Heatwaves and French nuclear: why reactors power down when rivers run hot

This is not (only) “nuclear overheating”

When it is very hot and rivers run low, headlines stack: “EDF powers down reactors,” “record unavailability,” sometimes a blackout hint. Three objects get mixed:

1. the physics of cooling;
2. environmental discharge limits into waterways;
3. a missing-capacity figure on a given afternoon.

This piece separates the three. It is not about your electricity bill — another machine (contract, markets, taxes). Here: why a riverside fleet voluntarily cuts output when water is too warm or too scarce.

First: what the water around a plant is for

France has, in the press / EDF framing often repeated, on the order of 57 reactors, covering a majority share of power generation (often cited around ~70% depending on the period). Almost all sit beside a river or the sea: water cools the secondary circuit / condensers — it is not “spraying the core” like a car radiator in viral imagination.

In normal operation, a plant takes in water, uses it to reject heat, then discharges some of it warmer (and, depending on site and regime, under other effluent rules). The lower the river flow and the warmer the upstream water already is, the more the discharge heats a stressed ecosystem.

Consequence: cutting reactor power cuts the heat that must be rejected. In EDF’s wording as relayed by AFP, this is not a “nuclear safety on the brink” gesture: the company often stresses units can still run with warm water without a safety risk, and that caps mainly aim to protect flora and fauna.

ASNR / EDF: discharge caps, not a moral thermometer

Each site has limits — authorisations, prescriptions — framed on the safety / environment side by today’s ASNR (nuclear safety and radiation-protection authority), and applied by operator EDF. In heatwave / drought periods, EDF talks about “environmental constraints” or “external causes linked to the environment”: power reductions, or even a temporary unit stop, so temperature / flow / ecosystem thresholds are not breached.

Keep in mind:

• this is not the same as a forced outage or a long maintenance stop;
• nor is it proof that “nuclear never works when it is hot” in an absolute sense — coastal sites, cooling towers and flow regimes differ;
• it is a local, temporary constraint: low river + warm water = less headroom to reject heat.

Site-by-site thresholds are technical; for readers, the core idea is enough: the environment limits discharge, so sometimes power.

Drought + heatwave = two stresses at once

Drought and heatwaves act together:

• low flow: less water to dilute / carry away rejected heat;
• already-warm upstream water: less thermal headroom before a cap;
• prolonged heat: days when the same riverside sites stay constrained.

The fleet is not uniform. Some reactors are more exposed (summer-sensitive rivers); others less. A “national environmental unavailability record” can therefore concentrate on a few sites while the rest of the fleet still generates — which is why a percentage is not a nationwide off-switch.

August 2026: the ~15.6% figure (cite carefully)

According to AFP calculations from public EDF data (series since ~2015), relayed on 11 August 2026 by Boursorama, Capital and others: one August Sunday (often dated 9 August in those wires), the fleet reached about 15.6% nuclear power missing for heat / drought causes — presented as a record in that series.

The same coverage cites, for that day, reductions at several units (examples often named: Tricastin, Saint-Alban, Bugey…) and stops at others (Golfech, Chooz, Cattenom… — treat lists as a press snapshot, not a frozen official EDF table). Another peak in the number of constrained units had already been mentioned around mid-July 2026.

Sober reading of 15.6%:

• it is a share of nominal / available power missing at a moment, not “15.6% of French homes without power”;
• it is a media / open-data calculation, not a single ASNR release;
• EDF, in those wires, stresses the overall production impact remains limited at system scale, and that France often stays a net exporter — RTE cited for exports sometimes equivalent, on some days, to “more than a dozen reactors” in exchanged power.

In other words: an environmental unavailability record ≠ a blackout. Both sentences can be true; slogans pick only one.

What it does not imply

1. Imminent blackout. France’s grid leans on a mix, interconnectors, import/export and RTE dispatch. A summer constraint on a fraction of nuclear is a supply stress, not an automatic national outage.
2. “Nuclear is incompatible with climate.” That is an energy-policy debate. The observed fact here is narrower: some riverside sites cut when discharge limits bind. Other technologies also face weather constraints (low hydro, PV at night, wind without wind).
3. Your monthly bill. Wholesale prices, regulated tariffs, taxes, load profile: another article. Unavailability can weigh on markets if it lasts and stacks; it is not a “heatwave” line on your statement.
4. Safety = environment. In the EDF / AFP frame, summer caps mainly protect the aquatic environment. Mixing “we cut for biodiversity / water temperature” with “the core is melting” is a viral mash-up, not an ASNR reading.

Jellyfish, local incidents, media noise

Summer 2026 also saw other highly covered frictions — for example jellyfish forcing temporary stops at Gravelines units (AFP / France 24, 11 August). Useful for seeing that the fleet faces varied external hazards; dangerous if you stack them into one “nuclear is collapsing” story.

Method: local incident → cause (screens, heat, maintenance) → MW impact → system effect (exports, prices, reserves). Jumping from headline to blackout skips three steps.

Why this returns every summer (and why 2026 sounds loud)

Three structural reasons:

• France’s fleet is heavily nuclear and heavily riverside;
• hot / dry summers are more frequent in the climate narrative (without every summer being “the worst” everywhere);
• EDF data / AFP maths make the phenomenon countable: a percentage hits harder than a vague line.

2026 sounds loud because 15.6% breaks a series record and arrives in a summer already packed with water restrictions and heat alerts. The signal is real; the “no electricity anywhere” reading is not.

A useful habit when the next afternoon spike lands: separate cause (river heat / low flow), operator response (derate or stop named units), system outcome (still exporting? prices up? reserves thin?), and household story (almost never a direct outage line). Four boxes; far fewer panic titles. The same habit also keeps this topic distinct from political slogans about “keeping reactors online at any cost” or “shutting them all”: those are policy fights. The August mechanism is narrower — discharge limits bind, so output falls.

Mini glossary for the next headline

• Environmental unavailability: power not produced because of water / temperature / ecosystem limits — not necessarily a breakdown.
• Power reduction: the reactor still runs, but lower.
• Stop: unit at zero temporarily (check cause: environment, maintenance, other).
• ASNR: safety / radiation-protection authority — prescription frame.
• Net exporter: France sends more power than it receives over a period — compatible with local cuts.
• Missing power %: ratio to capacity, at an instant — not share of homes cut off.

What to keep without slogans

When rivers are low and warm, EDF cuts or stops some units to respect discharge limits aimed mainly at protecting the environment — ASNR / operator frame. In August 2026, an AFP calculation on EDF data places a peak around ~15.6% missing power one afternoon — a series record, to be cited as such. That is not proof of an imminent blackout, nor the same topic as your bill. It is water physics + environmental regulation + a mostly riverside fleet, under a summer that stresses them together.

Going further

• Boursorama / AFP — unavailability record — 15.6%, named sites, exports.
• Capital / AFP — same wire — EDF framing of safety vs environment.
• France 24 / AFP — summer 2026 heat and hazards — broader context (including local incidents).
• ASNR — institutional safety / radiation-protection frame.
• EDF — operator publications on output and constraints.