Who is betting on what in AI, and against whom.
Physics AI and AI for science
Ideas are cheap now; experiments are the bottleneck. Replace the lab with a universal learned simulator that also tells you which direction to improve.
In plain words
Ideas are cheap now; experiments are the bottleneck. This camp wants a learned simulator good enough to replace most of the lab bench, and a loop around it that says which direction to improve a design.
The signal is different from text: simulation data, plus the governing equations themselves, which act as a reward you can check. The representation is continuous fields in space and time, handled by neural operators that work at any resolution. The proof case already exists: learned weather models beat the best physics-based forecasts from 2023 on.
The camp has two strands, and the lab pages below are sorted by them. Simulator-first labs (Prometheus, PhysicsX, Accelerated Understanding, Physical Superintelligence) bet on scale in the model: a learned simulator good enough to design inside. Lab-in-the-loop companies (Periodic, Lila, Radical, CuspAI, and the drug designers Isomorphic, Chai and Latent) bet on the loop itself: a model proposes, an automated lab tests, the result trains the next model. Bezos's Prometheus, at $41B, is the largest bet on the first strand; drug design is the most mature part of the second.
Unfamiliar terms are in the glossary.
Where it stands
Weather forecasting is already won by learned models. Materials, fusion and device design are early. Money moved fast in 2026: Prometheus raised $12B, PhysicsX and CuspAI raised $300M and $450M, and Accelerated Understanding came out of stealth in August with the most aggressive scale claim. Periodic and Lila bet on closing the loop with real robots in real labs.
What would prove them right. A device, material or molecule designed in the loop by the model that beats human plus classical simulation on a real-world benchmark, not just screened from a candidate list.
The disagreements, both sides cited
Scale the simulator, or close the loop with a real lab?
Is Prometheus in this camp?
What to watch
- A material, device or molecule designed in the loop that beats human-plus-classical-simulation on a real benchmark, not screened from a candidate list.
- Accelerated Understanding's first customer results after its August 2026 launch; the 5-trillion-data-point claim is unverified outside the company.
- Clinical-stage readouts from Isomorphic, the camp's most advanced commercial test. Alphabet's AI drug discovery platform Isomorphic Labs raises $600M from Thrive TechCrunch 2025-03-31
People (20)
Researchers 17
Kristin PerssonUC Berkeley and Lawrence Berkeley National Laboratory, Professor; director, Materials Project
Founders 3
Researchers have held a research or faculty role; scientist-founders count. Founders, executives and investors are listed separately so nobody mistakes a boardroom for a lab. Where someone argues for a different camp than the one they work in, it says so.
Where it is studied
Universities and institutes with people in this camp, from the roles recorded here. Not a ranking, and not complete: a place is listed when someone in the atlas works there.
- Lawrence Berkeley National Laboratory Gerbrand Ceder, Aditi Krishnapriyan, Kristin Persson
- MIT Regina Barzilay, Markus J. Buehler, Tommi Jaakkola
- UC Berkeley Gerbrand Ceder, Aditi Krishnapriyan, Kristin Persson
- Brown University Jérôme Darbon, George Em Karniadakis
- Caltech Anima Anandkumar, Andrew M. Stuart
- University of Cambridge Pietro Liò, Carola-Bibiane Schönlieb
- Carnegie Mellon University Amir Barati Farimani
- Cornell University David A. Muller
- Drexel University Joshua C. Agar
- EPFL Philippe Schwaller
- ETH Zurich Siddhartha Mishra
- Imperial College London Aron Walsh
- JKU Linz Johannes Brandstetter
- Mila Jian Tang
- Nanyang Technological University Yew-Soon Ong
- National University of Singapore Shyue Ping Ong
- Texas A&M University Shuiwang Ji
- UCL Marc Deisenroth
- University of Alabama Runlong Yu
- University of Amsterdam Max Welling
- University of Chicago Ian T. Foster
- University of Oxford Michael Bronstein
- University of Pennsylvania Paris Perdikaris
- University of Stuttgart Mathias Niepert
- University of Tennessee, Knoxville Sergei V. Kalinin
- University of Tokyo Koji Tsuda
- University of Toronto Alan Aspuru-Guzik
- University of Wisconsin-Madison Kyle Cranmer
Also publishing here (32)
Senior authors in this camp's core literature, found through the alphaXiv research index or an institutional profile, and listed on the strength of one paper each. Being on this list means they publish in the field, not that they have taken a side. The full list is on the People page.
- Joshua C. Agar Drexel University paper
- Alan Aspuru-Guzik University of Toronto paper
- Johannes Brandstetter JKU Linz paper
- Michael Bronstein University of Oxford paper
- Markus J. Buehler MIT paper
- Jérôme Darbon Brown University paper
- Marc Deisenroth Google DeepMind and UCL paper
- Amir Barati Farimani Carnegie Mellon University paper
- Ian T. Foster Argonne National Laboratory and University of Chicago paper
- Rafael Gómez-Bombarelli Lila Sciences paper
- Tommi Jaakkola MIT paper
- Shuiwang Ji Texas A&M University paper
- Sergei V. Kalinin University of Tennessee, Knoxville and PNNL paper
- Simon Kohl Latent Labs paper
- Karsten Kreis NVIDIA paper
- Aditi Krishnapriyan UC Berkeley and Lawrence Berkeley National Laboratory paper
- Pietro Liò University of Cambridge paper
- Siddhartha Mishra ETH Zurich paper
- David A. Muller Cornell University paper
- Mathias Niepert University of Stuttgart paper
- Shyue Ping Ong National University of Singapore paper
- Yew-Soon Ong Nanyang Technological University and A*STAR paper
- Paris Perdikaris University of Pennsylvania paper
- Philippe Schwaller EPFL paper
- Panos Stinis Pacific Northwest National Laboratory paper
- Jian Tang Mila and HEC Montréal paper
- Koji Tsuda University of Tokyo paper
- Rama K. Vasudevan Oak Ridge National Laboratory paper
- Aron Walsh Imperial College London and CuspAI paper
- Kevin G. Yager Brookhaven National Laboratory paper
- Runlong Yu University of Alabama paper
- Linfeng Zhang DP Technology paper
Labs and companies (17)
Simulator first
Learn a universal simulator of physical systems (neural operators, field models, 'world engines') and optimize designs inside it. Prometheus, PhysicsX, Accelerated Understanding, Orbital.
NASDAQ:GOOGLGraphCast and AlphaFold; AlphaEvolve for algorithm discovery. The AlphaFold team dispersed in 2026, with Jumper leaving for Anthropic.
Lab in the loop
Models propose molecules or materials, automated labs and instruments test them, the results train the next model. The signal here is real experiments, not simulation, which is what separates this strand from the first. Isomorphic, Chai, Latent, Recursion, Lila, Periodic, Radical, FutureHouse, Adaptyv.
Also active here: NVIDIA (Neural operators (FNO, FourCastNet) came out of NVIDIA research.); Sakana AI (The AI Scientist: automated research loops.).
Signature ideas
Neural operators (FNO)
Networks that learn maps between functions, not between fixed grids, so one model handles any resolution. The Fourier neural operator (2020) made them fast.
FourCastNet
Nvidia's 2022 global weather model on neural operators; the first to match numerical forecasts at a tiny fraction of the compute.
Direct 4D rollout
Predict a whole space-time field forward in one shot rather than stepping a solver. Accelerated Understanding's pitch.
PDE residual as reward
Plug the model's prediction back into the governing equation; the size of the violation is a training signal you never run out of.
Differentiable simulation
Write the simulator so you can take gradients through it, then optimize a design by descent instead of by trial.
Autonomous labs
Robots that synthesize and test what the model proposes, closing the loop without a human at the bench.
Cross-physics transfer
One model pretrained across fluids, solids, electromagnetics and chemistry, on the bet that the maths overlaps.
Read in this order
- Physics Informed Deep Learning (Part I) 2017
PINNs. The governing equations become part of the loss.
- Fourier Neural Operator for Parametric Partial Differential Equations 2020
Learn the solution operator, not one solution; resolution-invariant and orders of magnitude faster than solvers.
- Highly accurate protein structure prediction with AlphaFold 2021
The result that convinced funders a learned model can replace an experiment.
- Neural Operator: Learning Maps Between Function Spaces 2021
The general framework behind the simulator-first strand.
- Scaling deep learning for materials discovery 2023
GNoME: 2.2 million new crystals predicted, hundreds made in a lab. The lab-in-the-loop strand in one paper.
How money reaches this camp
| Lab | Access | Note |
|---|---|---|
| Google DeepMind | NASDAQ:GOOGL | Alphabet (GOOGL): GraphCast, AlphaFold lineage, Isomorphic. |
| NVIDIA | NASDAQ:NVDA | Nvidia (NVDA): neural operators, Earth-2, and an investor in PhysicsX, Periodic, Lila and Radical. |
| Prometheus | private | $41B; JPMorgan, BlackRock, Goldman on the round. |
| Periodic Labs | private | $1.3B seed valuation; reported talks at $7.5B. |
| PhysicsX | private | $2.4B; Siemens and Applied Materials as strategics. |
| Lila Sciences | private | $1.3B+; Flagship-incubated. |
| Isomorphic Labs | private | Alphabet subsidiary with outside investors since 2025. |
Not investment advice. Private valuations are what the last round implied; "reported" means the press, not the company, gave the figure. Full table on the Capital page.