Two open channels cut through the dunes feed 5.4 gigawatts of reactors on the North Sea coast, and a jellyfish bloom in one intake channel took 3 of those units off the grid overnight
The seawater comes in through a channel cut into the coast, and everything in it arrives with the water.
Weed, plastic, small fish, sand. A bank of moving screens takes it all out before the flow reaches anything that matters, and they turn all day without anyone thinking about them.
On a Monday night in the second week of the month, they stopped being able to keep up.
Which leaves an odd question standing, and it is how an animal with no bones and no muscle worth the name takes down a 900 megawatt machine.
Three units tripped themselves off the grid within the hour, and a fourth dropped to half power. Which leaves an odd question standing, and it is how an animal with no bones and no muscle worth the name takes down a 900 megawatt machine.
How a boneless animal stops a reactor
It never touches the reactor.
A coastal station pulls enormous volumes of seawater to carry away the heat its steam circuit rejects. That water crosses the moving screens first, and they are sized for the debris a shoreline normally hands over.
One jellyfish weighs almost nothing.
Several tons of them arriving together behave like a wall. The mesh does not tear and the pumps do not break, but the flow through the screens drops sharply, and the instruments watching that flow do exactly what they were built to do.
They shut the unit down.
That is not a failure, it is the design working. A protective trip on falling cooling flow is automatic and entirely deliberate, and the operator loses a reactor rather than risking one.
The station that lost three units in an hour
Gravelines sits on the North Sea coast between Dunkirk and Calais, not on the Channel, and it is the largest nuclear station in Western Europe.
Six pressurized water reactors of 900 megawatts each give it 5.4 gigawatts of installed capacity, packed into one stretch of shoreline.
In 2025 the site produced 32.27 terawatt hours.
That covers roughly 60 percent of everything the surrounding region consumes in a year, so when three of those units leave the network at once, the national grid notices inside seconds.
Units 2, 3 and 4 were the ones that went.
Unit 5 was already out for its ten year inspection, and one more unit stayed at full output through the night. Six reactors on paper, one running normally by morning.
What the restart record actually shows
Unit 2 came back first, on the 13th.
Unit 3 followed on the 18th at 3:55 in the morning, and unit 4 stayed down longer than either. Unit 1 also spent two days offline that week for an unrelated low water level, worth separating out rather than folding into the same story.
Recovery took about a week.
The year before it took closer to two, and that gap is the one measurable sign that the response has improved. It is not a sign that the bloom itself was smaller.
The earlier event was in fact bigger.
Four units tripped that night rather than three, and the calendar dates sit on top of each other, which turns an unlucky August into something an engineer designs around rather than waits out again.
What the operator bought after the first time
A three year partnership with a sustainable fishing organization, at about 1.7 million dollars.
The money buys observation and sampling at sea rather than a barrier. A fishing network runs daily visual checks, a small inflatable is on call to take samples within a day, and targeted trawling can be ordered between May and September when a bloom is spotted.
A sea rescue society adds a second set of eyes for about 35,000 dollars a year.
Cameras went onto the filter drums and the intake channel so the operator can watch the mat build rather than infer it from falling flow. The operator’s own review lists all of it.
None of it was a wall, and none of it stopped the swarm.
Warm water is doing this in more places than one, from the North Sea food web to reefs that crews are now shading by hand.
What engineers do next
Filters are the obvious answer and the wrong one.
No screen sized for weed and sand handles a mat measured in tons, and building one that does means a bigger structure, a longer intake and a permanent pressure loss the plant pays for every hour, bloom or not.
Reporting on the outage put it at three units with a fourth at half, so the work is moving upstream of the water.
Detection, timing and clearing the bloom offshore before it reaches the channel are cheaper than steel, which is exactly why the money went to boats and cameras rather than to a new intake.
Two new reactors are planned for this same shoreline.
Preparatory work at the site was authorized in August, and those units will draw from the same water as the six already there. Whether a jellyfish swarm is treated as weather or as a design load is the question that has to be answered before the concrete, and the honest answer today is that nobody has settled it yet.
Hugo is an engineer with strong technical expertise. Multilingual from an early age, his writing combines technical clarity with a strong interest in science and energy.
