Latent Space · 2026-02-25 · 34 min

🔬Searching the Space of All Possible Materials — Prof. Max Welling, CuspAI

Max WellingBrandonRJ

Podcast ↗

TL;DR

AI for science is “not just emerging, it’s exploding,” but Welling says the leap from hundreds of millions to billions also means “we are creating a new bubble here.” Protein folding and machine-learning interatomic potentials are successful examples; a Jeff Bezos startup’s cited $6.2 billion Series C is the capital-market extreme, while the opportunity moves AI beyond ad placement into drugs, energy materials, and carbon capture.
Materials are the upstream constraint on nearly every technology thesis, including AI itself. An LLM depends on a GPU, whose wafer fabrication and EUV lithography become materials problems as conventional scaling reaches its limits; in energy, Welling contrasts solar’s roughly 22% capture today with a theoretical 50% using perovskite layers. “Underlying almost everything is a material.”
The larger bet is to turn materials discovery from artisanal hypothesis-testing into a search engine over “the space of all possible molecules.” A user would specify what they want, while computation and remote experiments generate candidates and feedback; the opportunity extends from better batteries to plastics that decompose after weeks and become fertilizer.
CuspAI, founded by Welling and Chet Edwards, has become a roughly 40-person company with $130 million raised after about 20 months. Its climate premise is demanding: staying within two degrees would require zero emissions by 2050, then another 50–100 years of atmospheric carbon removal at about half today’s emissions rate.
The platform combines candidate generation, progressively costlier digital-twin screening, and experiments that Welling wants to treat as a “physics processing unit.” Cheap calculations eliminate obvious failures before multi-scale, multi-fidelity models and physical tests take over; agents increasingly search literature and orchestrate the workflow, while high-throughput experimentation is being added. Welling says the design is “not rocket science,” but implementation and data are “surprisingly hard”—and are the moat.
Automation will advance through a “retreat” of expert intervention, not a near-term dark laboratory that removes chemists. Teams first assemble modular workflows manually, then automate bounded tasks such as choosing and validating the right DFT calculation; Welling does not expect complete automation of domain expertise in the next five years and expects chemists to keep judging candidate materials.
Practical breakthroughs can arrive before full automation because each added capability is immediately useful. CuspAI is pursuing an undisclosed lighthouse material alongside narrower paid projects and a PFAS-removal partnership with Kamira; unlike fusion or quantum computing, the platform need not wait decades for an all-or-nothing payoff, though every new materials vertical still requires retraining and renewed human guidance.

Digest · the substance, structured for research

1. AI for science is exploding—and already resembles a bubble

Max Welling said his early career followed an internal “sensor” for deep questions—black holes, universe boundaries, quantum mechanics. With retirement still roughly ten years away, impact became a second axis: two-dimensional quantum gravity might yield papers but little worldly effect, while materials could combine difficult science with climate leverage.
“Physics is the thread”: symmetries from particle physics and general relativity informed his equivariant machine-learning work, while diffusion models led him toward stochastic thermodynamics. He now sees common mathematics spanning reinforcement learning, Schrödinger bridges, MCMC sampling, and physical systems outside equilibrium.
His enduring fascination includes Gerard ’t Hooft’s claim that quantum mechanics is wrong and his proposed replacement, which Welling called courageous even though “nobody understands what he’s saying.” Welling’s own ambition is more modest: understand quantum mechanics without “all the weird multiverses and collapses.”
Protein folding and machine-learning force fields showed that modern AI could move scientific methodology. As researchers sought impact beyond “ad placement” and multimedia, investment climbed from hundreds of millions into billions; Welling cited a $6.2 billion Series C for a Jeff Bezos startup, called the field a virgin field, and conceded: “We are creating a new bubble here.”

2. Materials discovery is becoming a search problem

Welling’s stack runs downward from software to matter: beneath an LLM sits a GPU, and beneath that sit wafer materials patterned with EUV light. With dimensional scaling near its limits, further semiconductor improvement increasingly becomes “an actual material problem.”
The energy transition is likewise materials-bound through batteries, fuel cells, and solar panels. Welling’s concrete example was adding perovskite layers over silicon, potentially capturing “theoretically up to 50%” of light versus “I don’t know, maybe 22” today.
Traditional discovery proceeds slowly from papers to hypotheses, experiments, and revised understanding. Welling wants instead to “search the space of all possible molecules”—including molecules neither manufactured nor found in nature—then refine a query as computation and experiments return a list of candidates.
That search-engine framing extends to sustainability: Welling argued that plastics could plausibly be designed to destroy themselves after a few weeks and become fertilizer. “These things can be done, right? And we should do it.”

3. CuspAI links digital models to a physics processing unit

CuspAI began about 20 months earlier from Welling’s climate concern and was started with co-founder Chet Edwards. His premise for staying within two degrees requires emissions reaching zero by 2050, followed by 50–100 years of directly removing CO₂ at roughly half today’s emissions rate—an unsolved problem whose failure could mean four degrees, which he described as “very bad.”
The company grew to about 40 people and collected $130 million, unusually large by Welling’s European benchmark. Its platform generates candidates, then walks them through a multi-scale, multi-fidelity digital-twin ladder: “Do the cheap things first,” discard obvious failures, reserve expensive calculations for survivors, and send the final few into experiments for feedback.
Literature-search agents and autonomous orchestration are being added at various levels of maturity. LLMs are being integrated, and CuspAI is moving toward high-throughput experimentation; Welling said self-driving-lab capabilities should also be added. He wants to treat experiments as a “physics processing unit”—nature, “possibly” the fastest computer known, computing alongside the data center.
Welling’s distinction is sharp: the architecture is “not rocket science,” but obtaining data and making the whole platform work is “surprisingly hard.” He said the moat is in the data the company can get its hands on and in actually building the platform.

4. Humans and industrial partners remain inside the loop

Automation starts with modular tools assembled manually by chemists. If a promising porous material collapses when shaken, the team adds a stability test; only after the workflow is understood might a Bayesian optimizer or an LLM trained to be a good chemist learn which tools to invoke and in what order.
DFT illustrates the gradual “retreat” of expertise: today a non-specialist consults an expert about the right DFT calculation for the problem, how long to run it, and whether the result is good; software might automate those bounded decisions. The eventual interface may return candidate materials from a query, but the chemist still decides what is actually good.
One host tested the stronger vision of a fully automated process. Welling rejected a “completely dark lab” that receives “find something interesting” and independently defines interestingness: automating away lifelong domain expertise “in the next five years? I don’t think that’s going to happen at all.”
CuspAI only starts investing in a materials direction after finding a good industrial partner. It is pursuing an undisclosed lighthouse material as an AI-enabled real-world proof point, while accepting narrower force-field projects; its PFAS water-filtration work with Kamira exemplifies a deep partnership. Each capability is immediately useful, but new verticals require different experiments, instruments, retrained and fine-tuned models, and renewed human direction.

5. Physics priors help, but the bitter lesson survives

Welling explained equivariance with a rotated bottle: an ordinary network may have to relearn every orientation, while a symmetry-aware network generalizes from one, reducing data needs. Rotations, translations, and graph permutations can be built into the model as constraints on its weights.
The host’s pushback—worth keeping: why not use data augmentation? Welling conceded that augmentation sometimes works better because exact equivariance complicates the optimization surface; encoding a true symmetry saves data, but an imperfect constraint can make good minima harder to find.
His synthesis was “a trade-off between data and inductive bias.” A wrong bias imposes a performance ceiling, and the applicable bitter lesson from LLMs is that architectures must scale unless the dataset is tiny.
Welling’s forthcoming Generative AI and Stochastic Thermodynamics, which he said he had sent to the publisher, develops the same cross-fertilization. It maps generative AI onto nonequilibrium statistical mechanics, connecting variational free energy, Geoff Hinton and Radford Neal’s work, Karl Friston’s free-energy principle, and fluctuation theorems. He hoped to have it in hand for an ICLR keynote in April, while stressing that publication timing remained outside his control.

Speaker 0

I want to think of it as what I would call a physics processing unit, like a PPU, right? You have digital processing units, and then you have physics processing units. It’s basically nature doing computations for you. It’s possibly even the fastest computer known.

swyx

Yeah.

Speaker 0

It’s a bit hard to program because you have to do all these experiments. It’s also quite bulky. It’s a very large thing you have to do. But in a way, it is a computation, and that’s the way I want to see it.

You can do computations in a data center, and then you can ask nature to do some computations, right? Your interface with nature is a bit more complicated, but these things will have to seamlessly work together to get to a new material that you’re interested in.

swyx

It’s a pleasure to have Max Welling as a guest today. Max has done so much over his career that I’ve been excited about. If you’re in the deep learning community, you probably know Max for his work on variational autoencoders, which has literally stood the test of time—officially stood the test of time.

If you are a scientist, you probably know him for his pioneering work on graph neural networks and equivariance. And if you’re in materials science, you probably know him for his new startup, CuspAI. Max has a long history of doing lots of cool problems. You started in quantum gravity, which is very different from all these other things you’ve worked on.

The first question, for AI engineers and scientists: What is the thread in how you think about problems? What is the thread in the types of things that excite you? And how do you decide what the next big thing you want to work on is?

1. Curiosity Meets Real World Impact

Speaker 0

It has actually evolved a lot. In my younger days, let’s put it that way, I would just follow what I found super interesting. I have this sensor that I think many people have, but maybe don’t really use very much, which is that you get this feeling of getting very excited about some problem.

It could be what’s inside of a black hole, what’s at the boundary of the universe, or what quantum mechanics is actually all about. And so I followed that basically throughout my career, but I have to say that as you get older, this changes a little bit.

swyx

Mm-hmm.

Speaker 0

In a sense, there’s a new dimension coming to it, and this is impact.

swyx

Mm-hmm.

Speaker 0

Working in 2D quantum gravity, you pretty much guarantee there’s going to be no impact in what you do relative to this world. Maybe a few papers, but not in this world, at this energy scale.

As I get closer to retirement—which is fortunately still 10 years away or so—I do want to make a positive impact in the world. I got pretty worried about climate change, and politics seems to have a hard time solving it, especially these days. So I thought I’d better work on it from the technology side, and that’s why we started CuspAI.

But there are also a lot of really interesting science problems in materials science. It’s combining both the impact you can make with it and the interesting science. So it’s these 2 dimensions: working on things where you feel, “Oh, there’s something very deep going on here,” and, on the other hand, trying to build tools that can actually make a real impact in the world.

Alessio Fanelli

The thread, when I look back at the different things that you’ve worked on, is that some of them seem pretty connected, like the physics to equivariance and graph neural networks, maybe. And that seems to be somewhat related to CuspAI. Do you have a thread through there?

2. Physics Connects AI And Science

Speaker 0

Yeah. I think physics is the thread. Having spent a lot of time in theoretical physics, I think there are, first, very fundamental and exciting questions—things that haven’t actually been figured out in quantum gravity—so that is really the frontier.

There are also a lot of mathematical tools that you can use. In particle physics, but also in general relativity, symmetries play an enormously important role, and this goes all the way to gauge symmetries as well. Applying these kinds of symmetries to machine learning was a very deep and interesting mathematical problem.

I did this with Taco Cohen, and Taco Cohen was the main driver behind this. It went all the way from simple rotational symmetries to gauge symmetries on spheres and stuff like that. Maurice Weiler, who’s also here, wrote an entire book when he was a PhD student with me, which I can really recommend, about the role of symmetries in AI and machine learning. I find it’s a very deep and interesting problem.

More recently, I’ve taken a different path: the relationship between diffusion models and a field called stochastic thermodynamics. This is basically thermodynamics, which is a theory of equilibrium, but formulated for out-of-equilibrium systems.

It turns out that the mathematics we use for diffusion models—and even for reinforcement learning, Schrödinger bridges, and MCMC sampling—has the same mathematics as this physical theory of nonequilibrium systems. That got me very excited.

Actually, when I taught a course in Muizenberg, South Africa, close to Cape Town, at the African Institute for Mathematical Sciences, or AIMS, I turned that into a book. Two years later, the book is finished. I’ve sent it to the publisher, and this is about the deep relationship between free energy, diffusion models—basically generative AI—and stochastic thermodynamics.

I find physics very deep. I also think a lot about quantum mechanics.

swyx

Yeah.

Speaker 0

It’s a completely weird theory that nobody really understands.

swyx

Mm-hmm.

Speaker 0

There’s a very interesting story that may be good to tell to connect my PhD back to where I am now. I did my PhD with a Nobel laureate, Gerard ’t Hooft. He’s just the most brilliant man I’ve ever met. He was never wrong about anything, as long as I’ve seen him, and now he says quantum mechanics is wrong and has a new theory of quantum mechanics.

Nobody understands what he’s saying, even though what he’s writing down is not mathematically very complex. But he’s trying to address the understandability, let’s say, of quantum mechanics head-on. I find it very courageous, and I’m completely fascinated by it.

So I’m also trying to think: Can I actually understand quantum mechanics in a more mundane way, without all the weird multiverses and collapses and stuff like that? Physics has always been the thread, and I am trying to apply physics to machine learning to build better algorithms.

swyx

You’re still very involved in understanding physics and the world, even beyond just applications to machine learning or introducing new formalisms. That’s really cool.

Speaker 0

Yes, I would say I’m not contributing much to physics, but I’m contributing to the interface between physics and science, and that’s called AI for science or science for AI. It’s actually a new discipline that’s emerging.

It’s not just emerging; it’s exploding, I would say. That’s the better term, because investments have gone from the hundreds of millions into the billions. There’s now actually a startup by Jeff Bezos that has a $6.2 billion Series C round, right? It’s insane. This is the largest startup ever, I think, and that’s in this field, AI for science. It tells you something: we’re creating a new bubble here.

swyx

Yeah.

Speaker 0

Right?

swyx

So why do you think it is? What has changed that has motivated people to start working on AI-for-science-type problems?

Speaker 0

There are 2 reasons, actually. One is that people have been applying the new tools from AI to the sciences, which is quite natural. Of course, I think there are 2 big examples.

Protein folding is, of course, a big one, and the other one is machine-learning force fields, or what are sometimes called machine-learning interatomic potentials. Both of them have been very successful. Both also have something to do with symmetries, which is also cool.

People in AI for science saw an opportunity to apply the tools they had developed beyond ad placement or multimedia applications to something that could actually make a very positive impact in society, like health, drug development, materials for the energy transition, and carbon capture. These are all really cool, impactful applications.

swyx

Yes.

Speaker 0

Besides that, the science itself is also very interesting. The fact that these 2 fields are coming together, and that we’re now at the point where we can actually model these things effectively and move the needle on some of these scientific methodologies, is also a very unique moment, I would say.

And people recognize that, okay, now we're at the cusp of something new. This is also what our company is called after. We're at the cusp of something new, and of course, that always creates a lot of energy. It's like, okay, there's something. It's sort of a virgin field, right? It's like nobody's been there. I can rush in, and I can start harvesting there, right?

Speaker 1

Yeah. Yeah.

Speaker 0

And I think that's also what's causing a lot of enthusiasm in the field.

Speaker 2

If you're an AI engineer—which many of the people who listen to this podcast will be—and you maybe don't have a strong science background but are excited, how does somebody who is not a scientist on a day-to-day basis get involved? Most, I would say, most AI practitioners, be they engineers or scientists, would consider themselves scientists, and they have some background—a little bit of physics, a little bit of chemistry—in college, maybe even graduate school, and have been working or are starting out. How does somebody who is not a scientist on a day-to-day basis get involved?

Speaker 0

Well, they can read my book once it's out. We should create curricula that are on this interface. I'm not sure. There are possibly already, at some universities, actual courses you can take—maybe online courses you can take. These workshops where we are now are actually very good as well, and we should probably have more tutorials before the workshop starts. Actually, I've proposed this at some point: maybe first have an hour of a tutorial so that people can get new people into the field.

But yeah, there's a lot out there. Most of it is, of course, inaccessible. But I would say we will create many more books and other content—

Alessio Fanelli

Yeah.

Speaker 0

—that is more accessible, including this podcast, I would say. I think it will come. These days, you can watch videos and things. There's a huge amount of content you can go and see.

swyx

So maybe a follow-up to that. How do people learn and get involved, but why should they get involved? I mean, a lot of people in our audience will be interested in AI engineering, but they may be looking for bigger impacts in the world.

Speaker 0

Yeah.

swyx

What opportunities does AI for science provide to make an impact, to change the world, that working in the world of pure bits would not?

3. Materials Become A Search Engine

Speaker 0

My view is that underlying almost everything is a material.

swyx

Yeah.

Speaker 0

We're focusing a lot on LLMs now—

swyx

Yeah, yeah.

Speaker 0

—which is kind of the software layer. But I would say, if you think very hard, underlying everything is a material. I was saying, you know, there's the LLM, and underlying the LLM is a GPU on which it runs. In order to make that GPU, you have to put materials down on a wafer and shine it with EUV light in order to etch the structures in. But that's now an actual material problem because, more or less, we've reached the limits of scaling things down, and now we're trying to improve further with new materials. So that's the fundamental materials problem.

We need to get through the energy transition fast if we don't want to mess up this world.

swyx

Mm-hmm.

Speaker 0

For instance, there are batteries. That's a complete materials problem, right? There are fuel cells. There are solar panels. They can now make solar panels with new perovskite layers on top of the silicon layers that can capture, theoretically, up to 50% of the light, whereas now we're at, I don't know, maybe 22% or something.

swyx

Yeah.

Speaker 0

Right? So these are huge changes, all through material innovation. And yeah, I think wherever you go, I can probably dig deep enough and then tell you, “Well, actually, the very foundation of what you're doing is a material problem.”

I think it's just very nice to work on this very foundation, also because I think this is maybe something that's happening now: we can start to search through this material space. This has never been the case, right? The normal way of working is you read papers, and then you come up with a hypothesis, you do an experiment, and you learn, et cetera. So that's a very slow process.

Now we can treat this as a search engine. Like we search the internet, we can now search the space of all possible molecules—not just the ones that people have made or that are in the universe, but all of them.

swyx

Yeah.

Speaker 0

Right? And we can make this kind of fully automated. That's the hope, right?

swyx

Yeah.

Speaker 0

We can just type what we want. It becomes a tool where you type what you want, and something starts spinning—

swyx

Mm-hmm.

Speaker 0

—and some experiments get going.

swyx

Yeah.

Speaker 0

Right? Then out comes a list of materials, and you look at it and say, “Maybe not,” and then you refine your query a little bit.

swyx

Yeah.

Speaker 0

You kind of do research with this search engine, where a huge amount of computation and experimentation is happening somewhere far away, in some lab or some data center or something like this. I find this a very, very promising view of how we can build a much better materials layer underneath almost everything, and also more sustainable materials.

Our plastics are polluting the planet, right? If you can come up with a plastic that destroys itself after, I don't know, a few weeks, and actually becomes a fertilizer, these are things that are not impossible at all. These things can be done, right? And we should do it.

swyx

Can you tell us just generally a little bit about CuspAI? Then I have a ton of questions.

4. CuspAI Targets Carbon Removal

Speaker 0

Yeah. So CuspAI started about 20 months ago because I was worried about—I’m still worried about—climate change.

swyx

Mm-hmm.

Speaker 0

I realized that in order to stay within 2 degrees, let's say—

swyx

Mm-hmm.

Speaker 0

—we would not only have to reduce our emissions to zero by 2050, but then have another half-century or even a century of removing carbon dioxide from the atmosphere—not by reducing our emissions, but actually removing it at a rate that's about half the rate at which we now emit it.

That is an unsolved problem. And if we don't solve it, 2 degrees is not going to happen, right? It's going to be much more. I don't think people quite understand how bad that can be—4 degrees, very bad.

So this technology needs to be developed, and this was my motivation, along with my co-founder, Chet Edwards, to start this startup. We also saw that the technology was ready, which is very good. If you're going to do it, the time is right.

In the meanwhile, we've grown to about 40 people. We've collected $130 million in investment into the company, which for a European company is quite a lot. It's interesting that right after that, other startups got even more, so that tells you how fast this is growing.

We are now at the point where we've built the platform, of course, but it's for a series of material classes, and it needs to be constantly expanded to new material classes. It can be more automated because we're now putting LLMs in, so the whole thing gets more and more automated.

Now we're moving to high-throughput experimentation, connecting the actual computational platform to the experiments so that you can also get fast feedback from experiments. I kind of think of experiments as something you do at the end, although that's what we've been doing so far. I want to think of it as what I would call a physics processing unit, like a PPU.

swyx

Yeah.

Speaker 0

You have digital processing units, and then you have physics processing units. It's basically nature doing computations for you. It's the fastest computer known, possibly even.

It's a bit hard to program because you have to do all these experiments. It's also quite bulky. It's a very large thing you have to do.

swyx

Yeah.

Speaker 0

But in a way, it is a computation, and that's the way I want to see it. You can do computations in a data center, and then you can ask nature to do some computations, right? Your interface with nature is a bit more complicated. But then these things will have to seamlessly work together to get to a new material that you're interested in. That's the vision we have.

We don't say superintelligence because I don't quite know what it means, and I don't want to oversell it. But I do want to automate this process and give a very powerful tool to chemists and materials scientists.

swyx

That actually brings up a question I wanted to ask you. First of all, can you talk about your platform to whatever degree you can? Explain how it works and what your thought processes were in developing it?

5. Inside The Automated Materials Platform

Speaker 0

Yeah. Actually, it's been surprisingly—it's not rocket science, I would say.

It's not rocket science in the sense of the design. Basically, the design that I wrote down at the very beginning is still more or less the design, although you add things. I wasn't thinking very much about multi-scale models, and it came on our radar that multi-scale is actually very important. In the beginning, I wasn't thinking very much about self-driving labs, but now I think we're at the stage where we should be adding that. So there are bits and details that we're adding.

But more or less, it's what you see in the slide decks here as well: there's a generative component that you have to train to generate candidates, and then there's a digital twin, multi-scale, multi-fidelity—

swyx

Mm-hmm.

Speaker 0

…digital twin, which you walk through the steps of the ladder. They do the cheap things first. You weed out everything that's obviously not useful, and then you go to more and more expensive things later. So you narrow things down to a small number. Those go into an experiment: you do the experiment, get feedback, et cetera.

Now, things that have also been added more recently are more agentic parts. We have agents that search the literature and come up with chemical suggestions for doing experiments. We have agents that autonomously orchestrate all of the computations and experiments that need to be done. They're in various stages of maturity, and they can be continuously improved, I would say. So that's basically it. I don't think that part is rocket science. But the design of that thing—

swyx

Mm-hmm.

Speaker 0

…is not surprising. What's surprising is that it's surprisingly hard to actually build it, right? So that's the thing where the moat is in the data that you can get your hands on and in actually building the platform.

swyx

Yeah.

Speaker 0

And I would say there are 2 people in particular I want to call out: Felix Hunke, who is actually building the scientific part of the platform, and Alessandro De Maria, who is building the MLOps part of the platform.

Recently, we also added Aaron Walsh to our team, who is a very accomplished scientist from Imperial College. We're very happy about that. He's going to be our chief science officer. We also have a partnerships team that seeks out all the customers, because I think this is one thing I find very important.

It's so complex to actually bring a material into the real world that you must do this in collaboration with the domain experts, which are typically the companies. So we only start to invest in a direction if we find a good industrial partner to go on that journey with us.

swyx

Makes a lot of sense. Over the evolution of the platform, did you find that human intervention—human feedback—was needed at different steps? I guess you could start out with a purely automatic process. You could imagine 2 directions: 1, you start out making everything purely automatic—automated, agentic, and so on.