The week's biggest story? Reality checks. From BMW proving humanoids can actually work production lines to a legendary roboticist declaring the whole endeavor "fantasy," this week forced the industry to confront the gap between demo videos and deployable technology.
BMW Brings Figure Robots to Europe After U.S. Success
The headline numbers are hard to ignore: 30,000 cars produced, 90,000+ parts handled, 1,250+ hours of runtime. That's what Figure AI's humanoid robots achieved at BMW's Spartanburg plant in South Carolina over 11 months.
Now BMW is taking the experiment to Europe. On February 27, the automaker announced it will deploy humanoid robots at its Leipzig plant in Germany—the first time "Physical AI" of this kind has entered European automotive production.
Why it matters: This isn't a demo. Figure 02 ran 10-hour shifts, Monday through Friday, on an active assembly line. The robots loaded sheet-metal parts with 5-millimeter precision in just 2 seconds per cycle. When you're building X3 SUVs, that kind of consistency matters.
The deployment also generated critical data that shaped Figure 03's design. The forearm—the robot's most failure-prone component—was completely re-architected for the new model. Every intervention, every failure mode, every hour of runtime informed the next generation.
Our take: This is the deployment milestone the industry needed. Flashy videos of robots folding laundry are one thing; running an automotive production line for nearly a year is another. BMW's expansion to Germany signals that the pilot exceeded expectations. For more on Figure's latest, see our Figure 03 review and Figure AI company analysis.
iRobot Cofounder: Humanoid Vision Is "Pure Fantasy"
Rodney Brooks, the MIT roboticist who cofounded iRobot (makers of the Roomba), unloaded on the humanoid robot industry this week. His verdict on Elon Musk's vision of humanoid assistants: "pure fantasy thinking."
Brooks argues that today's humanoid robots "will not learn how to be dexterous" regardless of how many billions VCs pour into training. The problem? Touch.
Human hands contain 17,000 mechanoreceptors for detecting pressure and texture. While AI has been trained on massive datasets of speech and images, "we do not have such a tradition for touch data," Brooks wrote. Training robots by filming humans performing tasks—the approach used by Tesla and Figure—won't solve this fundamental gap.
Why it matters: Brooks isn't some armchair critic. He's been building robots for three decades. His claim that robots won't look like humans in 15 years—instead sporting wheels, multiple arms, and only being called humanoids—directly challenges the form factor every major player has bet on.
Our take: Brooks has a point about the touch data problem, but dismissing the entire humanoid effort feels premature. BMW's deployment shows real-world value exists even with current limitations. The question isn't whether today's robots are perfect—it's whether they're useful enough to justify continued investment. Still, his critique about transparency is valid. When companies hide their teleoperation rates, the public can't properly evaluate progress. For context on what these robots actually cost, check our humanoid robot pricing guide.
China Owns 90% of the Humanoid Robot Market
The numbers are stark: Chinese companies shipped roughly 90% of all humanoid robots sold globally in 2025. Unitree moved 5,500 units. Agibot shipped 5,168. Meanwhile, Figure, Agility Robotics, and Tesla each sold around 150.
That's not a typo. Unitree shipped 36 times more humanoid robots than its closest American competitor.
"China has a more robust hardware supply chain—much of it built up through the EV sector, from sensors to batteries—and the world's strongest manufacturing base," analyst Selina Xu told TechCrunch.
Even Elon Musk acknowledged the competitive reality at Davos: "China is very good at AI, very good at manufacturing, and will definitely be the toughest competition for Tesla. To the best of our knowledge, we don't see any significant competitors outside of China."
Why it matters: This is the EV playbook all over again. Early state support, industrial policy, rapid iteration, cost advantages—and before Western competitors could scale, Chinese companies owned the market. Global humanoid shipments were just 13,317 units last year. By 2035, that's projected to reach 2.6 million. The early leader often becomes the permanent leader.
Our take: The U.S. still leads in AI and software. Figure's deployment at BMW demonstrates capabilities Chinese competitors haven't matched publicly. But hardware matters, and China's supply chain advantages are formidable. For the latest on Chinese robots, see our Unitree G1 review, Unitree H2 review, and analysis of the best humanoid robots on the market.
MIT Technology Review: The Hidden Human Labor Behind Humanoids
A worker in Shanghai recently spent a week wearing a VR headset and exoskeleton while opening and closing a microwave door hundreds of times a day—training the robot beside him. Welcome to the strange new world of humanoid robot training.
MIT Technology Review published a deep investigation into the human labor powering the "autonomous" humanoid industry. Key revelations:
- Data harvesting at scale: Figure partnered with Brookfield, which manages 100,000 residential units, to capture "massive amounts" of real-world data in homes.
- Delivery workers as sensors: One company had workers wear movement-tracking sensors while moving boxes, generating training data.
- Teleoperation everywhere: 1X's Neo robot, shipping to homes this year, relies on remote operators to handle tricky tasks. If the robot gets stuck, a person in Palo Alto takes over.
"Just as our words became training data for large language models, our movements are now poised to follow the same path," the report notes.
Why it matters: If home humanoids aren't genuinely autonomous, the business model is "a form of wage arbitrage that re-creates the dynamics of gig work while, for the first time, allowing physical tasks to be performed wherever labor is cheapest."
Our take: Transparency matters. When companies hide their teleoperation rates, the gap between marketing and reality becomes dangerous—literally, as Tesla's Autopilot lawsuits demonstrate. We're not saying teleoperation is bad; 1X gets customer consent. But the industry needs honest communication about what these machines can actually do today. For more on what you can actually buy, see where to buy humanoid robots.
Honor Enters the Humanoid Race
Chinese smartphone giant Honor will unveil its first humanoid robot at Mobile World Congress 2026 in Barcelona. The announcement, made via teaser video on X, marks another major consumer electronics company entering the humanoid space.
Honor joins Xiaomi, which launched CyberOne in 2022, in the phone-maker-to-robot-maker pipeline.
Why it matters: When smartphone companies start building humanoids, it signals the technology is approaching a commercial tipping point. These companies have massive manufacturing capabilities, consumer distribution networks, and experience shipping millions of complex hardware units annually.
Our take: The consumer electronics crossover validates the humanoid form factor for home applications. Honor's robot remains mysterious for now, but we'll be watching MWC closely. For more on the smartphone-to-robot trend, see our Xiaomi CyberOne review.
Market Pulse
| Metric | Data |
|---|---|
| Total 2025 humanoid shipments | ~13,300-18,000 units globally |
| China's market share | ~90% of global shipments |
| Top sellers | Unitree (~5,500), Agibot (~5,168), UBTech, Leju, Engine AI, Fourier |
| Projected 2035 market | 2.6 million units; $38 billion |
| Figure 02 status | Fleet-wide retirement beginning |
What to Watch Next Week
- Mobile World Congress (March 1-4): Honor's humanoid reveal, plus any robotics announcements from other consumer electronics players.
- Figure 03 deployment updates: With Figure 02 retiring, watch for announcements about where the next-generation robots will be deployed.
- China policy signals: The "Made in China 2025" robotics push is accelerating. Watch for new funding announcements or pilot programs targeting the 2.6 million unit 2035 goal.
Stay updated with the latest humanoid robot news by visiting our humanoid robot news hub. Ready to buy a humanoid robot? Check out our marketplace.
