One Sun, Ten Prices: The Capture-Rate Divergence Trade
Europe's most-quoted solar stress metric broke this year. German negative-price hours fell 23% while curtailed volume rose 20%. In April the same solar MWh earned seven times more one border away, and the December forecasts stretch the gap past ten. The stress didn't heal. It moved. This is where it went, and how a desk trades it.
Executive Summary — The Broken Dial
Every energy feed on LinkedIn quotes the same number this summer: negative-price hours. Spain logged 596 in the first half of 2026, Portugal 462, France 370. The count is easy to fetch, easy to chart, and it has stopped meaning what everyone thinks it means.
Germany is the proof. Negative hours fell from 389 to 298 year on year on the EPEX convention, a 23% improvement on the dial everyone watches. Over the same six months, commercially curtailed renewable volume rose 20%, from 1,216 GWh to 1,463 GWh on Montel's estimates: renewable output offered below zero in the day-ahead sell curve and left uncleared. The Solar Peak Act zeroes subsidy value for new solar during negative intervals, so those assets bid at zero or switch off instead of bidding negative. The price signal went quiet. The physical surplus got worse.
Meanwhile the value of a solar MWh has detached from the solar resource itself. France, with excellent spring irradiance, captured 9.5% of the average baseload price in April, the lowest figure recorded in any European market; Italy cleared roughly 71% in the same month. Great Britain held 84.9% across Q2 and is forecast by Kpler near 97% for December. Italy-North is forecast to average about €120/MWh in capture prices through year-end. Same continent, same physics, a seven-to-tenfold gap in what the MWh is worth, decided entirely by what surrounds the panel: nuclear floors, missing interconnectors, permitting queues, and storage.
Stop reading negative-hour counts as stress. Read curtailed volume and capture spreads instead. The trade is owning the flexibility that monetises the divergence, in the zones where it is widest, for a window we estimate at roughly two years before storage buildout does to Europe what it is already doing to California.
The Metric Migration: How Germany Silenced Its Own Alarm
The Solarspitzengesetz entered into force on 25 February 2025. For qualifying new PV of 2 kW and above, the EEG support value drops to zero during negative-price intervals; the lost intervals are partly compensated later by extending the support period, so the law defers support more than it destroys it. Systems below 100 kW stay exempt until a smart meter is installed, and new small systems without control hardware face a 60% export cap. The design goal was rational: stop paying people to produce power nobody wants.
It worked, and that is exactly why the popular dashboard number broke. An asset whose support is worth zero in a negative interval rationally bids at zero rather than below it. When zero-priced renewable offers exceed demand, the auction clears at exactly €0 and rations the volume pro rata: Montel's worked example is 100 GWh offered at zero against 80 GWh of demand, leaving 20 GWh unsold. Where 2025's surplus printed as 389 hours of sub-zero prices, 2026's surplus increasingly prints as zero-price clustering and rejected volume. The market never shows the stress because the stressed volume no longer bids below the line.
- H1 2025: 389 negative hours, 1,216 GWh commercially curtailed (Montel estimate, all renewables).
- H1 2026: 298 negative hours, 1,463 GWh commercially curtailed, a record pace on Montel's modelling.
- Also H1 2026: wind and solar together overtook fossil generation in Germany for the first time.
The control group makes the case stronger than Germany alone ever could. France ran the same summer under the opposite subsidy design, and got the opposite result: negative hours rose about 14% while commercially curtailed volume fell 32% (1,035 to 700 GWh), because supported French assets keep producing through shallow negatives and only curtail below roughly −€0.10/MWh. The Netherlands cut both. Same sun everywhere; the dial now measures policy design, not physics. A desk that models solar stress from negative-hour counts has a dial that moves the wrong way in Germany and the right way in France, and no way to tell which is which without curtailment and capture data.
Curtailment is supply withheld at €0 revenue. Every curtailed GWh is latent supply waiting for a battery, an interconnector or a flexible load to arrive and buy it at near-zero. That is the raw material of the spread trade in Section 07, and it is invisible if you only watch the price count.
One Sun, Ten Prices: The Divergence Map
Europe produced a record 129 TWh of solar electricity in Q2 2026 across Montel's market coverage, 23% above Q2 2025. The average solar capture rate across markets tracked by Kpler fell to 38% in May, an all-time low and seven points below the previous record low. The average hides the story. The dispersion is the story.
| Zone | Capture Signal (2026) | H1 Neg. Hours | What Sets the Price |
|---|---|---|---|
| France | 9.5% capture (Apr), −75% YoY | 370 | Solar stacked on a nuclear fleet with a high operational floor; mild spring demand |
| Spain | ~38% capture (Q2) | 596 | ~50 GW of correlated solar behind ~2.8 GW of exits to France |
| Portugal | 42% forecast Aug–Dec | 462 | Same Iberian pool, same border queue |
| Germany | 25–30% (Apr, by provider) → 43.6% (Q2) | 298 ▼ | Surplus migrating from price to curtailment under the Solar Peak Act |
| Italy | ~71% (Apr); €124.7 (Mar) → €108.6 (Jun) national | 0 (€0 floor) | GME day-ahead floors at zero; stress prints as zero-price runs, never negative |
| Great Britain | 84.9% (Q2, monthly 79–92%); ~97% Dec fcst | — | Low solar penetration; evening scarcity still prices the day |
| Italy-North | ~€120/MWh capture price fcst Aug–Dec | — | Permitting keeps solar scarce; gas still sets sunny-hour prices |
The pattern generalises: capture value is set not by irradiance but by what the solar output lands on. A zone with a high inflexible supply floor, weak borders and no storage destroys its own solar revenue. A zone where solar is scarce relative to demand keeps paying nearly full price for it. No harmonised dispersion index exists across bidding zones, but on like-month observations the April range ran from about 10% in France to about 71% in Italy: 61 percentage points, the widest the country data has shown. That gap is the widest tradeable structure in European power.
Negative-hour counts follow the Montel/Kpler H1 tally. Conventions diverge since the day-ahead moved to 15-minute settlement in October 2025: EPEX counts an hour negative when the mean of its four quarter-hours is; on that convention Germany printed 298. Montel's EnAppSys series puts France at 413 against the 370 above, and Modo's fractional 15-minute count puts Spain at 729.5 by 30 June, past its entire 2025 total. Direction agrees everywhere; levels depend on method, so never mix vendors in one series.
France: The Nuclear Sandwich, Both Seasons
April 2026 in France: solar generation up to roughly 3.4 TWh from 2.8 TWh a year earlier, nuclear output simultaneously higher at about 29 TWh, demand softened by mild weather and holidays. Prices went negative on 90% of April days across 139 negative hours, up from 90 a year earlier. The intraday low touched −€479/MWh on 26 April. The share of French solar output sold into negative intervals jumped from 29% to 45% in a year. Result: a 9.5% capture rate, the lowest ever recorded in a European market, worth €3.33/MWh in cash terms on Modo's April series.
The mechanism matters more than the record. French reactors do load-follow, and they ramped down materially this spring. What the system cannot shed is the combination: high fleet availability, a large minimum economic and technical output block, and cycling costs that make deep daily modulation expensive, so remaining output is offered very low in the merit order. Midday solar lands on top of that block, in a market that cannot export enough of the sum. Solar does not displace gas in France in April. It fights nuclear for room on a full grid, and loses on price.
Here is the part we find genuinely elegant: the same physics runs our winter thesis. In Priced Like 2021, Filed Like 2022 we argued French Cal-27 at €56.91 underpriced a nuclear fleet whose filed October availability sat below September's. Three weeks later the market agrees: Cal-27 printed €64.41 on 30 July and the Q1-27 premium over the calendar widened from €12–14 to roughly €30. One asset class, two seasonal extremes: in April the fleet's floor crushes midday solar value; in January its ceiling defines national scarcity. France is simultaneously Europe's cheapest midday market and its most dangerous winter one, for the same reason.
Short the French midday (capture-linked structures, negative-interval optionality), long the French winter peak (Q1-27 against Cal-27, per the July thesis, already ~€16 in the money on the spread). Both legs are the same position: a supply floor set against a load curve it no longer matches.
Iberia: An Island With 2.8 GW of Exits
Spain added roughly 9 GW of PV in 2025 and now operates close to 50 GW against broadly flat demand. Iberian solar output is highly correlated: when Seville produces, so does Faro. The peninsula's exits to the rest of the continental market total about 2.8 GW of physical interconnection with France, of which only 680 MW cleared as firm year-round Spain-to-France capacity for 2026. Spain logged 596 negative hours in H1, up 30% year on year and the most in Europe; Portugal 462.
And unlike Germany, Iberian stress is now physical, not just commercial: Montel puts Spanish network-driven curtailment at 2,391 GWh in H1, with 1,112 GWh in June alone against 193 GWh a year earlier. Set against that, Spain's operating battery fleet totals roughly 210 MW, while its grid-access queue holds about 40 GW of storage. The deepest trough in Europe currently has almost nothing to buy it, and the largest storage queue in Europe waiting to.
This is not "too much solar." It is too much coincident solar behind a constrained border, with a neighbour whose own midday is already saturated by nuclear and its own record solar. France is not an outlet in April; it is part of the same trap. The Bay of Biscay link is designed to lift the border to 5 GW, and it commissions in 2028, which is precisely why the divergence window has a date on it. Until then, Iberian capture economics are the continent's worst: Portugal and Spain are forecast to average just 42% and 50% capture across August–December. For unhedged merchant solar that is a revenue crisis. For flexibility it is the deepest, cheapest charging trough in Europe, twice a day, on schedule.
GB & Italy-North: Scarcity Premiums With Expiry Dates
The high-capture zones are not miracles; they are simply earlier on the same curve. Great Britain's capture strength is a penetration story: solar was 6.9% of 2025 electricity, so an extra solar MWh still displaces gas rather than other solar, and evening scarcity still sets daily prices. GB held 84.9% across Q2 (observed monthly range 79–92%). Kpler's ~97% December forecast is real but seasonal: scarce winter solar concentrated in valuable daytime hours. The CfD regime does not prop up the wholesale capture rate; since AR4 the contracts pay nothing in below-zero reference hours, and Modo expects negative prices to become the main drag on GB solar CfD revenues as the 4.9 GW cleared in the latest round (strike ~£65/MWh) and its successors connect, taking the annual rate from ~85% toward ~58% by 2035. The premium is real and it is melting on a published schedule.
Italy-North is a permitting story. Italy holds one of Europe's largest PV connection queues, roughly 140 GW of applications, while actually-built capacity crawls through EIA backlogs and land-use restrictions. Scarce solar plus gas-set sunny-hour prices equals the strongest capture prices in Europe: €124.7/MWh nationally in March, €108.6 in June, and a forecast around €120 for Italy-North across August–December. Note the design detail: GME's day-ahead floors at €0, so Italian stress never prints negative. It prints as zero-price runs instead, like the nine consecutive zero hours in the south on 12 May. Every reform that unblocks the queue converts a little of the northern premium into Iberian-style saturation.
For a desk, the two zones are the long leg of a convergence pair with a known catalyst calendar: CfD connection cohorts in GB, permitting reform milestones in Italy. The divergence trade is not a bet that the gap persists forever. It is a bet that the gap closes slower than the flexibility premium pays you to wait.
The Trade: Own the Spread, Not the Average
The divergence does not trade as a single instrument. It trades as three expressions, in order of directness.
A battery is a machine that buys the trough cannibalisation digs and sells the peak it spares. German two-hour systems annualised €218k/MW/yr in April; GB two-hour systems took ~£73k/MW/yr over the twelve months to April, and GB's June day-ahead spread averaged £90/MWh, the widest in over a year. The raw material keeps widening: the mean pan-European two-hour day-ahead spread ran €116/MWh in H1 2026 against €60 in 2024 (CBRE), and Germany's average daily high-minus-low reached €130.4/MWh across 2025 against €115.9 in 2024 (FfE), gross one-cycle optionality before losses and fees, not realised revenue. June sketched the ceiling: German middays averaged €21.90 while evenings averaged €258, with a €747 print on 24 June. And on 1 May, German day-ahead hit −€499.99/MWh at midday and spiked the same evening; the year's best arbitrage day was also its most negative price. Curtailed GWh (Section 02) is this expression's fuel gauge: withheld supply at zero is the battery's cost basis.
Do not short the average: the curve already prices it. German summer-27 peakload trades below baseload, with Q2-27 peak at €64.38 against base at €79.72, a 19% discount for the hours that used to be the valuable ones, and Spanish Q2-27 base at €44.94 sits 28% under the €62.50 year. The market knows about midday solar. What a 12-hour peak block cannot express is the inside of the day: it nets the −€500 trough against the €747 ramp and calls it one number. April is the tail warning, with France's capture factor falling from 0.42 to 0.10 in a year, far beyond anything a quarterly peak/base ratio implies. The underpriced residual is quarter-hourly convexity, curtailment risk and zonal basis, so own convexity: cap products, shaped hedges, and the capture-rate guarantees now being written as PPA products. The contract market already moved to the same conclusion. New German PPAs increasingly suspend at the first negative quarter-hour where legacy vintages waited six hours, and a ten-year German pay-as-produced strike near €50 against a €103.91 Cal-27 base settlement is the profile discount written down in one line.
Long the scarcity zones' capture premium (GB, IT-North) against short the saturated zones' (Iberia, FR midday) through capture-linked PPAs, financial transmission positions and shaped hedges. The carry is the divergence itself; the risk calendar is public (Biscay 2028, CfD cohorts, Italian permitting reform). Add the France barbell (Section 04) for the seasonal version, where the winter leg has already repriced from €12–14 to roughly €30 since mid-July.
What all three expressions share: they are long dispersion and indifferent to the average price level. The consensus argument this summer is about where baseload settles. The divergence thesis says baseload is the least informative number on the screen.
The Kill Switch: Australia Is the Memo From the Future
Every good thesis names the thing that kills it. Ours is already printed, in two grids that run years ahead of Europe on the storage curve. Australia's NEM crossed 9 GW of grid-scale storage and the average battery charge-discharge spread collapsed 85% in a year, from A$342/MWh in Q2 2025 to A$51 in Q2 2026, with fleet net revenue down from A$130.5m to A$57.5m. California is milder but unambiguous: CAISO's April four-hour top-bottom spread fell 35% year on year ($189 to $123/MW) as the fleet passed 16 GW, and the 7pm day-ahead price that pays every evening discharge dropped from $43 to $24. Batteries eat the spread they feed on. Solar keeps digging the trough, but storage shaves the peak that pays for the round trip.
Europe's version of that future is under construction, faster than most models assumed. German utility-scale BESS (1 MW and up) reached 3.61 GW / 5.93 GWh by end-June across 466 systems, after a record H1 that added 2.5 GW / 5.6 GWh across all segments, up 76% in energy terms, with average new-build duration stretching from 1.7 to 2.3 hours; Modo pencils a nominal 5.7 GW grid-scale fleet by year-end and projects two-hour revenues could halve by 2030. BW ESS broke ground on 1 July on Klostermansfeld, a 1 GW / up to 5.7 GWh single site targeting 2028, and RWE started a 236 MW / 470 MWh build at Hambach on 28 July. On the demand side, 73% of EU households still sit on fixed-price contracts, which is why demand flexibility barely moves the needle today and why it will later: Ember's modelling has behind-the-meter batteries shifting up to a quarter of German rooftop solar from midday to evening by 2030.
2026–27: divergence and dispersion dominate; European storage is still too small to matter system-wide, and Iberia (210 MW operating) has effectively none. 2028+: Biscay commissions, the German and Spanish pipelines land, long-term flow-based allocation matures, and the compression begins. The stop-loss is mechanical: four consecutive quarters of weather-normalised top-bottom spreads falling while PV output and curtailment keep rising means the trade has been arbitraged away, whatever the physical stress still says. This is a trade with a window, not a worldview.
What Falsifies This — Watch These Prints
If German curtailed GWh stops growing while negative hours keep falling, the surplus is genuinely clearing and Section 02 weakens. Monthly Netztransparenz and Bundesnetzagentur prints.
German H1 additions hit 2.5 GW / 5.6 GWh with durations stretching. If the ~5.7 GW grid-scale year-end level lands on time, the NEM scenario arrives in 2027, not 2028, and the spread leg shortens.
Bay of Biscay slipping past 2028 extends the Iberian leg; acceleration compresses it. Same logic for any France-Spain capacity auction changes.
Politically rejected today. Any revival reprices every German expression here: a split would print the north's surplus and the south's scarcity directly.
Since October 2025 the day-ahead settles on quarter-hours. Widening intra-hour dispersion confirms the shape-underpricing view in Expression 2; flattening refutes it.
Dynamic-tariff and smart-meter uptake beyond the current 27% of households moves midday demand up and erodes the trough earlier than modelled.
Falsified by prints, not by opinions. Every one of these is a number with a publication schedule.
Watching It Live on Voltstack
A desk somewhere argued about this exact problem in July: it could count a zone's 180 negative hours, but it could not say what a solar MWh in that zone actually earned. Capture rate per bidding zone is the most argued-about derived number in European power, and no free European tool publishes it. Both legs of the calculation, hourly day-ahead prices and hourly solar generation per zone, flow through Voltstack's live feeds today; the derived capture layer is on our build path, and this thesis is the reason it moved up the queue.
What you can already do on the platform: watch the German day-ahead curve print its midday trough per 15-minute interval, track Iberian negative clusters against interconnector flows, and see the evening spread that pays the batteries, on the public demo desk, no login required. When the curtailment prints and capture spreads in Section 09 move, you will see it there first, which is the entire point: this thesis is falsifiable on the same screens it was built from.
Data & References
- Kpler — "Europe's solar capture rates hit record lows as market divergence widens", 10 Jul 2026 (capture rates, H1 negative-hour tallies, GB/Iberia forecasts)
- Montel EnAppSys — European price-sensitive curtailment report, 27 Jul 2026 (H1 negative hours and commercial-curtailment estimates by country); Montel commentary "From negative prices to zero hours" (the zero-clearing mechanism); Q2 market summary (129 TWh European solar)
- Euronews — Solar Peak Act mechanics and German H1 coverage, 27–29 Jul 2026; EEG §51/§51a (gesetze-im-internet) for the deferral and exemption detail
- pv magazine — "Europe's record solar output drives surge in negative electricity prices", 15 Jul 2026; "Solar capture factors fall across Europe", 13 May 2026
- Pexapark — April capture-factor collapse (like-month France ~0.10 vs Italy ~0.71); Euro Composite PPA index, Apr 2026
- Modo Energy — European solar capture rates and prices Q2 2026 (incl. FR €3.33/MWh April); GB solar forecast (Apr 2026); GB BESS revenues (Jun 2026); CAISO battery research (2026); German BESS buildout and revenue benchmarks
- Reuters Open Interest — "Europe's solar boom is masking growing strain", 1 Jul 2026
- EEX settlements via Fraunhofer Energy-Charts (DE/FR Cal-27 base, 30 Jul 2026); EEX Q-27 tables (7 Jul 2026); OMIP screens (ES YR/Q-27, FR Q1-27, 29–30 Jul 2026); ICE Endex via public quote pages (FR/IT peak Yr-27, 24 Jul 2026)
- FfE — German day-ahead daily max–min spread, calendar 2025 (€130.4/MWh); CBRE — European power markets & battery storage update (2h spread €116/MWh H1 2026); European BESS Index — trailing TB2 by market
- MaStR via Batteriespeicher-Report — German utility-scale fleet Q2 2026; BW ESS (1 Jul 2026) and RWE (28 Jul 2026) — Klostermansfeld and Hambach; REE / ess-news — Spanish fleet and pipeline
- AEMO/NEM reporting via Modo and Renew Economy — Q2 2026 battery spread and revenue compression; Aurora Energy Research — Spanish network curtailment
- EPEX SPOT — 15-minute MTU go-live (30 Sep 2025) and hourly-index convention; JAO — Core flow-based intraday go-live (28 Apr 2026)
- Ember — demand-flexibility and behind-the-meter battery modelling to 2030; European Electricity Review 2026
- ENTSO-E Transparency Platform — prices, generation, interconnection (via Voltstack live feeds)