The analogy I find myself reaching for most often when explaining what we do is the one to silicon. People sometimes hear it as a marketing claim, but I think it's more precise than that. Let me try to work through why.

Silicon, in semiconductors, is not a device. It's a substrate that devices get built on. If you walk through a semiconductor fab, you don't see anyone making a "silicon product." You see people putting layers of dopants and metals on top of silicon wafers, etching patterns into the layers, and slicing the wafers into chips. The chips are devices: a transistor, a diode, a logic gate, a memory cell, an image sensor, a solar cell, a power MOSFET, a MEMS accelerometer. The devices are very different from each other, both in what they do and in who buys them. But all of them are arrangements on silicon and a handful of dopants, processed with broadly the same kind of equipment, in the same kind of fab.

The semiconductor industry has roughly seven device categories sitting on top of one substrate. Those seven categories have produced something like a hundred end-product industries, from consumer electronics to satellites to medical imaging. The substrate is one material. The industries are many.

Now look at electrochemistry. A proton exchange membrane is a thin polymer film whose function is to let protons through while blocking other things. By itself, it doesn't do anything. To turn it into a useful device, you put catalysts on either side, sandwich it between electrodes, and connect the whole thing to a power source or a load. The result is a device.

The device categories are: fuel cell (hydrogen and air in, electricity out), electrolyzer (electricity and water in, hydrogen out), flow battery (electrolyte cycling, electricity stored), electrochemical hydrogen compressor (hydrogen in at low pressure, hydrogen out at high pressure), electrodialysis cell (mixed ionic streams separated), and membrane reactor (electrochemical synthesis of something specific, such as reduction of CO₂ to formate). That's six categories, depending on how you count.

Sitting on top of those six categories are the end-product industries: power generation, fuel production, energy storage, materials separation, chemical synthesis, water purification, recycling. The same substrate. Different arrangements. Different industries.

What I find striking when I lay it out this way is how exact the structural correspondence is. It's not that the membrane is "like" silicon in some hand-wavy sense. The relationship between substrate and devices and industries has the same shape in both cases. One material at the bottom. A small number of device categories in the middle. A much larger set of industries at the top.

The implications of this shape are also similar. In semiconductors, the company that controls the substrate sets the floor for the entire industry. The substrate decides what's possible. For a long time, Fairchild held that position by being able to make better silicon than anyone else. Today the position is held by a handful of companies most people have never heard of, supplying the entire semiconductor industry — one of them holds over 30% of the market. Either way, whoever can do the substrate best gets to decide what the device makers can build, which in turn decides what the industries on top can do.

The same logic should apply to membranes, and it largely does, except for one important detail: nobody has really competed on the membrane in thirty years. The substrate has been frozen since 1994 at roughly the same performance level, set by a polymer that DuPont invented for NASA in the 1960s. Everybody downstream has been working with the substrate they were given, the way the chip industry would have looked if silicon manufacturers had stopped improving the wafer in the 1970s.

This is the part that I think makes the analogy unusually load-bearing, not just decorative. The thing that would have happened in semiconductors if silicon wafer quality or size had stopped improving is the thing that has happened in electrochemistry. If Fairchild hadn't figured out planar silicon, Intel couldn't exist. The downstream industries are real, but they're constrained, and the constraints all trace back to the same physical layer. Push that layer forward and everything on top of it moves.

I sometimes meet people who think the analogy goes the other way — that electrochemistry is "like" semiconductors in the sense that it's an emerging hard tech field, but that the parallel doesn't extend to the structure of the industry. I think the analogy is much more literal than that. The substrate is doing the same job, in the same place in the stack, with the same gating effect on what's possible above it. The reason the analogy holds is that both industries have the same shape.

What that means in practice, and what that implies for the longer time horizon, is what I find most worth working out.

‍