Who is betting on what in AI, and against whom.
Physical AI and robotics
The architecture is roughly settled; data is the bottleneck. Scale vision-language-action models on teleoperation, human video and simulation until robots generalize.
In plain words
The robotics camp mostly agrees with the LLM camp about architecture: a vision-language model with an action head, trained on teleoperated demonstrations, human video and simulation, then fine-tuned on the robot. What it needs is data, and the whole field is organized around getting it: fleets, teleoperation farms, simulators, and cheap hardware.
The proof case is a robot doing a task in a home it has never seen. Physical Intelligence's pi0.5 showed generalization of that kind in 2025; Figure reports robots on a BMW line; Agility's Digit has the most warehouse hours. Humanoid demos are still teleop-heavy, and Tesla admitted in January 2026 that no Optimus was doing useful work.
Money follows the platform thesis. Skild and Physical Intelligence are priced like AI infrastructure, not robot makers. China dominates installations, and Unitree's August 2026 listing in Shanghai gave the camp its first large public pure play. Levine's wing is philosophically closer to the experience camp: robots should learn from their own deployment.
Unfamiliar terms are in the glossary.
Where it stands
Less a scientific bet than an application layer: it imports architectures from camp 1 and simulators from camp 2. Real deployments exist in warehouses and laundries; humanoid demos are still teleop-heavy and Tesla missed its 2025 Optimus target entirely. Valuations moved faster than deployments: Skild tripled to $14B in seven months, Unitree closed its first trading day up 460%. Levine's wing is philosophically closer to the experience camp.
What would prove them right. A general-purpose robot doing unseen household tasks at useful reliability without per-site data collection.
The disagreements, both sides cited
Humanoid or task-specific?
One brain for every body, or vertical integration?
Is this a camp at all?
What to watch
- A general-purpose robot doing unseen household tasks at useful reliability without per-site data collection.
- Optimus production numbers from Fremont in the second half of 2026, against a 2025 target of 10,000 that was missed entirely.
- Unitree's and Agility's first public quarters: revenue per robot and installed base, the first audited figures in the camp. Unitree Robotics kicks off STAR Market IPO pricing Shanghai Stock Exchange / Global Times 2026-08-06
People (26)
Researchers 19
Founders 6
Executives 1
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.
- UC Berkeley Pieter Abbeel, Ken Goldberg, Sergey Levine, Jitendra Malik, Koushil Sreenath
- Carnegie Mellon University Deepak Pathak, Guanya Shi
- Peking University Wang He, Shanghang Zhang
- Stanford University Chelsea Finn, Shuran Song
- Beijing Institute for General Artificial Intelligence Siyuan Huang
- Caltech Aaron D. Ames
- ETH Zurich Marco Hutter
- Fudan University Yanwei Fu
- Georgia Institute of Technology Sehoon Ha
- Institute of Automation, Chinese Academy of Sciences Zhaoxiang Zhang
- KAIST Minjoon Seo
- MIT Leslie Kaelbling
- MIT CSAIL Daniela Rus
- Nanyang Technological University Ziwei Liu
- Oregon State University Alan Fern
- Princeton University Anirudha Majumdar
- Simon Fraser University Xue Bin Peng
- Tsinghua University Hang Zhao
- UC San Diego Xiaolong Wang
- University of Texas at Austin Yuke Zhu
- University of Washington Dieter Fox
- Westlake University Donglin Wang
Also publishing here (26)
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.
- Aaron D. Ames Caltech paper
- Alan Fern Oregon State University paper
- Yanwei Fu Fudan University paper
- Baining Guo Microsoft Research paper
- Sehoon Ha Georgia Institute of Technology paper
- Siyuan Huang Beijing Institute for General Artificial Intelligence paper
- Marco Hutter ETH Zurich paper
- Tsung-Yi Lin NVIDIA paper
- Ziwei Liu Nanyang Technological University paper
- Anirudha Majumdar Princeton University paper
- Suraj Nair Physical Intelligence paper
- Jiangmiao Pang Shanghai AI Laboratory paper
- Xue Bin Peng Simon Fraser University and NVIDIA paper
- Ian D. Reid MBZUAI paper
- Guanghui Ren AgiBot paper
- Minjoon Seo KAIST paper
- Guanya Shi Carnegie Mellon University paper
- Shuran Song Stanford University paper
- Koushil Sreenath UC Berkeley paper
- Jie Tan Google DeepMind paper
- Donglin Wang Westlake University paper
- Xiaolong Wang UC San Diego and Meta Superintelligence Labs paper
- Shanghang Zhang Peking University paper
- Zhaoxiang Zhang Institute of Automation, Chinese Academy of Sciences paper
- Hang Zhao Tsinghua University and Galaxea paper
- Yuke Zhu University of Texas at Austin and NVIDIA Research paper
Labs and companies (23)
NASDAQ:GOOGLGemini Robotics 2 (July 2026) controls whole humanoid bodies; its models go into the next Boston Dynamics Atlas.
NASDAQ:NVDAGR00T N1 (March 2025) was the first open humanoid foundation model; N2 previewed at GTC 2026.
NASDAQ:TSLAOptimus: vertically integrated humanoid. Production slipped to summer 2026; 2025 targets missed.
Also active here: Amazon AGI Lab (Hired the Covariant founders in 2024 for warehouse robot learning.); OpenAI (Rebuilt a robotics and hardware group from late 2024; its hardware lead resigned in March 2026 over a Pentagon contract.).
Signature ideas
Vision-language-action models
Take a vision-language model and add an output head that emits robot actions. RT-2 (2023) showed the language pretraining transfers to manipulation.
Cross-embodiment datasets
Pool demonstrations from many different robots so one policy learns from all of them. Open X-Embodiment gathered 22 robot types.
Flow-matching action heads
Generate continuous action chunks with a diffusion or flow model instead of predicting one discrete action at a time. pi0 made it the default.
Sim-to-real (Isaac)
Train in a physics simulator with randomised appearance and dynamics so the policy survives the transfer to a real robot.
Humanoid form factor
Build the robot to fit a world designed for humans. Expensive, but every tool and doorway already fits.
Data flywheels from deployment
Every deployed robot collects data that trains the next policy. Tesla's FSD model, applied to bodies.
Read in this order
- Diffusion Policy: Visuomotor Policy Learning via Action Diffusion 2023
The action head most current robot policies use.
- Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware 2023
ALOHA. Cheap teleoperation made imitation data collection possible at scale.
- RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control 2023
A language model that outputs robot actions; the VLA idea in one paper.
- Open X-Embodiment 2023
Pooling data across 22 robots. The camp's answer to the data problem.
- π0: A Vision-Language-Action Flow Model for General Robot Control 2024
Physical Intelligence's generalist policy; folding laundry from one model.
- GR00T N1: An Open Foundation Model for Generalist Humanoid Robots 2025
NVIDIA's open humanoid model and the synthetic-data pipeline behind it.
Layers that sit on top
Research programs that live inside this camp without disagreeing with it about how intelligence works.
Evaluations and forecasting 7 organizations
Whoever measures the field shapes it. Benchmarks, arenas and forecasting shops decide which claims count, and their numbers are what the money watches. They sit across camps 1 and 5 because that is where the products are.
- METR Task-horizon measurements of autonomous agents
- Epoch AI Compute, data and capability trends
- LMArena Crowd-sourced pairwise model rankings
- Artificial Analysis Independent speed, price and quality benchmarks
- ARC Prize The one benchmark built to resist scaling
- SWE-bench Real GitHub issues as the coding-agent yardstick
- Apollo Research Evaluations for deception and scheming
How money reaches this camp
| Lab | Access | Note |
|---|---|---|
| Unitree | SSE:688836 | Unitree (SSE 688836) since August 2026. |
| Tesla (Optimus) | NASDAQ:TSLA | Tesla (TSLA): Optimus, production starting 2026. |
| Google DeepMind | NASDAQ:GOOGL | Alphabet (GOOGL): Gemini Robotics, and an investor in Apptronik. |
| NVIDIA | NASDAQ:NVDA | Nvidia (NVDA): GR00T, Isaac, and on the cap table of Figure, Skild, PI, Dyna, Agility. |
| Amazon AGI Lab | NASDAQ:AMZN | Amazon (AMZN): Covariant team, Agility customer, Dyna investor. |
| Boston Dynamics | indirect | Hyundai Motor (005380.KS) owns Boston Dynamics and funds the RAI Institute. |
| Agility Robotics | private | SPAC with Churchill Capital XI at $2.5B, closing 2026. |
| Figure | private | $39B. |
| Skild AI | private | $14B; SoftBank-led. |
| Physical Intelligence | private | $5.6B; CapitalG-led. |
| Apptronik | private | $5.5B; Google and Mercedes. |
| Galbot | private | $3B+; Hong Kong IPO planned. |
| AgiBot | private | Hong Kong IPO planned at ~$6B. |
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.