less deep in late 2025: the LGC link

Here’s a subtle but important point: negative prices can become more frequent while becoming less extreme.

AEMO observed exactly this in Q4 2025: 86% of negative prices were between -$30 and $0, up from 33% a year earlier. AEMO links this shift to a collapse in LGC prices: LGCs averaged about $9/certificate in Q4 2025 vs $34/certificate in Q4 2024.

Lower LGC value means renewable generators have less incentive to tolerate very negative prices. They curtail sooner, or bid less negative, and the market doesn’t dive to the floor as often—even if oversupply remains common.

The Clean Energy Regulator’s own quarterly carbon market reporting also describes LGC spot prices falling as supply increases (for example, spot prices fell to $10.80 from $16.50 over the September quarter 2025, with forward prices tracking down).

So LGCs don’t just “subsidise renewables”; they actively shape the depth of negative prices.

5) “Can wind and solar just turn off?”

Mostly yes—technically it’s easy:

• Solar farms can curtail output quickly via inverter control.

• Wind farms can pitch blades / feather and reduce output quickly.

So why don’t they always curtail the instant price goes negative?

Because curtailing is not free in their business model. Curtailment means giving up:

• LGC creation for those MWh

• contracted revenues tied to generation (depending on contract design)

• and sometimes operational preferences (frequent cycling / ramping constraints / compliance details)

As the AER puts it: wind and solar may bid negative to “guarantee dispatch” when conditions are optimal and competition for dispatch is intense, and some can still be profitable with certificates and contracts even at negative spot prices.

6) So is this “beneficial for everyone”… or just rational?

“Beneficial” depends on what you mean.

But it’s rational for each group:

• Coal and some gas: negative prices can be cheaper than shutdown/restart + wear + risk.

• Wind/solar: negative spot prices can still be profitable when certificates/contracts are included.

• The market/system: negative prices are how the NEM clears surplus supply and signals that flexibility (storage, demand response, transmission, curtailment capability) is valuable. The market price floor exists specifically to help clear low-demand/surplus conditions.

The deeper punchline is that negative prices are the shadow of a system learning how to run on variable generation. They’re not proof renewables “don’t work”; they’re proof the grid is running into a new constraint:

Not “can we generate enough?”
but “can we control enough?”

In the next article I’ll zoom in on the LGC machinery: what it is, who pays for it, why the value collapsed, and why that one policy lever quietly changes real-time price behaviour.

Notes

Note A — Price floors aren’t a “subsidy”:
The market floor (-$1,000/MWh) is a rule setting that allows the market to clear surplus supply and signals oversupply. It’s discussed explicitly in the AER’s market overview and in the NEM wholesale market settings review.

Note B — SA imports coal share:
The “~two-thirds of SA imports coal-generated in 2025” figure is calculated from NEMWeb. It’s not a published official statistic.

But in the world of electricity markets, an hour is a long time — because the market doesn’t buy “a day of energy”. It buys control, ramping, and fast response, sliced into tiny time-intervals.

This article is about the Market Machinery: the rules and protocols that turn physical electricity into a high-speed auction where speed itself is valuable.

1) The Five-Minute Revolution: “Baking and selling by the slice”

The single most important market reform of the last decade is Five-Minute Settlement (5MS).

Before 2021, the NEM dispatched generators every five minutes, but paid them on a 30-minute average. Under that setup, a five-minute price spike could be diluted into irrelevance.

From October 2021, dispatch and settlement aligned at five-minute intervals. [1]

Why this matters:

• Batteries can hit full output extremely quickly once they’re already charged and enabled. That means they can position themselves to show up for short, high-value intervals.

The 5-minute settlement (5MS) repriced flexibility: fast assets can monetise short spikes; slow assets need longer sustained conditions to benefit. [1]

2) FCAS: the high-paid twitch

Most people think batteries make money by buying low and selling high.

Sometimes, yes. But the real surprise is that batteries also get paid for being the grid’s fast-twitch stabiliser — Frequency Control Ancillary Services (FCAS).

The grid has a heartbeat: 50 Hz. If a big generator trips or an interconnector constraint bites, frequency can deviate. FCAS markets pay providers to correct frequency — some in seconds, some in minutes. Batteries are extremely good at this because they can change power output almost instantly.

The “Very Fast” FCAS markets (1-second services)

Two new FCAS markets — Very Fast Raise and Very Fast Lower — commenced on 9 October 2023. [2]
These are the “blink and the grid notices” services.

And this is where SA got a spectacular demonstration of how markets reward speed:

• In August 2025, volatility in South Australia’s Lower 1-second FCAS market was so extreme that Modo Energy summarised it as: SA batteries earned ~$43k per MW from FCAS in that month. [3]

• Their summary also notes prices spiking to the market price cap (and even triggering administered pricing in that market episode). [3]

That’s not a steady-state income stream. It’s a “weather event” for market revenues: when network outages and constraints force FCAS to be sourced locally, small markets can go non-linear very quickly.

FCAS isn’t always “the big money”

Sometimes FCAS dominates; sometimes energy arbitrage dominates.

AEMO’s Quarterly Energy Dynamics shows this clearly at the NEM level:

• Q3 2025: battery net revenue rose to $111.9m, with FCAS revenue $53.7m (about half the revenue stack). SA batteries accounted for $47.4m of that FCAS revenue (88% of the NEM total that quarter). [4]

• Q4 2025: battery net revenue was $70.4m, with energy arbitrage $61.4m and FCAS $9.0m (FCAS only 13%). [5]

So the best mental model is:

• Batteries are option sellers. They sell the option of fast response.

• Some quarters the option is worth a fortune; other quarters it’s just steady arbitrage.

3) Negative prices: abundance, incentives, and the “self-curtailment reflex”

By late 2025, South Australia spent a shocking share of time at zero or negative prices.

AEMO reports that in Q4 2025, SA recorded negative prices in 48.4% of all intervals (highest in the NEM). [6]

But the shape of negative prices matters as much as the frequency.

AEMO also reports that in Q4 2025:

• 86% of negative prices were between -$30/MWh and $0/MWh, far “less extreme” than earlier periods, and

• this shift coincided with much lower LGC prices (average $9/certificate in Q4 2025 vs $34 in Q4 2024). [6]

That’s the key: negative prices aren’t just “solar did it.” Negative prices are an equilibrium produced by:

• rooftop PV crushing operational demand,

• wind/solar output variability,

• minimum stable levels for thermal plant (in other regions),

• and the economics of certificates and contracts.

I’ll explain more about the LGC in the next article.

Do solar farms have an “automatic off-switch”?

Not as a market rule.

The market allows prices to fall to the market floor (and the floor exists for a reason), but the practical behaviour is driven by plant controls and bidding strategy, not a magical safety catch. [Note A]

What is true in practice:

• Most utility-scale renewable assets now use some mix of automated bidding, dispatch-target tracking, and inverter control logic to reduce output when prices become sufficiently negative.

• The threshold at which it becomes rational to curtail depends on the asset’s situation: certificate value, contract terms, marginal cost, and the cost (and rules) of stopping/starting output. The “- $10/MWh” type thresholds are not universal.

So the “negative price canyon” isn’t a death trap so much as a sorting mechanism:

• It forces generators to reveal whether their incentive is “must run” (contracted/technical) or “only run if price makes sense.”

An important point - batteries like negative prices because charging becomes cheap (or paid), but batteries are limited by energy capacity and state of charge — they can’t absorb infinite surplus.

4) The final piece: from spinning iron to digital inertia

Historically, SA needed multiple synchronous machines online — “spinning iron” — to provide inertia, fault current, and other services that make the grid behave.

By 2 September 2025, AEMO and ElectraNet reduced the minimum synchronous generator requirement in South Australia from two units to one, when system conditions allow. AEMO also explicitly notes that under certain conditions (low demand, separation risk, lack of fast start, credible contingencies), two synchronous generators may still be required. [7]

This matters for two reasons:

1. It reduces the amount of gas generation that must run “just in case”, which reduces curtailment and directions costs. [7]

2. It opens the door for inverter-based resources (including batteries) to provide more of the system services historically bundled into synchronous plant.

Grid-forming inverters (what they are, and what they change)

A grid-following inverter needs a stable grid waveform to follow. A grid-forming inverter can create a stable waveform and contribute to system stability more directly.

Blyth Battery is often cited here, and it’s a useful example — but the most accurate description is:

• Blyth was built and entered commercial operations initially in grid-following mode, and is progressing through work to transition/retrofit grid-forming capability (including changes to performance standards and additional equipment). [8]

• Neoen also describes Blyth as equipped with grid-forming inverter technology, supported by ARENA, to provide stability services traditionally provided by synchronous generation. [9]

The point isn’t “batteries replace everything.” The point is: a growing part of the grid’s stability toolkit is becoming software-defined.

5) So how much money are we talking about? (SA battery revenue: energy + FCAS)

AEMO’s reports estimated net revenue for NEM (= “National Electricity Market”, covering Australia excluding WA & NT) grid-scale batteries of:

• Q2 2025: $130.6m net (energy arbitrage $120.8m, FCAS $9.8m) [10]

• Q3 2025: $111.9m net (energy arbitrage $58.2m, FCAS $53.7m) [4]

• Q4 2025: $70.4m net (energy arbitrage $61.4m, FCAS $9.0m) [5]

What we know specifically for SA

In Q3 2025, SA batteries contributed $47.4m of battery FCAS revenue (88% of the NEM battery FCAS total that quarter). [4]
That’s a huge number for one region — and it’s exactly the kind of “local constraint + scarce service” event that makes batteries look like money printers.

A reasonable SA estimate for calendar 2025 (flagged as an estimate)

Because AEMO reports the full energy/FCAS stack NEM-wide (and only sometimes breaks out regional contributions in the text), the honest way to estimate SA’s 2025 totals is:

• FCAS: We have $47.4m in Q3 alone, plus additional FCAS earnings in other quarters. A conservative annual estimate is ~$50–60m FCAS revenue for SA batteries in calendar 2025, recognising Q3 was exceptional. [4][5]

• Energy arbitrage: NEM energy arbitrage net revenue was very large in Q2 and substantial in Q3–Q4. SA’s share of this will depend on fleet size and operation (and it changed fast through 2025). A reasonable ballpark is ~$50–80m of energy-arbitrage net revenue for SA batteries across 2025.

That puts a plausible SA total around ~$100–140m net revenue for battery services (energy + FCAS) in 2025. This is revenue to the battery fleet, not “profit after capital costs”, and it will not be evenly distributed across assets.

Conclusion

The market isn’t rewarding batteries because someone decided “batteries are the future” at a press conference.

It’s rewarding batteries because the market buys what the grid physically needs most during stressed intervals: fast response, fast control, and (increasingly) fast stability services.

South Australia doesn’t have “a day’s worth of batteries.” It has just enough storage and inverter capability to be very profitable in the shoulder periods — and, sometimes, a high-paid twitch in the FCAS markets.

In the next article, I’ll explain the bidding mechanics in more detail — including the practical meaning of negative prices (“why supply power when you have to pay?”), and how constraints and FCAS procurement create the conditions for battery windfalls.

Notes and References

Note A — Market limits and price caps:
Market prices can fall to a market floor and rise to a market cap, and those values change over time. For example, the market price cap increased to $20,300/MWh from 1 July 2025. [11]

[1] AEMO: Five Minute Settlement and Global Settlement (5MS commenced 1 Oct 2021)

https://www.aemo.com.au/initiatives/trials-and-initiatives/past-trials-and-initiatives/nem-five-minute-settlement--program-and-global-settlement

[2] AEMO: Fast Frequency Response / Very Fast FCAS market commencement (opened 9 Oct 2023)

https://www.aemo.com.au/initiatives/major-programs/nem-reform-program/nem-reform-program-initiatives/fast-frequency-response

[3] Modo Energy: “South Australian BESS earn $43k/MW from FCAS in August 2025”

https://modoenergy.com/research/en/australia-nem-battery-energy-storage-revenues-august-2025-fcas-price-volatility-south-australia

[4] AEMO Quarterly Energy Dynamics Q3 2025: battery net revenue $111.9m; FCAS $53.7m; SA batteries $47.4m FCAS

https://www.aemo.com.au/-/media/files/major-publications/qed/2025/qed-q3-2025.pdf

[5] AEMO Quarterly Energy Dynamics Q4 2025: battery net revenue $70.4m; energy arbitrage $61.4m; FCAS $9.0m

https://www.aemo.com.au/-/media/files/major-publications/qed/2025/qed-q4-2025.pdf

[6] AEMO QED Q4 2025: SA negative prices 48.4% of intervals; 86% of negative prices between -$30 and $0; LGC avg $9 in Q4 2025

https://www.aemo.com.au/-/media/files/major-publications/qed/2025/qed-q4-2025.pdf

[7] AEMO: 2025 Transition Plan for System Security — reduction from 2 to 1 minimum synchronous generators in SA (conditions apply)

https://www.aemo.com.au/-/media/files/major-publications/tpss/2025-transition-plan-for-system-security.pdf

[8] ARENA / Neoen Blyth Battery lessons learnt report (status: GFL operations; progressing GFM transition/retrofit)

https://arena.gov.au/assets/2026/01/Neoen-Blyth-Battery-Lessons-Learnt-3.pdf

[9] Neoen press release: Blyth Battery and grid-forming inverter technology (ARENA support)

https://neoen.com/en/news/2024/neoen-completes-joint-financing-of-the-2nd-tranche-of-goyder-south-wind-farm-and-blyth-battery-in-australia/

[10] AEMO Quarterly Energy Dynamics Q2 2025: battery net revenue $130.6m; energy arbitrage $120.8m; FCAS $9.8m

https://www.aemo.com.au/-/media/files/major-publications/qed/2025/qed-q2-2025.pdf

[11] AER: Q2 2025 high electricity price events (notes MPC increased to $20,300/MWh from 1 July 2025)

https://www.aer.gov.au/news/articles/communications/aer-reports-q2-2025-high-electricity-price-events

The Problem of Speed

When the sun sets and South Australia’s 2.5 GW rooftop solar fleet vanishes, the grid faces a massive “ramp”.

• The “Slow” Guard: Large gas units in SA and coal stations in Victoria or NSW (linked via the Heywood and the new 150 MW Stage 1 PEC) often cannot throttle up fast enough to catch 5-minute spikes.

• The Maintenance Tax: Forcing thermal plants to ramp up and down aggressively increases maintenance costs, which generators bake into their bids as a “risk premium”.

As the early evening “shoulder period” begins, the market sends a desperate price signal. Batteries are the only participants fast enough to answer.

The Signal vs. The Hype

The general population, often guided by journalists repeating business and government press releases, believes that because batteries are getting “cheaper by the day,” they are the total solution for the night.

In reality, utility-scale batteries remain expensive to build. However, the extreme market spikes—which reached over $5,000/MWh seven times on a single July day in 2025—create a financial environment where even an expensive battery can turn a massive profit in just 30 minutes.

Watching the Market Breathe

The following data shows the reality of battery deployment. First, the change 2015 - 2025 in supply by generation type onto the grid:

The battery share by time of day over the same period:

The changes just in 2025 by quarter:

And an animation of the 2015-2025 change:

As you can see, the total supply from batteries remains a small percentage of the total energy mix, but it is precisely targeted. By late 2025, batteries provided nearly 9% of the total grid supply during the critical 6:30 PM peak in some quarters.

The “One Hour Tank”

While the growth in capacity is impressive, we must look at the volume of energy stored.

South Australia’s total electricity consumption for 2025 is estimated at 16.8 TWh, or approximately 46 GWh per day. On average, the state requires roughly 1.9 GWh of energy every hour.

At the end of 2025, after a historic 1.0 GWh surge in utility-scale installations (led by the 477 MWh Blyth Battery and 330 MWh Templers BESS), South Australia finally has enough stored energy to power the wholesale grid for a little over an hour.

This “one-hour tank” isn’t a failure—it’s a feature. These assets aren’t trying to replace the night; they are trying to “kill” the evening price spike.

Next Step: In the final article of this series, we will explore The Market Machinery. We’ll explain why solar farms sometimes pay to supply the grid, how the 5-Minute Settlement rule changed the game, and why the most important thing a battery does isn’t actually supplying energy, but providing “synthetic inertia.”

Notes

Battery additions by year

• 2017 (0.129 GWh): Hornsdale Power Reserve initial build (129 MWh).

• 2018 (0.008 GWh): Dalrymple ESCRI (8 MWh) commissioned/operational in 2018.

• 2019 (0.052 GWh): Lake Bonney BESS (52 MWh) energised/registered in 2019.

• 2020 (0.065 GWh): Hornsdale expansion (+64.5 MWh).

• 2023 (0.250 GWh): Torrens Island BESS (250 MWh) commissioned in 2023.

• 2024 (0.084 GWh): Tailem Bend Battery Project (84 MWh) with BESS COD Q4 2024.

• 2025 (1.007 GWh):

  • Blyth BESS (477 MWh) commissioned April 2025.

  • Mannum BESS (200 MWh) commercial ops Oct 2025.

  • Templers BESS (330 MWh) commercial ops Dec 2025.

Here’s an graph that shows the change as SA has embraced more renewables - the annual average wholesale price vs time of day for 2015, 2020 and 2025:

And - a bit more cluttered - every year from 2015-2025:

This is the “Pricing Paradox”: how “free” fuel from the sun and wind has created a market of extreme highs and lows that rarely results in a “cheap” bill for the average household.

Here’s a video, showing the same progression from 2015-2025:

The annual average wholesale price:

1. The “Canyon Curve”

In 2025/26, South Australia’s curve has evolved into what traders call the “Canyon Curve.”

• The Floor (Midday): On many days, solar output is so high it exceeds the state’s total demand. In Q4 2025, South Australia recorded negative prices in 48.4% of all intervals.

• The Walls (The Ramp): At 5:00 PM, as solar vanishes, the grid must “ramp up” thousands of megawatts of alternative power in minutes.

• The Peak (Evening): Because South Australia has no coal, this evening gap is filled by Natural Gas and Batteries. Because gas is an international commodity, it is expensive. When supply is tight, gas generators bid at the market cap (up to $16,600/MWh).

The annual average market price is around $100/MWh. Here are the top ranked prices - you can see how high the market extremes are - more than 100x the average price, and how much each of them impact the annual price:

Here’s the graph of the prices in 2025 across the year:

2. The “Gas Marginal” Effect: Why the Last Megawatt Wins

The most important rule of the Australian market (and most market systems) is that the last generator needed to meet demand sets the price for everyone.

• During the day, the “last generator” is often a solar farm bidding $0.

• During the evening, the “last generator” is almost always a Gas Peaking Plant.

• Even if 90% of the energy in a given hour is “cheap” wind or battery power, if that final 1% comes from an expensive gas plant bidding $300/MWh, then every generator in the state gets paid $300/MWh.

3. The “Shadow Costs” You Don’t See

While wholesale prices are the “celebrities” of the energy news, they only make up about 35-40% of your retail bill. The rest is made up of “Shadow Costs” that actually increase as the grid becomes more complex.

• FCAS (Frequency Control): As the grid loses the “natural stability” of spinning coal turbines, it must pay batteries and gas plants to stand by and “twitch” the grid back to 50Hz. In volatile years like 2024, these “System Strength” costs added millions to the state’s total energy bill.

• Network “Gold-Plating”: To handle the massive exports of solar and the new interconnectors like Project EnergyConnect, the poles and wires must be upgraded. These are “fixed costs” that show up on your bill regardless of whether the wholesale price is negative.

4. The Retailer’s “Fear Premium”

You might wonder why your retail bill isn’t $0 if the wholesale price is negative half the time. Retailers (like AGL or Origin) act as Shock Absorbers. They buy power at the volatile 5-minute spot price but sell it to you at a flat, stable rate (e.g., 40c/kWh).

• To survive a single hour of $15,000/MWh (which occurred 11 times in early 2025), a retailer has to charge a massive “Insurance Premium” in their flat rate.

• In February 2026, despite “record low” wholesale averages in late 2025, Adelaide households are still paying roughly $1,800 – $2,200 a year—among the highest in the developed world.

In the next article we’ll look a little deeper into the reasons for the high evening price, why grid suppliers will pay to supply energy at certain times, and how the market mechanism has created the demand for (expensive at grid-scale) batteries.

Genuine progress, and also some excellent demonstrations of the challenges of increasing renewables.

1. The Two-Lever Advantage: Wind AND Sun

One reason South Australia has outpaced its peers like Germany and the UK is its unique geographic “two-lever” advantage. While northern Europe has world-class wind, its high latitude makes solar a weak second player.

• Germany & UK: Rely heavily on wind. Germany’s solar capacity factor is roughly 10% (980 full load hours/year). During “wind droughts,” they often fall back on coal or gas because the sun simply isn’t strong enough to carry the load.

• South Australia: Boasts irradiance levels comparable to the Middle East or Spain, with solar capacity factors of 25–32%.

• The Mix: In 2024–25, SA generated 44.5% of its power from wind and 32.5% from solar (22.4% rooftop, 10.1% utility). When the wind stops, the sun is actually powerful enough to contribute a massive chunk of the load, and vice versa.

2. The Nameplate Paradox: Built for 200% Capacity

The sheer scale of the infrastructure is staggering when compared to the state’s actual demand.

• Peak Demand: The all-time record set in January 2025 was 4,486 MW. Normally, a “busy” summer day peaks at around 3,274 MW.

• The Fleet: SA has built a renewable fleet that could theoretically power two whole South Australias at once.

  • Wind Nameplate: ~2,341 MW (with Goyder South expanding this further).

  • Solar Nameplate: ~3,500 MW (Combined rooftop and utility-scale).

• Total Capacity: At over 5,800 MW, the renewable nameplate is roughly double the average daily demand.

This leads to a phenomenon unique to SA: Negative Demand. On October 19, 2024, the state hit a record low operational demand of -209 MW. Households weren’t just using their own solar; they were generating so much excess they were powering the state’s high-voltage lines and the interconnector to Victoria.

Here’s one part of the story from 2015 to 2025:

Imports, primarily from Victoria, are important, but haven’t increased over a decade:

Exports:

3. The Coal Umbilical Cord

Despite its “100% renewable” boast, SA is not an island. It remains physically “strapped” to the massive, coal-heavy grids of Victoria and New South Wales.

• The Large Neighbor: While SA generates 75% renewables, its neighbors still rely on coal for over 50% of their mix.

• The Sync: When SA’s wind and solar drop, it survives by importing “dirty” power through the Heywood Interconnector. In 2024-25, SA imported 668 GWh from Victoria to fill these gaps.

• The Safety Valve: This is why SA is building Project EnergyConnect (PEC) to NSW. By having a second “umbilical cord” to another coal-rich state, SA can export more of its midday solar glut (to avoid wasting it) and import more reliably during the evening peak.

AEMO’s SA Electricity Report (SAER) makes the trade story explicit: net imports from Victoria fell from 927 GWh (2023–24) to 668 GWh (2024–25) as exports increased.
In other words, SA’s interconnectors are not merely “backup”—they’re also the route by which SA exports surplus wind/solar.

That’s a big reason SA can run such high renewable shares without curtailing everything into the ground: when SA has excess, it can often push it out.

However, imports are very important. Here is November 2025. The orange line is imports as % of demand - it was often above 40%:

Note - at certain times demand on the SA grid is extremely low due to rooftop solar, and in these very low times the import % can be misleading. So the graph was produced excluding imports during the lowest 30% of demand.

4. The Grid’s New “Heartbeat”: Batteries & Physics

Without coal turbines spinning, the grid loses its natural “inertia”—the momentum that keeps the lights from flickering when something goes wrong.

• The Shift: In 2024–25, battery developments dominated the grid, adding approximately 600 MWh of new storage.

• The “Engines Off” Milestone: By late 2025, AEMO confirmed it could run the SA grid with “engines off”—meaning zero gas turbines online—by using a combination of batteries and synchronous condensers to maintain grid stability.

In the next article we’ll see how solar power crashes the wholesale price and also massively increases the wholesale price. Two sides of the same solar coin.

Germany increased wind and solar hard — but also saw some de-industrialisation (industry moved) and a big change in how much electricity it traded with neighbours. The UK turned coal off, leaned heavily into offshore wind… and increased the burning of woodchip. Both countries also leaned more on imports — especially French nuclear-dominated exports and Norwegian hydro.

Today, we look at the bill.

This article is the “follow the money” version: how consumer electricity pricing changed from 2015–2024 in Germany, the UK and France — and what drove it.

France is the “control group” because it didn’t do a Germany/UK-style renewables surge. It’s a nuclear-heavy system with long-lived plants, a very different political relationship to retail prices, and (crucially) a different habit of shifting costs off the bill and into the state budget / EDF.

The spruikers of the energy transition often promise that because wind and solar have zero fuel cost, electricity prices will inevitably fall. However, a look at the data from 2015–2024 for the three major European powers—Germany, the UK, and France—tells a different, and more complex story.

1. The Physical Reality: Generation Mix (2015–2024)

Before looking at the price, let’s review what is actually keeping the lights on.

The UK Story: A total exit from coal, replaced by a massive surge in wind and a controversial reliance on biomass (wood pellets). The “hidden” reality here is the UK’s increasing reliance on imports from the continent:

The German Story: The Energiewende in full view. High wind and solar growth, but look closely at the total TWh—it is shrinking. Germany isn't just "greening" its grid; it is de-industrializing. In 2023/24, Germany became a net importer as it shuttered its final nuclear plants. Importing French nuclear generated electricity..

The French Story: The control group. France’s mix is dominated by the green "Nuclear" bars. While wind and solar are growing, they are a "cap" on a massive baseload foundation. Note the 2022 "dip"—the year France’s nuclear fleet suffered systemic corrosion issues, forcing them to import and driving European prices to record highs.

2. The Financial Reality: Deconstructing the Bill

To understand the pricing, we have to look inside the unit rate. These are broken down into three layers:

1. Wholesale (power plants etc): The market price for power.

2. The Balance (The Grid): The “Blue Layer”—grid fees, redispatch costs, and supplier operations.

3. Extra Costs (The Taxes): Policy levies for renewables, social schemes, and VAT.

For each country we can see the costs in local currency but also in US$.

UK:

The UK in US$:

Germany

Germany in US$:

And France - 2022 was an anomaly where the state provided huge subsidies:

France in US$:

As a minor comment, I was somewhat surprised. France has significantly lower retail prices, but not as much a gap as I expected. Note that 2022 was an anomaly year due to the “France nuclear guarantee” not being available for France, or for Germany or the UK.

3. What is Driving the Pricing?

The data reveals that the “Retail Unit Rate” is now a political and engineering construct, not a simple market price.

Germany: The Tax-funded Mirage

Germany’s “Extra Costs” (Red) dropped significantly in 2023/24. This was not a saving; it was a re-indexing. The government abolished the EEG-Umlage (renewable surcharge) from the bill and moved it to the federal budget.

• The Hidden Item: You no longer pay for solar subsidies on your electricity bill; you pay for them via CO2 taxes on your heating oil and petrol.

• The Grid Explosion: Notice the “Balance” (Blue) layer expanding. This is the physical cost of the transition—”Redispatch.” Because wind is in the North and industry is in the South, Germany pays billions to tell wind farms to turn off and gas plants to turn on to keep the grid from collapsing.

The UK: The “Spark Gap” and the Shadow Grid

The UK has some of the highest electricity prices in the world.

• The Taxation of Electricity: The UK loads ~82% of all social and environmental levies onto electricity, while gas remains relatively untaxed. This creates the “Spark Gap,” making “clean” electric heating nearly four times more expensive than gas.

• The Capacity Market: Hidden in the “Balance” layer is the “Shadow Grid”—over £1 billion a year paid to gas and coal plants just to stand still and wait for the wind to stop blowing.

France: The Political Shield

France’s graph shows a unique phenomenon in 2022: a negative balance.

• The “Tariff Shield”: When wholesale prices (Orange) exploded in 2022, the French government simply froze the retail price. The state-owned utility, EDF, absorbed a €17.9 billion loss to protect consumers.

• The Return to Reality: In 2024, the “Red” and “Blue” layers are growing again as the government restores taxes to pay off the 2022 debt. Even with these hikes, France remains cheaper than the UK and Germany because its nuclear “Wholesale” basement is structurally lower than gas-marginal markets.

4. Conclusion

If the promise of renewables was “cheap, abundant energy,” the 2015–2024 data suggests a different reality.

1. System Costs Overwhelm Commodity Savings: While the “Wholesale” cost of a windy day might be low, the cost of managing that variability (the Blue Layer) and the “legacy” subsidies of the past (the Red Layer) are driving the total price upward.

2. Decarbonization via Imports: Both the UK and Germany have “decarbonized” partly by importing French nuclear and Norwegian hydro. They have outsourced their grid stability to neighbors with firmer foundations.

3. The Nuclear Dividend: France, despite its 2022 technical crisis, maintains a more stable and lower-priced retail environment by avoiding the massive “Redispatch” and “Capacity” costs that plague its neighbors.

In the next article, we will look at South Australia.

But annual charts of “Public Net Generation” only tell half the story. To understand how a system can lose its “thermal basement” while reporting record-breaking renewable shares, we have to look at the three levers of the British “magic”: the Shrinking Denominator, the Subsea Safety Net, and the Carbon Loophole.

The Shrinking Denominator: Less “Made in the UK”

The most overlooked factor in the UK’s high renewable percentage is that the total pool of electricity generated within British borders has shrunk significantly.

• Generation vs. Demand: Between 2015 and 2024, total internal generation fell from 338.9 TWh to 285.0 TWh—a 16% contraction.

• The Demand Gap: While total demand also fell (by ~11% due to efficiency and industrial decline), it did not fall as fast as domestic production.

• The Statistical Effect: When you keep wind output growing but reduce the total “pool” of domestic power (by closing coal and nuclear), the percentage share of renewables rises automatically. The UK has essentially “shrunk its way” toward a greener-looking mix.

The Subsea Safety Net: A One-Way Energy Straw

As domestic firm capacity (coal and nuclear) declined, the UK built a massive infrastructure pivot to tether itself to Europe. Since 2015, the UK has more than doubled its interconnector capacity.

UK Interconnector Growth (2015–2024):

• 2015: 4.0 GW (IFA1, BritNed, Moyle, EWIC)

• 2021: 7.4 GW (Added Nemo Link, IFA2, North Sea Link)

• 2024: 9.8 GW (Added Viking Link to Denmark).

While the 2022 anomaly (shown in the chart below) saw the UK become a net exporter for the first time in 40 years—driven by French nuclear outages and a gas supply crunch—the long-term trend returned with a vengeance in 2024.

UK Electricity Imports by Country (2015–2024)

In 2024, net imports rose by 40%, with the UK’s net import dependency climbing to 43.8%. These interconnectors are not just “trading platforms”; for 92% of the hours in the year, they are a unidirectional straw pulling in French nuclear and Norwegian hydro to balance the UK’s variable wind.

Here are the imports and exports over 2015-2024:

Zooming In: May–June 2024 Hour-by-Hour

Annual bars show “how much,” but hourly data shows “how.” Looking at May and June 2024, we see the high-resolution dance of the interconnectors.

• France (Yellow): Provides the persistent “floor” of the import mix, taking advantage of the UK’s consistently higher wholesale prices—the highest in the IEA for industrial users.

• Norway (Light Blue): Acts as a surge protector, showing high volatility as it reacts to wind lulls in the UK.

• The Price of Connection: This reliance isn’t free. UK consumers pay an estimated £735 million premium annually just to “pull” this energy in over the market price in Europe.

Now imports in GW, and as a percentage of demand, at key times:

The Shadow Grid: Paying to Do Nothing

While the UK might celebrate the days wind provides 50% of their power, they also pay a separate bill—the Capacity Market—to ensure gas plants and interconnectors stay ready for the days it doesn’t.

• The Bill: For the 2024/25 delivery year, the UK will pay approximately £1.26 billion in capacity payments.

• The Beneficiaries: Gas remains the “principal form of UK generation,” setting the system price 97% of the time even when its share of generation falls.

• De-rating Reality: While a gas plant is paid for ~95% of its capacity, wind is “de-rated” to <10% in these auctions because it cannot be trusted to deliver during a still, cold winter evening (note 1).

The Carbon Loophole: Biomass and “Zero” Emissions

The final pillar of the UK renewable story is Biomass, which provided nearly 14% of the mix in 2024.

• The Account: Drax is the UK’s single largest emitter, releasing 13.3 million tonnes of CO2 annually.

• The Zero: Under current accounting, this is reported as 0g CO2/kWh.

• The Physical Reality: Scientific analysis suggests the Carbon Payback Period—the time it takes for a replanted forest to re-absorb those emissions—is between 44 and 104 years.

Physically, the UK is “borrowing” carbon from the next century to report a clean grid today. In the atmosphere’s ledger, those 13.3 million tonnes are very much present.

Summary

The UK’s transition is a masterpiece of complex engineering and creative accounting. They have replaced domestic coal with:

1. Variable Wind (The headline stat).

2. A Subsea Safety Net (£8bn in cables, record net imports).

3. A Shadow Grid of Gas (£1.26bn in standby fees).

4. A Biomass Loophole (A 100-year carbon debt).

The grid is cleaner, but it is not yet “self-reliant.” It is an island grid that has chosen to outsource its reliability to its neighbors.

Notes

De-rating vs. Capacity Factor: Not Just Rhetoric

This is a critical technical distinction used by the National Energy System Operator (NESO) to ensure the lights stay on.

• Capacity Factor (CF): This is a retrospective average. If an offshore wind farm has a CF of 45%, it means over a year, it produced 45% of its theoretical maximum.

• De-rating Factor (DRF): This is a prospective risk assessment. It asks: “If we have a one-in-ten-year system stress event (e.g., 6 PM on a freezing, windless Tuesday in January), how much can we statistically rely on this asset?”

The reason wind is “de-rated” to <10% (specifically ~7.5% to 12% for offshore) is that there is a high probability of “correlated wind lulls” across the whole UK during high-pressure winter cold snaps. In contrast, Gas plants are de-rated to ~90-95% because their availability is limited only by technical breakdowns, not the weather.

Crucial Point: A generator is only paid for its de-rated capacity. If a wind farm has a 100MW nameplate, it only gets paid for ~8MW of capacity because that’s all the grid “trusts” it to provide during a crisis.

Here’s electricity generation by source for the UK from 2015 to 2024. Reminder - TWh are just like the kWh on your electricity bill, but a billion times bigger.

Domestic electricity generation has dropped – by 55 TWh, or 16%, from 2015 to 2024. Demand fell by 31 TWh, or 10%. Net imports increased to make up the balance. We’ll look at this in a follow-up article.

The UK story in one sentence:

Coal collapsed, wind surged, nuclear declined, gas stayed stubbornly central, and biomass became a non-trivial chunk of “renewables”.

Some headline numbers (from these categories):

• Coal: ~76 TWh (2015) → ~2 TWh (2024). That’s basically a disappearance.

• Wind: ~40 TWh → ~83 TWh. Roughly doubled.

• Nuclear: ~70 TWh → ~41 TWh. A long slide.

• Other fossil (mostly gas): ~102 TWh → ~88 TWh. Down a bit, but still huge.

• Biomass & other renewables: ~29 TWh → ~40 TWh. Up meaningfully.

• Solar: ~7.5 TWh → ~14 TWh. Up, but not the main character in the UK.

If you squint, the UK is doing two things at once:

1. Decarbonising by removing coal and adding wind.

2. Holding onto gas as the system’s structural “backbone”.

That second point is the one that tends to get lost in celebratory headlines.

Year-to-year Changes

Here’s the year-on-year change in generation by source – the same data as the first graph, but shown as annual deltas:

This is the “year-to-year turbulence” view. It answers: what moved the dial each year?

A few patterns jump out:

• Coal falls in steps, not smoothly. The UK didn’t gradually “use less coal.” It closed plants and coal output drops hard.

• Wind rises in bursts (new capacity coming online), but with big weather-driven wobble on top.

• Nuclear mostly trends down with occasional small recoveries—classic aging fleet behaviour.

• Gas flexes around everything else. When wind and nuclear are down, gas tends to fill the hole (and vice versa). You can see gas behaving less like a “being phased out” fuel and more like a balancing fuel.

This is one of those graphs that quietly undermines simplistic narratives. The UK didn’t just “replace coal with wind.” It replaced coal with a mix: wind + more biomass + gas doing the heavy lifting of variability.

Capacity Additions: Nameplate vs Reality

Here are capacity additions – the “nameplate” MW or GW that always get cited in press releases:

Capacity charts are often used as PR because they’re easy to game: a gigawatt of wind “sounds like” a gigawatt of gas until winter happens.

Still, capacity change tells you what the build/retire machinery is doing:

• Coal capacity collapses over the decade. That matches the generation story: coal isn’t “managed down”; it’s shut.

• Wind capacity climbs steadily, with some chunky years.

• Solar grows, but more modestly than you’d guess (unless you’ve actually live in the UK) from the way solar dominates the social-media imagination.

• Other fossil capacity doesn’t vanish. It bounces around. The system still wants dispatchable power.

This is the bit where a grid engineer shrugs and says: “Yes, of course.” Electricity systems are not moral projects; they’re physical machines with a reliability constraint.

As with our China and Germany view, here’s the generation capacity converted into a more realistic picture of what each can produce:

This chart takes the capacity changes and multiplies each technology by a representative capacity factor (CF). It’s a blunt tool, but a useful one.

Interpretation:

• 1 GW of wind is not 1 GW of average power—it’s more like ~0.3 GW average (order-of-magnitude).

• 1 GW of nuclear is close to ~0.8 GW average if it’s running well.

• Biomass behaves closer to dispatchable plant than wind/solar (higher CF), because it’s burnable fuel.

So this chart asks: If we translate new capacity into rough average output, what’s the trend?

You get a clearer picture of why gas remains central:

• Wind additions deliver real energy, but not in a controllable way.

• Nuclear retirements remove large chunks of dependable energy.

• Biomass additions add energy that looks “renewable” in policy categories but behaves like a fuel.

This chart is also where you start to feel the shape of the follow-up post: if you remove coal and some nuclear, then add variable renewables, the system must compensate with something that can respond quickly (gas, imports, storage, demand response, etc.).

The Decade Shift: Absolute and Percentage Changes

Here’s the absolute change in generation over the decade:

And by percentage:

They both paint the same picture.

The political version is: “Coal is gone, renewables are up.”
The engineering version is: “Coal is gone, wind is up, nuclear is down, and gas is still doing a lot of work.”

Both are true; one is incomplete.

“Biomass & other renewables” — what is it?

In UK electricity statistics, this bucket is mostly bioenergy (things you burn), plus tiny amounts of wave/tidal.

What’s being burned?

• Wood pellets / woody biomass (including substantial imports, historically)

• Waste wood and residues (varies by plant/supply chain)

• Energy-from-waste (incineration; mixed fossil + biogenic content)

• Biogas (landfill gas, sewage gas, anaerobic digestion)

And this brings us to the question curious readers ask:

Is biomass really “CO₂ neutral”?

We’ll come back to this. But as a spicy placeholder, in 2024, Drax emitted approximately 13.3 million tonnes of CO2.

In official UK grid intensity figures, this is reported as 0g CO2/kWh, despite the physical stack emissions being higher per MWh than a coal plant.

Where this is Heading

Just like with Germany, we’ll primarily look at the imports and the biomass accounting fiction. If you shut down your coal plants and it’s not windy (or sunny — and if you haven’t lived there, lack of sunshine is a key feature), how do you keep the lights on?

Notes

1. Some small categories in the charts are rolled up or omitted for clarity.

2. This is generation within the UK, not “electricity used in the UK after imports/exports.”

When the wind is low in Germany, it’s usually low in northern Europe.

This is where Spain enters the conversation. As shown below (data from WindEurope), Spain is a wind powerhouse:

More importantly, meteorologically speaking, Spain is often “Anti-Germany.” When a high-pressure system sits over Berlin, creating a wind lull, it typically forces Atlantic storms south toward Madrid.

On paper, this is the “Holy Grail” of renewables: The Mediterranean Balancing Pole. Research by Grams et al. (2017) - summary in note 1 - suggests that if we built more wind in Spain and the Balkans instead of doubling down on the North Sea, we could nearly eliminate the “zero-wind” days that currently haunt the German grid.

However, there is a massive physical hurdle between the wind blowing in Zaragoza and the factories humming in Bavaria. Electricity is a physical commodity that needs a pipe. Currently, the “pipe” across the Pyrenees is tiny—only about 2.8 GW. To put that in perspective, Spain has over 31 GW of wind capacity. Even if Spanish wind is roaring while Germany is calm, we can currently only send about 9% of that potential north.

In the rest of this article, we’ll look at why this “interconnector reality” is a bottleneck of multi-billion euro costs and decades-long delays.

The 2015 Precedent: A Billion-Euro Tunnel

The last major upgrade to this connection was the Baixas–Santa Llogaia link which “started in 2008” and was commissioned in 2015. To cross the Pyrenees without destroying the landscape, engineers had to build a 65 km underground line, including an 8.5 km tunnel through the mountains.

• Cost: €700 Million.

• Technology: HVDC (High Voltage Direct Current).

• Efficiency: Doubled the capacity at the time, but as Spain’s wind fleet grew, this billion-euro “extension cord” was full almost instantly.

The Bay of Biscay Saga: From €1.7B to €3.1B

To bypass the (politically sensitive) mountains entirely, the current “flagship” project—the Bay of Biscay Interconnector—goes under the sea. It is a 400 km subsea link designed to increase capacity to 5.0 GW.

However, the cost of this project has exploded. Initially estimated at €1.75 Billion in 2017, the final price tag has surged to €3.1 Billion as of early 2026. Why the jump?

1. Material Inflation: The cost of copper and specialized HVDC subsea cables spiked during the 2021-2023 supply chain crisis.

2. Technological Complexity: To avoid the subsea “Capbreton Canyon,” the cable has to transition back to land for a section, requiring micro-tunnels that protect coastal dunes while enabling the transition from marine to terrestrial infrastructure.

The Decades of Delay: Navarra and Aragón

If the Bay of Biscay project is a struggle, the future projects are currently ghosts. Two major terrestrial links were planned to follow: Navarra–Landes at 2GW and Aragón–Atlantic at 1.5GW.

• The Original Plan: Both were intended to be operational by 2030 to help Europe meet its 15% interconnection target.

• The New Reality: Due to environmental opposition and the sheer technical difficulty of HVDC procurement, Navarra has been pushed to 2036, and Aragón is now estimated for 2040–2041.

The Technical “Tax”: Why is it so Expensive?

Building a long-distance interconnector isn’t like stringing a wire between two poles. At these distances, you cannot use standard Alternating Current (AC) because the line losses would be astronomical. You must use HVDC (High Voltage Direct Current), which brings three distinct “taxes” to the bill:

• The Converter “Tax”: At each end of the line, you need a massive Converter Station—the size of several football fields—to turn AC into DC and back again. These stations alone cost roughly €200 million per Gigawatt.

• The Cable “Tax”: Subsea cables can cost 5 to 10 times more than overhead lines. Laying them requires specialized vessels that cost hundreds of thousands of euros per day to operate.

• The “Transit” Tax: This is the hidden hurdle. Power doesn’t just “teleport” through France once it crosses the border. France’s grid operator, RTE, is currently planning a €100 billion upgrade through 2040 to reinforce its internal “backbone.” To get Spanish wind to Germany, France must build massive 400kV “highways” to carry that power from the Pyrenees all the way to the Rhine. Without these internal French upgrades, the Spanish electron stays stuck in Bordeaux.

• The “Grid-Forming” Tax: Following the grid stability issues of 2025, new links now require Grid-Forming Inverters. This software-heavy hardware allows the cable to help “restart” a grid after a blackout, but it adds a significant premium to the electronics.

The Bottom Line: A 1,500km Relay Race

If it costs €3 billion just to move 2 GW across the border, why do we do it?

Because the potential is too big to ignore. The Bay of Biscay link alone is expected to avoid 7,400 GWh per year of Spanish wind energy being “curtailed” (wasted). When the wind blows in Spain and Germany is in a lull, the price difference can be over €100/MWh.

However, we have to view this as a 1,500km relay race.

1. Leg 1: Get the power over the Pyrenees (The €3B Biscay/Navarra projects).

2. Leg 2: Move it across the busy French “intersection” to the German border (Internal RTE reinforcements).

3. Leg 3: Push it through the final cross-border links into the German industrial heartland.

At current spreads, these multi-billion euro “extension cords” can pay for themselves in 12 to 15 years, but only if the entire chain is built. For now, the “Mediterranean Balancing Pole” remains a meteorological mirage. Until every leg of this relay race is funded and finished, Spain’s wind will remain stranded behind the mountains, while Germany continues to look to its immediate neighbors to keep the lights on.

References

Balancing Europe’s wind-power output through spatial deployment informed by weather regimes, Grams et al. (2017), Nature Climate Change

Notes

Summary of Grams et al. (2017):

“Balancing Europe’s wind-power output through spatial deployment informed by weather regimes”

The fundamental challenge addressed by Christian Grams and colleagues is the high variability of wind power at the synoptic scale. The paper moves beyond simple distance-based correlations to evaluate wind power through the lens of seven distinct Atlantic-European weather regimes (WRs). Using 36 years of ERA-Interim reanalysis data, the authors simulated the potential wind power output of current and future European fleets.

The Weather Regime Framework

The study identifies that European wind production is dominated by the phase of the North Atlantic Oscillation (NAO).

1. Zonal Regimes (NAO+, Atlantic Ridge): Characterized by high wind speeds in the North Sea and Central Europe but low wind in the Mediterranean.

2. Blocked Regimes (Scandinavian Blocking, European Blocking): These result in the dreaded Dunkelflaute (dark doldrums) across Germany and the UK. High pressure sits over Northern Europe, but—crucially—atmospheric “steering” pushes storms toward the Mediterranean and the Balkans.

The North-South Dipole

Grams et al. quantified a powerful meteorological dipole. During “Scandinavian Blocking,” which typically causes Germany’s lowest production periods, wind speeds in the Iberian Peninsula (Spain/Portugal) and the Balkans are significantly above average. Specifically, they found that when the North Sea fleet’s capacity factor (CF) drops below 10%, the Mediterranean and Iberian regions often see a corresponding surge in potential generation.

Key Findings on Spatial Deployment

The paper’s most provocative finding is that the current geographical distribution of wind turbines in Europe is suboptimal for grid stability. Currently, deployment is concentrated in the North Sea region, which exacerbates variability because these turbines are often affected by the same weather regimes simultaneously.

The authors modeled an “optimized” deployment scenario. They found that by strategically expanding capacity in Spain, Italy, and the Balkans, Europe could:

• Reduce total variability by 10%.

• Nearly eliminate periods of extreme low generation. In their model, an optimized pan-European fleet would almost never drop below a 10% aggregate capacity factor, as the Mediterranean “balancing pole” compensates for the Northern “lull.”

Conclusion and Policy Implications

The study concludes that “balancing” is not just a function of building more turbines, but of meteorological diversification. However, the authors concede a major technical caveat: this balancing potential is entirely dependent on a frictionless transmission grid. Without massive cross-border interconnection, the “Mediterranean balancing pole” remains isolated, forcing Northern Europe to rely on backup thermal plants (gas/coal) despite the abundance of wind elsewhere on the continent.

How do they do it? Why doesn’t everyone follow?

Better Grid Interconnection is Always Good

Well, “cost effective grid-scale storage” is way better. Here’s a photo:

A subject for another day.

In the absence of fairy tales or future massive engineering improvements, large grid interconnections are a good second choice.

Why? Because they quietly do three jobs at once:

1. Share cheap generation.

2. Share backup.

3. Share luck with the weather.

This was useful in the boring fossil era; it becomes essential once you start leaning hard on wind and solar.

Sharing cheap stuff, not average stuff

Think of an interconnector as a very long extension cord with Excel running at the other end.

Every hour, markets don’t trade “average electricity”, they trade whatever is cheapest at the margin. If:

• Germany has spare coal or wind that’s cheaper than a neighbour’s gas, power flows out of Germany.

• A neighbour has cheap nuclear or hydro and Germany is short, power flows in.

Interconnection doesn’t magically create low prices, but it lets each country lean on somebody else’s cheapest unit instead of their own expensive unit. That’s why cross-border flows bounce around so much hour-to-hour in the SMARD plots: the merit order keeps reshuffling and the interconnectors follow.

Sharing backup instead of building it twice

Without interconnectors, every country has to plan its system as if the neighbours don’t exist:

• You need enough dispatchable capacity to cover demand in a cold, windless week plus a safety margin.

• You need enough reserve to survive big outages.

That’s expensive. It’s like every town building its own spare hospital “just in case” and then keeping it half empty.

Once you connect systems, you can share capacity:

• Germany doesn’t need to build as many peakers if it can import French nuclear or Scandinavian hydro during tight hours.

• Those neighbours don’t need quite as much local backup either, because in some bad hours they will be leaning on German coal or gas.

Nobody gets to zero backup – you still need domestic capacity for when neighbours are also stressed – but the total fleet across the region can be smaller and run less often. Quiet, unglamorous savings.

Smoothing the weather (a bit)

Now add lots of wind and solar.

The problem isn’t that they’re random; it’s that they’re correlated inside one region:

• When it’s dark in Germany, it’s dark in France too.

• When a big high-pressure system parks over Central Europe, wind sags in most of that blob.

But the weather is not perfectly synchronised across a continent:

• A still, cloudy week in Germany might coincide with strong wind in the North Sea and decent hydro inflows in Scandinavia.

• A summer solar peak in Southern Europe might line up with moderate demand in the north.

Interconnection lets you harvest some of that geographical diversity. It doesn’t eliminate Dunkelflaute - dark windless periods, but it makes the dips shallower and shorter than they would be in a walled-off national system.

Why this matters for Germany’s story

Put those pieces together and you get a slightly counter-intuitive picture for a country like Germany:

• On some days, Germany looks like a renewables powerhouse feeding the neighbours – exporting when there’s plenty of wind and sun and local demand isn’t too high.

• On other days, Germany looks like a big flexible load – importing 15–18 GW for hours at a time when the weather is unhelpful and its own dispatchable fleet would otherwise have to work a lot harder.

From the outside, all you see is a squiggly line of imports and exports. Underneath that line, interconnection is doing the quiet work of:

• letting Germany lean on other people’s hydro and nuclear when its own wind/solar sag, and

• letting neighbours lean on German coal, gas and wind when their systems are tight.

Whether you think that’s a climate hero story, a shell-game story, or “complicated but mostly fine” depends on which hours you look at and what you assume is happening on the other side of the border.

Who Supplies Germany?

Here’s a graph of imports by country:

Across five years the total imports have gone up almost two and a half times. That’s exactly what you’d expect once you simultaneously:

• push renewables hard,

• shut down your nuclear fleet, and

• run coal and gas less often than you used to.

The country breakdown is interesting:

• France, Norway/Sweden and Switzerland dominate.

• Denmark looks surprisingly important.

• The Netherlands, Poland, Czech Republic and Belgium matter, but they’re supporting actors rather than stars.

Denmark isn’t as big as it looks

At first glance 2024 looks like “wow, Germany is living off Danish wind”. That feels wrong – if German wind is low, Danish wind is usually low as well; and Denmark only has 6 million people vs Germany’s 83 million. How can they be exporting so much?

The trick is that Denmark is a transmission corridor as much as a generator.

The short version of the Danish stats (2019–2024) is:

• Denmark is usually a net importer overall.

• It brings in on the order of 8–10 TWh per year from Norway and Sweden, and

• exports a similar amount south into Germany.

A very rough accounting – based on annual import/export flows, not full flow-tracing – suggests that something like half to two-thirds of “Danish” electricity imported by Germany is really Nordic hydro and Swedish nuclear passing through the Danish hub, with the rest coming from Danish wind and biomass. (note 1).

By 2024, with very strong Nordic–Denmark–Germany flows, that “two-thirds Nordic, one-third Danish” mix is probably closer to reality than a 50/50 split.

If we re-label accordingly, the “who really supplies Germany?” picture for 2024 looks more like:

• France: ~20.3 TWh

• Norway + Sweden (via Denmark and direct links): ~17.7 TWh

• Switzerland: ~8.4 TWh

Together those three “firm-ish” sources – big hydro and nuclear – account for roughly two-thirds of all German imports in 2024.

Exports: Who is Germany Propping Up?

Total exports fell by about one-third across 2019–2024.

That’s not because Germany has become altruistically modest. It’s because:

• domestic demand and climate policy keep squeezing coal and nuclear, and

• in tight hours Germany now needs more of its remaining dispatchable fleet for itself.

When Germany does export heavily, it is often shipping coal- and gas-backed power into neighbours who are also stressed – the mirror image of those big import spikes we see in winter.

A closer look at the worst weeks

To make this less abstract, here are two graphs for November–December 2024, the period that contains the highest import hours of the year:

And which countries contributed to the imports:

You can see:

• routine 10–15 GW imports,

• frequent spikes towards 17–18 GW, and

• a rotating cast of helpers: Denmark/Nordics, France and Switzerland occupy most of the stack.

When we match those hours to German generation in the SMARD “Actual generation” dataset, the top ten import hours of 2024 look roughly like:

• Imports around 17–18 GW,

• making up about 25–35% of total load in those hours,

• with German coal running hard, gas doing its share, and wind often miserable.

In other words, these aren’t little balancing blips; they’re large chunks of the system. Without those 15–18 GW of imports, Germany would either need:

• more domestic firm capacity online (coal/gas, storage, demand response), or

• to start shedding load.

So what should we take away from this?

A few things feel safe to say:

1. Interconnection is doing real work.
   Germany’s energy transition is not just “build lots of renewables”. It’s “build lots of renewables and lean heavily on neighbours’ hydro, nuclear and spare thermal capacity”.

2. Someone still burns fuel.
   When Germany is importing hard in a Dunkelflaute week, the missing wind isn’t magically replaced by “Europe”. It’s replaced by very specific plants: French and Swedish reactors, Norwegian hydro and Swiss dams & reactors, and quite a lot of coal and gas somewhere.

3. The story cuts both ways.
   The same interconnectors that rescue Germany in bad hours let Germany rescue its neighbours in their bad hours. From a regional perspective this is mostly a good thing: you get more reliability for fewer power stations.

4. You can’t scale this trick forever.
   Sharing is great as long as not everyone is in trouble at once. If every big country tries to copy Germany’s trajectory, all counting on “imports” during dark, still weeks, you eventually run out of neighbours with firm capacity.

For now, though, the data say something quite simple:

Germany can run a high-renewables, no-nuclear system largely because it is physically and economically welded into a wider European grid that still has a lot of firm, fuel-burning capacity. When the wind dies over Germany, the interconnectors hum.

That doesn’t make the German experiment meaningless. It just means that any honest discussion of “how they do it” has to show the imports as clearly as the wind turbines.

Notes

1. Back-of-envelope split based on annual import/export statistics from the Danish regulator (Energinet); a proper flow-tracing study would be needed to pin this down more precisely.

2. All import, export and generation figures are calculated from hourly data downloaded from SMARD, the German electricity market data platform operated by Bundesnetzagentur (the federal network agency). Dataset names: “Cross-border physical flows, Bidding zone DE/LU” and “Actual generation, Bidding zone DE/LU”, hourly resolution, 2019-01-01 to 2024-12-31. ResearchGate

   Bundesnetzagentur & SMARD, Electricity market data platform, accessed December 2025.

Germany is a simpler story from a PR point of view: whichever way you slice it, you see basically the same picture. The interesting part is how they’ve done it. That’s for a follow-up. First, the numbers.

The Big Picture

Here’s Germany’s electricity generation by source. TWh are just like the kWh on your electricity bill, but a billion times bigger.

The headline: domestic electricity generation has dropped a lot – by about 150 TWh between 2015 and 2024 (see notes). But demand hasn’t dropped that much.

In 2015 Germany was a net exporter of about 51 TWh. By 2024 it had flipped to about 32 TWh of net imports – none of which appear in the graph above, which shows domestic generation only. In a follow-up we’ll look at these cross-border flows, because they’re central to how a high-renewables grid actually works.

For now, also put a mental asterisk next to “Biomass & other renewables”. Most of that is biomass. It sounds very green but it’s not CO₂-free in the same sense as wind and solar. The good news (for simplification) is that in Germany it hasn’t changed much over this period, so we can mostly ignore it for this story.

Once you adjust for the change in trade balance, implied domestic demand is down by only about 70 TWh. Most of that drop is from heavy industry; household and services use are roughly flat over the decade.

The high-level changes over 2015–2024 are:

• Coal has roughly halved.

• Gas has increased, but nowhere near enough to replace coal.

• Nuclear has been fully phased out.

• Wind is up by about 75%.

• Solar has roughly doubled, but from a much lower base than wind.

Year-to-year Changes

Here’s the year-on-year change in generation by source – the same data as the first graph, but shown as annual deltas:

It’s less intuitive at first glance, but it shows how you get from 2015 to 2024:

• Big step-downs in coal in several years.

• Occasional bumps in gas when the system needs backup.

• Wind and solar adding useful chunks in most years, with some flat or bad-wind years.

• Nuclear dropping in a couple of big, discrete chunks as plants are shut down.

Capacity Additions: Nameplate vs Reality

Now look at capacity additions – the “nameplate” MW or GW that always get cited in press releases:

If you read the China piece, you’ll remember we rescaled those capacity additions by capacity factor (CF).

• A coal plant could in principle run all 8,760 hours of the year at full power (so CF up to ~100%), though in practice it’s less.

• Solar cannot: there’s night, lower sun angles, and clouds.

• Wind turbines also only produce a fraction of their nameplate rating over the year.

So “We added 10 GW of solar!” sounds like 10 GW of new coal-equivalent, but it isn’t. That’s why nameplate additions alone always flatter solar and wind.

Here’s the more realistic picture: capacity added or retired, scaled by a representative capacity factor for each source – i.e. “how much steady output would this capacity be equivalent to, if it ran like it typically does?”:

This isn’t a standard metric, and the CF assumptions are approximate, but it nudges us closer to reality: coal retirements are big negative chunks; wind and especially solar additions are respectable but not magic “coal replacement machines”.

The Decade Shift: Absolute and Percentage Changes

Here’s the absolute change in generation over the decade:

And here’s the percentage change:

Unlike China – where the same kind of graphs can tell wildly different stories depending on whether you look at absolute TWh, percentage, or “scaled capacity” – Germany’s charts all rhyme:

• Coal down hard.

• Nuclear to zero.

• Wind and solar way up.

• Gas nudged up, but not massively.

The story is robust to how you slice it.

How Big are Wind and Solar Now?

On the official gross-generation numbers:

• 2015

  • Wind: 12.4% of generation

  • Solar: 5.9%

  • Wind + solar = 18% of total generation.

• 2024

  • Wind: 27.7%

  • Solar: 14.8%

  • Wind + solar = 42.5% of total generation.

So you can say, quite fairly:

“In one decade, wind + solar have taken over an extra quarter of total generation.”

That’s impressive.

It’s also – inevitably – a bit flattering. Suppose Germany in 2024 still had:

• the same heavy-industry demand as 2015, and

• was still a net exporter instead of a net importer.

Then you’d be comparing 2024 wind + solar generation (in TWh) against the 2015 total. On that basis, wind + solar in 2024 are about 33% of 2015 total generation, up from 18%. Same physics, different denominator, less dramatic headline.

Where This is Going

In the next article we’ll look at what’s happening beyond the border:

• Germany used to be a big exporter.

• Now it often leans on imports when the wind is low or demand is high.

• Its neighbours – especially France and the Nordic countries – act as a kind of “extended grid battery”, with nuclear and hydro smoothing out some of Germany’s wind + solar variability.

Imports aren’t just boring footnotes at the bottom of a balance sheet. They’re an integral part of how a high-renewables system behaves. The interesting question is:

What happens on the day when Germany wants to import at the same time as its neighbours also want to import?

Right now, France’s nuclear fleet (yes, fleet – that’s the standard term for a group of reactors) means it’s much less exposed to the weather than Germany is. How long that remains true, and how Europe manages the tug-of-war between wind, solar, nuclear, hydro and gas, is the next part of the story.

Notes

1. Official total generation in 2015 from AGEB/Destatis is about 647 TWh.

2. The totals in the stacked graph are about 4–5% lower than that because “other energy sources” – mainly pumped-storage generation and a small mix of miscellaneous fuels – are omitted for clarity.