A 1 MW membrane free electrolyzer was destroyed on a test rig in Rossington in South Yorkshire during a routine shutdown, and its maker gave up building the machines altogether instead of redesigning the gas system

The machine had already survived the hard part.

It stood in a test bay in Rossington, wired up and pressurized, working through the last gate it had to clear before it belonged to a customer.

Then a fault the controller did not expect tripped it into a shutdown, which is the most ordinary thing a machine like this ever does.

At half past seven the following morning the manufacturer, Clean Power Hydrogen, asked for its shares to be suspended from trading on London’s junior market.

The shutdown is what wrecked it.

Nothing was salvageable afterward.

So how does switching a machine off destroy it?

How two gases that are never kept apart find each other

A conventional electrolyzer separates hydrogen from oxygen at the instant they are made.

A membrane sits between the electrode chambers, lets protons through, and blocks gas crossover.

It is a passive wall doing constant work while nobody watches.

The design in that test bay had no wall.

Its stack produces both gases as a single mixed stream, dries them, then pulls them apart by chilling the mixture until they part company by temperature.

The payoff is real: no perfluorinated polymer, no PFAS, and none of the platinum group metals a conventional stack has to buy.

The cost is that separation stops being a wall and becomes a routine.

Flow rates and pressures hold the two gases apart, and only while the routine runs.

Drop the pressure and the routine loses its grip.

A shutdown is exactly a controlled pressure drop.

A one megawatt unit that never left the test bay

The machine was a one megawatt membrane free electrolyzer, in the third and final stage of factory acceptance testing.

That is the last formal gate before a prototype becomes somebody’s equipment.

The buyer at the end of the chain was a water utility in Northern Ireland, reached through a civil engineering contractor that held the main contract.

An unexpected error put the machine into its standard shutdown sequence.

During the automated depressurization that follows, a hydrogen and oxygen mixture ignited inside the system and containment was lost.

The damage was past repair, and the manufacturer said so plainly: the unit could not be fixed well enough to finish the test program.

Nobody was hurt.

Health, safety and environmental procedures were followed throughout, and insurance assessors came out to the site.

What the company eventually put in writing

The ignition happened on May 28, 2026.

At half past seven the following morning the manufacturer, Clean Power Hydrogen, asked for its shares to be suspended from trading on London’s junior market.

Its first statement said only that an unexpected error triggered a standard shutdown and a subsequent failure caused significant damage.

The specifics arrived weeks later, in plain language: a hydrogen oxygen mixture ignited during an automated depressurization of the system, causing a loss of containment.

One line in that update matters more than all the rest.

The membrane free stack and its associated separators were not the cause.

A formal root cause analysis is due by August 31, and until it lands every claim about the ignition source is provisional.

Where a design without a membrane runs into its hardest wall

Clearing the stack is not the reprieve it sounds like.

It moves the problem into the balance of plant, which is where the company’s own engineering leadership put it.

The mixed gas system, they said, needs a substantial redesign before the technology can be operated safely under all conditions.

That sentence is heavier than it looks.

Dryers, pressure management software and gas handling logic are not the invention.

They are the part everyone assumes is already solved, and the part that company filings now place at the center of the failure.

Take the membrane out and those systems become load bearing.

Every startup, shutdown and pressure excursion becomes a moment when the gas boundary depends on control rather than a physical barrier.

Rivals chasing cheaper stacks attack the same cost problem from the other side, and they keep the wall.

A wall needs no controller.

What is left once the machine is gone

The board concluded it had neither the financial nor the engineering nor technical resources to run that redesign.

So the company stopped making machines and started selling the idea of them.

It holds 16 granted patents across 12 jurisdictions, including the United States, Japan, India, Europe and the Middle East, with 17 more pending and three licensing arrangements covering twelve countries.

The chief executive and the non executive chair both stepped down, and the subcontract behind the Northern Ireland unit was terminated by mutual settlement.

A first tranche of roughly $3.4 million landed at the end of June, with up to $10 million more put to shareholders, against a cash position that trade coverage reported as running only to mid July.

Meanwhile the wider industry is still opening electrolyzer factories for designs that keep the membrane.

Whether a licensee inherits a working technology or an unsolved control problem is the question nobody can answer yet.

The patents protect the geometry of the stack.

They do not settle how the system behaves on the way down.