The early conception of the hydrogen economy, going back to the 1970s, was modeled pretty explicitly on the natural gas economy. Huge centralized production plants. Long pipelines. A few dozen large producers serving thousands of customers through a fixed distribution network. The economics of natural gas had worked that way for half a century, and it seemed natural that hydrogen would work the same way.

I don't think this is going to happen. Or rather, I think it'll happen for a few specific use cases, but the dominant pattern is going to look much more like solar than like natural gas. Distributed, modular, on-site, sized to local demand, deployed in many small installations rather than a few large ones.

The reasons are pretty specific, and they're worth being explicit about, because the assumption that hydrogen production will be centralized is still embedded in a lot of policy and investment thinking.

The first reason is that hydrogen is genuinely difficult to transport. A natural gas pipeline can move methane across a continent because methane is a relatively well-behaved molecule that doesn't embrittle steel or leak through small openings. Hydrogen is the opposite. The molecule is much smaller, so it leaks through joints and welds that would hold methane. It diffuses into metals and embrittles them, which means hydrogen pipelines have to be built from specific alloys and operated at lower pressures. Building hydrogen pipelines is several times more expensive per mile than building natural gas pipelines, and the regulatory and right-of-way challenges are at least as hard. Some hydrogen pipeline buildout will happen — there are real proposals for industrial hubs in Texas, Louisiana, and the Midwest — but the universal-pipeline scenario isn't realistic right now.

The second reason is that liquefying hydrogen consumes roughly a third of its energy. That's a brutal round-trip efficiency penalty compared to liquefying natural gas, which costs about 10%. So long-distance shipping of liquid hydrogen is technically possible but economically painful for most use cases.

The third reason is the one I find most interesting, which is that the alternative — making hydrogen on site, from electricity, with an electrolyzer —keeps getting cheaper. The cost curve on electrolyzers has been steady, and it's accelerating though not as fast as we would like. The cost curve on electricity from renewables has been steeper. The crossover point at which on-site production beats trucked-in delivered hydrogen is moving down in scale every year. Right now it's somewhere around 500 kilograms per day for many use cases, but it's been moving by an order of magnitude per decade.

What this means in practice is that the hydrogen economy that's actually emerging looks more like solar than like natural gas. An electrolyzer can sit at a truck stop and serve a fleet of fuel-cell trucks. A two-megawatt unit can sit at a modular fertilizer unit and produce green ammonia closer to points of use. A ten-megawatt unit can sit at a datacenter and produce hydrogen for backup fuel cells. A handful of large units at a port can produce ammonia for marine fuel. All of these are sized to local demand. None of them require pipelines. All of them are physically located near where the hydrogen is consumed.

This is a fundamentally different architecture from the natural gas model, and it has different implications for the supply chain. The natural gas industry is organized around a small number of very large facilities. The distributed hydrogen industry is going to be organized around a large number of small to medium facilities, each at a customer site. The customer relationships look different. The financing looks different. The geographic distribution looks different.

It also has different implications for the equipment manufacturers. In a centralized model, the buyers are a few dozen large industrial gas companies and you're selling them very large plants. In a distributed model, the buyers are thousands of industrial customers and you're selling them much smaller, modular, standardized units. The unit economics, the manufacturing approach, and the engineering tolerances are all different. The competitive pressure on cost-per-stack is much higher in the distributed model, because the customer can comparison-shop in a way they can't with a one-of-a-kind centralized plant.

This is where the membrane fits in. The distributed model is much more sensitive to the cost and performance of the stack than the centralized model is, because the stack is a much larger fraction of the total system cost. A 50%reduction in stack cost is a more meaningful reduction in delivered hydrogen cost in a distributed system than in a centralized one. That's a long way of saying that membrane improvements compound faster in the distributed model.

The distributed model also benefits more from the kind of operational characteristics that a better membrane enables. A membrane that handles intermittent input lets the electrolyzer run directly off variable wind or solar without needing a large grid buffer, which matters a lot when you're sitting next to a solar farm or a wind site. A membrane that unlocks PEM electrolysis, which doesn't require a caustic electrolyte, matters more in distributed installations that may not have the same level of chemical management infrastructure, disposal, and expertise as a large central plant. A membrane that lasts 30 years instead of 7 means a distributed unit owner isn't replacing the unit three times during the useful life of the installation. Each of these is incremental in isolation, but they compound.

The bigger point, though, is that the distributed model is what makes the hydrogen economy actually scalable. Centralized models are great for industries that are already centralized — refineries, ammonia plants, steel mills. They're a poor fit for emerging applications like fleet operations, building energy, distributed power, and most of the other places where hydrogen actually has to compete with diesel and natural gas on the margin. The applications where hydrogen wins are mostly distributed, and they require an electrolyzer architecture, a supply chain, and a manufacturing strategy that match.

The thing I find encouraging about this is that it changes the politics. A centralized hydrogen industry would have looked like the existing oil and gas industry — a small number of very large players, concentrated geographically, dependent on huge infrastructure investments. A distributed hydrogen industry looks like the existing solar industry — many smaller players, geographically dispersed, deployed incrementally. The second pattern is much more compatible with the actual structure of American industrial geography, and it doesn't require the kind of permitting and pipeline buildout that the first one would.

So the question of what the hydrogen economy looks like isn't really an openone anymore. It's going to look distributed, because the technical and economic forces all point that way. The question is who builds the substrate and the products above it.

‍