Robot brains are getting smarter, but they still can't do a real day's work

Key points

• Unitree, China's leading humanoid robot maker, lost nearly half its market value after debuting at a $66 billion valuation.
• The Actuate developer conference in 2025 drew 1,500 attendees, triple its 2023 size, signalling fast-growing industry interest.
• Genesis AI raised a $105 million seed round in 2025 to build humanoid robots with hardware and AI designed together from the start.
• Foxglove launched a new data tool in 2025 that lets engineers search dense robot sensor data using plain spoken-language queries.
• Industry leaders disagree on whether a single breakthrough moment for robotics is possible, or what shape it might take.

Robots are getting better at moving. What they cannot do yet is earn their keep.

That gap between physical ability and real-world usefulness landed hard on Unitree this week. China's leading robot maker debuted on China's equivalent of the NASDAQ at a valuation of $66 billion, then shed nearly half that value as analysts pointed out the obvious: impressive movement skills don't automatically translate into tasks that make money.

Why can't robots just get to work?

Data is the short answer. Training an AI model to speak or write requires vast libraries of text. Training it to handle objects, sort parts, or navigate a factory floor requires something much harder to collect: libraries of physical experience.

That shortage, described at last week's Actuate conference in San Francisco as "the robotics data crisis", is the central problem facing the whole industry. Foxglove, the company that organises Actuate and helps robot developers manage training data, said 1,500 engineers attended this year, triple the count from 2023. We first covered Foxglove on 26 August, in the context of Wayve and Uber's robotics push.

Harry Mellsop, founder of simulation startup Antioch, put it plainly: physical AI is in its "GPT-2 era." GPT-2 was an early OpenAI language model that arrived a couple of years before ChatGPT, impressive on paper and largely useless in practice. More data and computing power were needed before things clicked. Mellsop argues robots are at that same awkward stage today.

Who is furthest ahead, and why?

Self-driving cars lead the field. They benefit from two advantages other robots don't have: every human-driven car generates useful training data automatically, and a car's core job is to avoid hitting things rather than pick them up or push them around, which is a simpler physical challenge.

That head start is why companies rooted in autonomous vehicles are now entering humanoid robotics. Tesla is already doing this with its Optimus robot. Wayve and Uber both opened humanoid robotics research labs in 2025.

Alex Kendall, CEO of Wayve, told TechCrunch that manipulation robotics, meaning robots that physically handle objects, is "like self-driving five years ago." His company licenses AI driving models to car makers, a business he sees as a multi-billion dollar opportunity, and plans to use the same infrastructure to build broadly capable embodied AI.

Not everyone agrees with that brain-first approach. Théophile Gervet, CEO of Genesis AI, argues it's too early to build a general AI brain and attach it to whatever robot body is available. His company raised $105 million in seed funding to design hardware and AI together from the start.

"No customer cares about the general purpose robot that works at 80% success rate," Gervet said. Robots tackling narrow, specific jobs are already shipping: Gritt builds solar farms autonomously, Agility deploys robots in industrial settings, and Bedrock operates excavators without a driver.

What would a real breakthrough look like?

Leaders in the field gave three different answers.

For Kendall, it arrives when self-driving works reliably with less than $1,000 worth of sensors in the car, a threshold that makes the technology affordable for ordinary buyers rather than just fleets.

Gervet's answer: the moment you can speak naturally to a robot and have it reliably handle basic tasks like pushing, pulling, or closing a laptop, with a success rate above 80 percent, straight out of the box.

Adrian Macneil, CEO of Foxglove, is sceptical there'll be one clean moment at all. "Distribution in the real world is way harder," he told TechCrunch. He'd settle for an Apple II moment: a home robot you can actually buy, that does something genuinely useful.

The robot data problem isn't going away quietly. The largest robot deployment in the world today is still consumer vacuum cleaners, which tells you everything about how far the field still has to travel.