The decline in the Unitree share price after the IPO reveals the early stage of physical AI
Shares in Unitree, valued at 66 billion dollars at the IPO, lost roughly half their value. According to industry experts, physical AI is still at an early stage comparable to the GPT-2 era, lacking high-quality training data and reliable commercial applications for general-purpose robots.
Shares in Unitree, a leading Chinese robot manufacturer, were valued at 66 billion dollars when the company listed on the Chinese equivalent of NASDAQ. This week, however, their price fell by roughly half. According to analysts, the main reason is that although the physical capabilities of robots are improving, robots still cannot perform work that creates value.
At the Actuate conference, dedicated to developing AI “brains” for robots, which had 1500 attendees this year (attendance has tripled since 2023), the industry was said to be facing a so-called data crisis—a shortage of high-quality training data for general-purpose robots—according to the organizer, Foxglove. Harry Mellsop, founder of the simulation company Antioch, compared physical AI to the “GPT-2 era,” referring to the model from OpenAI that preceded the arrival of ChatGPT. In his view, more data and computing power will be needed, especially GPUs optimized for the ray tracing used to create realistic simulations.
According to the article, autonomous vehicles are the furthest advanced in the field because they allow relevant data to be collected from human-driven journeys, and their main task is to avoid contact rather than manipulate their surroundings. Automakers are therefore now betting that investing in machine learning tools will enable them to compete with specialized manufacturers of humanoid robots—Tesla is trying this with the Optimus robot, while Wayve (focused on autonomous driving) and Uber have both recently launched robotics laboratories focused on the humanoid form as research initiatives.
Wayve CEO Alex Kendall considers it premature to bet on a specific hardware platform, while Genesis AI CEO Théophile Gervet argues that it is better at this point to design hardware and AI together. According to Gervet, companies focused on a specific task—such as Gritt (solar farms), Agility (industrial operations), or Bedrock (autonomous excavators)—are succeeding in deploying robots in practice, while general-purpose humanoid robots remain in laboratories.
Why it matters
The decline in the Unitree share price shows that the market may be reassessing how close physical AI is to commercially viable deployment. For companies and investors in the robotics sector, this means that, according to the sources, narrowly focused projects (construction, industry, solar energy) are reaching real-world deployment faster than bets on general-purpose humanoid robots, which lack sufficient high-quality training data and reliable commercial performance.
Two audiences, two different impacts
What this means
For individuals
For professionals following robotics and physical AI, the sources suggest that narrowly focused deployments (construction, industry, autonomous vehicles) are closer to practical use than general-purpose humanoid robots, which still remain mainly in laboratories.
More practical updates →For a business
According to the sources, companies considering investments in robotics should factor in the possibility that the market may have overestimated how close general-purpose humanoid robots are to commercial deployment, while narrowly focused vertical projects are already delivering real-world deployments and operational data.
StrategyCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.