Bubbles, Cryotrons, and CNTs
Edition #011: Imagine a bubble that only bursts when there’s somewhere better to go with Cyrus Mody, Author of Greedy Science: Creating Knowledge, Making Money, and Being Famous in the 1980s
How can you tell a fad from a trend? Most people rely on frameworks. I do too, here are my favourites.
1. The adoption-shape test (Rogers’ diffusion / the chasm). A fad spreads within a group: enthusiasts, insiders, the terminally online, and burns out from there. A trend broadens across groups, jumping from early adopters into segments that have nothing in common with them and no reason to signal membership. Speed matters less than whether it keeps bringing in new people.
2. The driver test (structural vs. attentional). A trend is downstream of a durable shift, in cost, capability, demographics, or regulation, that would keep going even if everyone stopped talking about it. A fad has nothing underneath except attention and status, so it dies the moment the attention moves on. This is Amara’s Law: the stuff we overestimate in the short run is usually running on attention, and the stuff we underestimate has some compounding mechanism grinding away underneath, which is why it surprises everyone later.
3. The cost-curve test (learning curves / Wright’s Law). Especially useful in hardware. Is there a curve underneath it e.g. cost per unit, performance per watt, yield, that improves predictably with volume and displaces the old thing? If there’s a slope, it’s probably substitution in progress. If the economics are flat, its likely a meme and there be dragons.
Notice, all 3 are reverse-engineered after the fact. Rogers, Amara, Wright each generalised from things that had already happened. They tell you the shape a trend usually takes, which is no help deciding whether the thing right now is a trend or not. Very hard for a founder, VC, or futurist out here. But the people who study these cycles for a living tend not to use frameworks. They use cases.
So Alvin told me to go talk to Cyrus Mody. Mody is the historian who wrote the long view on exactly this: 3 hype cycles, 3 crashes. He’s at Maastricht now, having left Rice in 2015. His last book is The Squares: US Physical and Engineering Scientists in the Long 1970s (2022); more recently he has a chapter on the hyped technologies of the ‘80s in Greedy Science: Creating Knowledge, Making Money, and Being Famous in the 1980s (2025). And he co-runs the ERC Synergy project Nanobubbles, which studies why the scientific record is so hard to correct.
He was the perfect person to tell me: is CNT a fad or trend?
What I Learned
1. A bubble ends when there’s an alternative tool to migrate to.
The STM-DNA case, 1990: people were claiming to image the double helix with the scanning tunneling microscope. The artefact critique came and went, the bubble carried on.
It burst when commercial atomic force microscopes turned up and the whole field migrated overnight.
Same handoff with cryotrons to Josephson Junctions in the 1960s, then Josephson Junctions to quantum in the ‘80s.
A field needs somewhere new to put its money and its careers before it’ll let go of the old thing.
2. Ultimate-result brands attract serial bubbles.
Quantum, AGI, fusion, nanotech all promise a leapfrog beyond whatever we can do now, and the promise outlives any specific failure inside it.
Mody’s superconductor arc: cryotrons in the ‘50s, Josephson Junctions in the ‘60s and ‘70s, quantum in the ‘80s, 3 bets across 30 years, all flying the same flag.
The next bubble is usually inflating before the last one’s finished popping, so the money keeps coming back even when the specific bet keeps failing. Which should worry me more than it does.
3. The CNT-as-copper-replacement story is 1970s alternative energy again.
Mody’s closest analogue is 1970s solar: high oil prices pulled every alternative-energy technology to the edge of commercial scale, until oil cracked in the ‘80s.
The companies that had bet everything on relative price bailed and missed the customers who would have paid extra for solar regardless.
If you’re riding only on where copper prices are in 2026, you’ll get crushed if copper stabilises. The cultural reasons people might want CNT (lighter aircraft, cleaner data centres, a less Chinese-supplied conductor) need to be real in the pitch.
* * *
I. How a Historian Reads a Bubble
Lawrence: Hey Cyrus, every VC asks this at some point. What’s your framework? You’ve studied multiple technology cycles, do you reach for Perez surge cycles, or Christensen, the diffusion curve, Wright’s Law, or some taxonomy of your own?
Cyrus: Not so much. Historians tend to be more interested in the differences between cases than in similarities. Sociologists, people in management studies, innovation-studies scholars, they want to find the thing that unites all these different cases. For a historian, what’s more interesting is how a specific case fits into a specific time and place and exemplifies that time and place. How it’s different from other cases that might on the surface look similar.
When I’m writing as a historian I pay very careful attention to that specific context. But I’d like the people who are reading it with the kind of applied view that you have to take something from it. For them, the commonalities, the comparisons to other cases, are going to be what’s relevant. If people want to make those comparisons, all power to them.
Lawrence: Take me through the Nanobubbles project then. How did you end up on something as specific as why the scientific record is so hard to correct?
Cyrus: It’s a long story. My dissertation and first book were a history of probe microscopy: the scanning tunneling microscope, the atomic force microscope, a whole family of a couple of dozen other probe microscopes. As part of that study I looked at an episode right around 1990 where people were using the STM to look at organic molecules in air. They were seeing some interesting things, generating interesting images. But there were good reasons to think that maybe the STM actually can’t image organic molecules in air, or at least not very well, and might be prone to real artefacts.
What happened was the people claiming they could see really interesting things (that maybe they could image the double helix, that maybe they could even use it to sequence DNA) got a lot of publicity very quickly. And then all of a sudden some people pointed out that, yeah, these are all artefacts. Virtually none of it is real. The bubble burst very quickly. Though it really burst right at the moment commercial atomic force microscopes were becoming available, and people could leap from the one to the other pretty easily.
Lawrence: So the bubble burst when the AFM was ready, more than when the critique arrived. Is that the mechanism you keep seeing?
Cyrus: People are very reluctant to move on if they don’t have a new place to move to. That’s true throughout the history of science and engineering. If the only thing you have is a critique, the critique often just sits there, ignored. If you have a critique plus an alternative tool, suddenly the field migrates.
* * *
II. The Long Cryotron-to-Quantum Arc
Lawrence: Was probe microscopy your way into nanotech as a theme, or did the field find you?
Cyrus: Probe microscopy was basically the first, then I branched out. My second book was about other areas of nanotechnology tied to the microelectronics industry: molecular electronics, superconducting computing (which is what’s now called quantum computing), nanobiology, things like that.
Lawrence: Quantum is a good example for me. As a VC I’ve never come across another field where so much money has been poured into something where the binary risk is still real. AI, crypto, you can argue the merits. With quantum it’s still unclear whether we’ll ever get a fault-tolerant quantum computer. How does a historian read that arc?
Cyrus: That story goes way, way back. As long as there have been solid-state digital computers, or dreams of them, there have been people saying you can try and make this thing with semiconductors but the real ultimate technology would be to do it with superconductors. The very first integrated circuit was actually with superconducting materials. It was called a cryotron, invented at MIT in the early 1950s. There was a mini-bubble around cryotrons in the ‘50s.
When semiconducting integrated circuits took off in the late ‘50s and early ‘60s, people said fine, we’ll give up on this cryotron thing. But almost immediately, other people came along and said, well, there’s this crazy Welsh guy Brian Josephson who says you can do totally insane devices with superconductors. Some of them actually took all the equipment that had been used for making cryotrons and adapted it for making Josephson Junctions. Over the 1970s, especially IBM (backed by the National Security Agency), but also Bell Labs and some others, tried to build a Josephson Junction computer. They got really close. Then in the early ‘80s they did a really rigorous study and basically said what you just said: actually, we can’t make this fault-tolerant. It’s going to have bugs in it no matter what. We’ve got to give up on this technology.
Almost immediately after that technology died, the quantum folks came in with all these ideas that are still with us today, that may or may not work. It’s a technology where you see these bubbles over and over again because people think there’s an ultimate result better than anything we have now.
Lawrence: So are the bubbles structural, or aren’t they.
Cyrus: Somewhere in between those two poles. I don’t know if it’s structural. I’m reluctant to say it’s inevitable. But once you have a kind of technology brand that has this idea of a leapfrog associated with it, of an ultimate result, you’re going to have entrepreneurial types (not necessarily commercial entrepreneurs, but also researchers who are “entrepreneurial” in the sense of getting people to run with their ideas) who take that and say, yeah, the way we tried to do that before didn’t work, but this new way, we can get further this time.
* * *
III. From STM to CNT: Where This One Fits
Lawrence: Take me back to CNTs specifically. You were at Rice from 2008 to 2015, a peak period for the field. How did you read it then?
Cyrus: I haven’t kept up with it really, in fairness. I was at Rice 2008 to 2015, obviously knew lots of CNT people back then, but it’s been a while. It was a funny material. On the one hand, if you really wanted to make a big impact with it, you were going to have to get old-school chemical-engineering expertise in to scale it up. But the applications people were talking about were more material science, electronics, really specialty materials. The handoff between the two always seemed a bit vexed, at least back then.
You could make incredibly high-quality CNTs in Rick Smalley’s lab. But scaling it up was the issue. People never seemed quite satisfied either with the cost, or the quality, or the applications. It got stuck in a valley somewhere.
Lawrence: The story I keep hearing in 2026 is that data-centre buildout and electrification have tightened the copper market enough that CNT-as-copper-replacement is finally selling. Does that pattern: demand created by a commodity-price shock, map to anything else you’ve studied?
Cyrus: Not anything in nanotech that I’ve seen. But it sounds rather similar to a lot of the cases in alternative energy in the ‘70s that I’ve been working on. That was an era when oil prices were sky-high and that allowed everyone to get into every conceivable alternative-energy technology, with the idea that high oil was going to make it possible to scale up from incredibly expensive applications (solar power was mostly being used on satellites, where NASA could pay any amount) to a tiny fraction of that cost, where consumers were willing to put it on their houses. People got right up to the edge of making that leap. And then suddenly the price of oil started to drop, and they said: actually, no, this isn’t going to work. You could imagine a different story where the price of oil stayed high, and that pulled solar and wind and nuclear into scale-up much earlier.
Lawrence: The risk being that if copper stabilises, the demand story collapses.
Cyrus: That’s always going to be a big factor. I tend to think it’s good to have some other factor alongside, because if you’re pinning everything on the price of some other commodity, that commodity’s price can go up and down.
Solar in the ‘70s is a good example. All the big companies that got into solar pinned their hopes solely on solar competing with an oil price that was very high. When that went down they lost interest, and they completely ignored the more cultural reasons people might have been interested in solar. They thought consumers would only be interested if it was cheap. That meant they ignored a whole market that might have been willing to pay a little extra.
Lawrence: Last one. The hottest VC bet right now is AI for materials discovery. See CuspAI just this week. People expect the model to surface a novel material with conductivity or strength properties no one has imagined. From a historian’s seat, how do you read that?
Cyrus: Materials are funny. You have these leaps where suddenly a new class of material becomes visible, and then suddenly you can make a lot of variants on that core type of new material very quickly. I could imagine a use for AI in coming up with a lot of those variants on something that’s known, and seeing whether this variant is good for that, whether this one is good for something else. It’s hard for me to imagine AI really getting us to an entirely novel class of materials like CNTs or high-Tc superconductors.
Lawrence: Anything in the nano space you’re really optimistic about?
Cyrus: Not so much. This is maybe a disciplinary thing. Historians are trained to be critical and pessimistic. Our inclination is always to burst bubbles, to say yeah, okay, you’re hyping this, but here are all the problems with it. To identify the naysayers and bring their voices out. There’s tons of nano startups, lots of academic entrepreneurs with a very clear vision about curing cancer or whatever. Maybe that’ll work out. But the current state is that it’s going a lot slower than they imagined. There’s always people pointing out that what was promised 10 years ago, 10 years later, still isn’t here.
Lawrence: My job is to be overly optimistic and how fast and how far change will come. So maybe together we could come up with good answers. I’ll be in touch!
* * *
So where does that leave the nanotubes? Further along than any of my frameworks would guess. Go back to #001 and this was a lab material with a mountain of papers and no customer attached. A few (short) years later, today it is something like 35,000 tonnes a year, 80% of it going into batteries, which is a proper supply chain. So the fad question is settled as far as I am concerned. What is left is a bit more boring but how scaling actually works. The standards gap from #009 is still open, identical-looking tubes still price from €40/kg to several thousand, and the reliability and longevity data that gets you designed into an accelerator takes thousands of hours. Copper replacement is the newest of the stories and the least proven. I think it gets there. I just think it gets there slowly, on qualification data and audited supply. That said, how slowlty is hard to predict, when we are talking about one of the fastest growing markets/infrastructure build outs of our time.
Also, don’t forget, we have a great back catalog for you:
CNTs & Batteries: Where the Tonnage Goes (#008). batteries already take most of the tonnage, silicon anodes swelling 300% need a CNT scaffold, and 85% of supply sits in Asia.
CNTs & The Standards Gap (#009). identical-sounding nanotubes price from €40/kg to several thousand and nobody can audit the difference, which is what stops OEMs speccing them in.
Liar, Liar, Battery Supplier (#010). Charlotte Hamilton of TAF Carbon on 120 years of cherry-picked battery data, and why one coin cell of data means walk away.
Thanks for reading the CNT Dispatch. If you have questions for Cyrus or want to argue with the historian-pessimism take, reply to this email. The lessons in this piece are mine. The historian’s job is to keep the cases separate. Hold onto the AFM case. If you’re a founder pitching a leapfrog beyond the current generation of anything (quantum, fusion, AGI, room-temperature superconductors, novel materials from AI), the historian’s first instinct will be to find the naysayers and bring their voices out. Make it harder for him. Thanks, Lawrence x