Tearing apart the iridium electrolyzer trap

The green hydrogen narrative has a single villain, and it isn't permitting timelines or grid interconnection queues. It's element 77 on the periodic table — iridium — and the story is simpler and more brutal than anything the energy transition press is currently running.

The dominant technology for splitting water into clean hydrogen is the proton exchange membrane electrolyzer. PEM is what serious projects use — high efficiency, dynamic response, pairs cleanly with intermittent wind and solar. Every major green hydrogen announcement you've read this year is built on PEM. The IEA's net-zero pathway runs on PEM. The EU's green hydrogen targets run on PEM.

Inside the PEM stack, the anode side — where oxygen evolves — runs in a corrosive acid environment at elevated temperature for tens of thousands of operating hours. Exactly one catalyst material survives that reliably at commercial scale: iridium oxide. That's not a preference. It's the result of thirty years of electrochemistry research converging on the same answer. No commercially proven substitute exists today.

The headlines will tell you about gigawatt-scale project announcements. The headlines will not tell you that the planet produces approximately seven to eight tonnes of iridium per year — total — and that number will not move regardless of what iridium costs.

This is not a standard supply crunch. In most commodity markets, a high enough price eventually calls new supply into existence. Iridium is extracted as a byproduct of platinum mining — for every tonne of platinum pulled from the Bushveld Complex, the ore yields roughly 39 kilograms of iridium. You cannot drill for iridium. The price signal has nowhere to go. The only way to get more iridium is to mine more platinum, and the major South African producers — Valterra, Impala, Sibanye-Stillwater — have been explicit in 2026 that current prices do not justify new greenfield shafts.

The machine consuming the iridium is being built faster than the machine producing it. That is the entire story. Everything else is footnotes.

Story off. Numbers on. The supply chain for iridium into a PEM stack has four nodes. The constraint is in the first one. Everything downstream is downstream of geology.

The flow from ore body to operating stack moves through four chokepoints. Three of them can be engineered around. One cannot.

I have looked at a lot of supply chains where the constraint is a policy decision, a permitting calendar, a labor contract. This one is different. The constraint is a geological accident: commercially meaningful iridium concentrations exist in one place on earth because an asteroid hit South Africa sixty-six million years ago. That is the whole story. There is no second story.

The vulnerability is not that iridium is rare. Everyone in this market knows it's rare. The vulnerability is that the market has priced the loading reduction roadmap as a certainty — and is treating $8,000/oz as a temporary friction rather than a structural signal.

The loading reduction argument runs like this: state-of-the-art 2021 stacks needed ~400 kg per GW. Heraeus now offers commercial catalysts at the equivalent of ~180 kg per GW. Laboratory results have demonstrated down to 50–60 kg per GW equivalent. Therefore, by the time 80 GW of PEM capacity is deployed, the iridium problem will have been engineered away.

Two things are wrong with this. The first: lower loading and commercial durability have not been demonstrated in the same stack at the same time. The ~0.10 g/kW results come from short lab runs. PEM electrolyzers need to run for 80,000 hours in industrial hydrogen plants. Degradation rate at ultra-low iridium loading — where the oxide layer is thinner and active sites are fewer — is not characterized at that timescale. You don't find out by running it for a year in a lab. You find out by running it for ten years in a plant.

The second problem is synchronization. The loading reduction roadmap and the deployment ramp are on parallel timelines, and they are not synchronized. The major European projects completing in 2026 — which Johnson Matthey specifically flags as a primary driver of iridium demand this year — are being built with today's commercial loading, not tomorrow's laboratory result. The demand is happening now. The solution is happening later.

And the recycling argument is a decade away from mattering. The first large wave of PEM stacks won't reach end-of-life until the mid-2040s. Telling the 2026 project developer that recycling will eventually close the gap is like telling someone building a house today that lumber will be cheap in 2042. Technically plausible. Operationally irrelevant.

I traded the cobalt market in 2018 on the thesis that battery loading reductions would close the supply gap. They did — eventually. It took longer than consensus expected, cost more than consensus expected, and the stocks of the companies caught holding long-duration contracts at peak prices did not survive the wait. The iridium story has the same shape. Different element, same trap.

The market is not mispricing the scarcity — at $8,000/oz all-time high, price is already speaking. What hasn't been priced is the second-order consequence: iridium availability will determine which green hydrogen projects actually get built in this decade, and which get a politely worded delay notice. That allocation problem is where the capital displacement happens.

Where does capital go in the base case — supply stays constrained, loading reductions trail deployment? Not into broad clean energy ETFs. Those are diluted with solar installers and wind turbine OEMs who have zero iridium exposure. The edge is in two specific places: PGM producers with iridium exposure and unhedged forward books (Valterra, Northam), and the alkaline electrolyzer manufacturers who quietly become the only viable path when PEM supply chains jam.

The green hydrogen roadmap is real. The iridium problem is also real. In my experience, when a deployment plan and a geological constraint disagree about the timeline, the geological constraint wins. It just doesn't send a press release when it does.