PEM vs Alkaline vs SOEC vs AEM: Which Electrolyzer Is the Real Bet?

Why the Machine Matters

When you make green hydrogen, you take water and electricity and split it. The electrolyzer is the box that does the splitting — and that box, plus the power to run it, is basically the whole cost of the hydrogen. So the machine you choose isn’t a procurement detail. It kind of is the business.

Here’s the number to hold onto: of every electrolyzer being built in the world right now, about three-quarters is one single type — not the fancy one everybody argues about online, but the old, boring, cheap one. That fact tells you who is actually winning, and the rest of this teardown explains whether they should be.

Nobody argues about whether a hammer is better than a wrench. You pick the one that fits the job — and the job is defined by your power and your site.

Four Machines, One Trick

Every electrolyzer does the same thing: it uses electricity to rip water into hydrogen and oxygen, which means shuttling a charged particle across a cell. The entire difference between the four architectures is which ion moves, and what moving it costs you.

Alkaline (AWE) moves a hydroxide ion through a bath of liquid potassium hydroxide and a porous separator. The chemistry is gentle, so the whole machine is built from cheap, tough nickel — no precious metals at all. That’s why it’s cheap. The catch: it’s bulky, and if you turn it down too far, hydrogen and oxygen start creeping across the separator and mixing — which is exactly what you don’t want.

PEM replaces the liquid with a thin solid membrane and shuttles a bare proton straight through it. Protons are quick and the film is thin, so PEM is compact, fast, and makes beautifully pure hydrogen. But the membrane behaves like a strong acid, and almost nothing survives sitting in acid while oxygen is being torn off next to it. Almost nothing — except iridium. Hold that thought.

Solid oxide (SOEC) goes the other way entirely: a ceramic electrolyte run at 600–850 °C, moving oxygen ions. At that temperature steam does part of the work electricity would otherwise have to do — the entire reason it’s the efficiency king. The price is in the description: 800 degrees is brutal on materials.

AEM is the dream mashup — alkaline’s gentle hydroxide chemistry and cheap metals, moved through a thin solid membrane the way PEM does it. Pull it off and you get PEM’s compact speed without PEM’s expensive metals. The catch is building a membrane that survives that environment for years. So far, nobody quite has.

The Water Problem

Before the money: the input almost everyone forgets. On paper it takes about 9 litres of water to make a kilogram of hydrogen. In the real world — purification rejects, cooling — it’s more like 20–30 litres per kilo. And clean is the key word: a PEM electrolyzer wants ultrapure, deionized, laboratory-grade water with almost no chloride, because chloride corrodes the cell. You don’t just need water; you need a water-treatment plant bolted onto your hydrogen plant.

Which is where somebody always says: build it next to the ocean. Here’s the thing — the water in seawater was never the problem. It’s the salt. Desalinate, reverse-osmosis, polish to ultrapure: that’s another plant, more energy, more capital, and a stream of brine to dispose of. The water molecule is cheap. Getting a clean one to the right place is not.

And this isn’t theoretical for me: when I’ve looked at sites for these projects, almost every serious one sits next to a river or a lake, in hydro country — Quebec, Manitoba, British Columbia, Washington State, and the same map everywhere hydro is king: Oregon and Idaho, upstate New York and Vermont, Norway and Sweden, Brazil, and China’s southwest, Sichuan and Yunnan. A big hydro grid hands you cheap, low-carbon, firm power and the rivers that come with it — electricity and water out of the same geography. Some deserts do sit on enormous fossil aquifers, but that water is tens of thousands of years old, doesn’t refill, is often brackish, and pulling it needs water rights that get harder to win every year. Green hydrogen needs three things in one place — cheap firm power, land, and clean water — and the map, not the spreadsheet, decides where you’re allowed to build.

The water molecule is cheap. Getting a clean one to the right place is not — water isn’t the line item that kills a project, it’s the map that decides where you can build.

The Cost War

A Western PEM system runs around $2,400 per kilowatt installed. A Chinese-made alkaline system: closer to $1,000, sometimes less ($750–1,300/kW). That’s a two-to-three-times gap for machines that make the exact same molecule. Put that next to the manufacturing number — China builds roughly 60% of the world’s electrolyzers, overwhelmingly alkaline — and the three-quarters figure explains itself: the cheapest, most proven machine, made at massive scale by the country that decided to own it.

So, case closed — cheapest wins? Not so fast. Cheapest and best-fit are not the same thing.

The Myth-Check: Can Alkaline Follow Renewables?

You’ll hear the same line everywhere: alkaline is cheap, but it can’t follow renewables — solar and wind bounce around all day, the old machine is too slow, so you need PEM. That used to be true. It’s now about half wrong. Modern alkaline designs can throttle from 5% to full power and ramp at roughly 10% per second; both alkaline and PEM respond to setpoint changes in fractions of a second. The gap everybody quotes has been quietly closing for years.

Where the gap is still real is at the very bottom. A PEM stack can idle down to about 5% of rated power and sit there happily. A classic alkaline machine really doesn’t like going below about 20% — partly because when you starve it, hydrogen and oxygen creep across to the wrong sides. That’s a safety problem, not a preference. PEM also packs far more power into the same footprint, which matters when land is tight.

This is the whole reason the machine has to match the power. Hydro is firm — it runs steady around the clock — so a hydro-fed plant can lean on cheap alkaline all day and never go near its weak spot. Wire your electrolyzer straight to bouncy solar and wind, and alkaline’s floor starts to bite. You can watch it play out in the real world: China’s showcase green hydrogen plant in Xinjiang was built with alkaline, and for a long stretch it ran at only about 20% of what it was supposed to make — not because the technology is broken, but because every time the renewable power dipped, the plant had to back off to stay safe. Doing this well takes smarter controls, not just a cheaper stack.

The Iridium Catch

If PEM is more flexible, why not build PEM everywhere? Because PEM has a catch of its own, hiding in the periodic table. PEM needs iridium — the one metal that survives the brutal acid environment on the oxygen side of the cell — and iridium is absurdly rare: the entire planet produces about 8 tonnes a year, nearly 90% of it from South Africa, as a byproduct of platinum mining.

The engineers have been heroic here. Iridium loadings have dropped from a couple of grams per kilowatt to about a third of a gram, heading for a twentieth. So this is a tax on scaling PEM, not a brick wall — the best analyses say thrifting plus recycling can support well over 1,000 GW of PEM by 2050. But it’s a real constraint, and it’s exactly why the metal-light options still matter.

The Efficiency King and the Wildcard

Solid oxide runs so hot that steam does part of the lifting, hitting 80–90% efficiency — one commercial demo made 20–25% more hydrogen per megawatt than the cooler machines. If you have cheap heat next door — a nuclear plant, a steel mill, an industrial site venting steam — it’s gorgeous. But today’s stacks last around 20,000 hours where alkaline and PEM last three to four times longer, and they hate being cycled. It’s not a bet you make everywhere; it’s a bet you make right next to cheap heat, once the durability problem is truly solved. Today: a watch.

AEM is on paper the best of both worlds — PEM’s speed and purity with cheap, abundant metals. The problem is the word “if”: today’s leading commercial AEM runs at a fraction of PEM’s throughput and has only been shown to hold up for around a thousand hours before the membrane starts falling apart, where the incumbents last tens of thousands. Highest upside, least proven. Genuinely exciting. Not yet.

The Verdict

Alkaline — a real bet. Today’s workhorse: mature, durable, no precious metals, and China will sell it to you cheap. It wins on firm, cost-driven sites — hydro country is its natural home. PEM — a real bet. The flexible future: marry it to bouncy solar and wind, or buy it when space and purity matter, and budget for the iridium question. SOEC — a watch: the efficiency king, but only next to cheap heat and only once it stops dying young. AEM — not yet: track it, don’t buy it.

The one line to take with you: cheap, flexible, durable, metal-light — every electrolyzer gives you three of those. None of them yet gives you all four. Anyone selling you the perfect machine is selling you something. And that reframes the China story too: China didn’t win electrolyzers by building the best one; it won by mass-producing the cheapest, most proven one and betting that cost and scale beat elegance. For today’s market, that’s a very smart bet. The next round — flexible, efficient, metal-light — is still wide open.

Cheap, flexible, durable, metal-light — every electrolyzer gives you three. None gives you all four. Read the site, then choose the configuration.