The cable threaded through a robot's arm should be invisible to everyone but an engineer. Yet in a Shenzhen workshop, that tendon-like rope is the reason one startup says its machine can miss a grab and still not shatter a glass or bruise a child standing nearby.
Key takeaways
- Stardust Intelligence (星尘智能), which markets its machines under the Astribot brand, claims to be the first company in the world to mass-produce "cable-driven" (绳驱) AI robots.
- On November 18, 2025 it closed a nine-figure RMB A++ round co-led by Guoke Investment (国科投资) and Ant Group (蚂蚁集团) — Ant's third lead in the company within a single year.
- The technical bet is a different body plan: replace rigid metal gears with rope-and-pulley "muscles" that flex on impact instead of fighting it.
- The company reports thousands of units in orders spanning research labs, cultural tourism, retail and industrial logistics.
Why a rope, not a gear
Most humanoid (人形机器人) arms move with rigid actuators and metal gears that hold a position with brute force. Stardust argues that design is the wrong starting point for a machine meant to share a room with people. Its cable-driven (绳驱) transmission mimics a human tendon: low friction, high follow-through, and a natural elastic buffer. The company says that means the robot "does not hurt others, itself, or the environment," because when a grasp goes wrong the arm gives instead of colliding.
That is a genuine engineering trade. Rigid gears are precise and easy to model. Rope systems are harder to control but safer in contact — exactly the property a service or home robot needs when it fumbles a cup. Stardust frames the whole philosophy as "Design for AI": build a body whose force data is clean and continuous enough that a learning model can actually read it.
The money: Ant's third bet
The November 18, 2025 round was described by the company as 数亿元 — several hundred million yuan, though the exact amount was not disclosed. At prevailing exchange rates that band implies roughly US$30M–130M (HK$230M–1.0B), but the figure should be read as indicative, not audited. It was co-led by Guoke Investment (国科投资) and Ant Group (蚂蚁集团), with Bloom Advance Capital, Times Beat (时代伯乐), Nanshan Strategic Emerging Investment (南山战新投) and returning shareholder Jinqiu Fund (锦秋基金) participating. Huaxing Capital (华兴资本) acted as exclusive financial adviser.
What makes the round notable is the pattern, not just the size. Ant Group led Stardust's A and A+ rounds in April 2025 and then this A++ round — three leads inside twelve months. For a private investor, that is a signal of conviction in a specific technical path rather than a spread of bets.
What they are actually shipping
The company says it cracked cable-driven mass production in June 2025 and began stable monthly batch deliveries. At the IROS robotics conference in October 2025 it showed an upgraded stack: a new half-body commercial robot, the Astribot S1-U; an ultra-long-distance "digital twin" teleoperation system validated across provinces and internationally; and an updated DuoCore full-body VLA (视觉—语言—动作, vision-language-action) model aimed at service scenarios.
Commercialization is the harder test. Stardust says it has taken thousands of units in orders and points to names across research, cultural tourism and entertainment, commercial service, and industrial logistics. On November 15, 2025 it struck a strategic partnership with Jinma Entertainment (金马游乐, Shenzhen-listed 300756) to build a line of cultural-tourism robots, including a retail store concept called "Robot MART" (机器人MART). It also lists ByteDance, Tencent, Baidu, CCTV and JD as customers or partners pushing the robots into real venues.
The honest case for cable-driven
Stardust is not alone in chasing safe, compliant bodies, but it is unusual in betting the company on rope instead of gears. If the control problem is solved, the safety and cost story is compelling: fewer precision machined parts, gentler contact, and force data a model can learn from. If it is not solved, the robots stay demo pieces. The company also signed with SEER (仙工智能), described as the world's largest controller-system vendor, in September 2025 — a supply-chain anchor that matters more than any single demo.
Honest limitations
Most of the hard claims here come from the company's own announcements, relayed through Chinese securities press (Shanghai Securities News, China Securities Journal, Securities Times). Independent, audited order and delivery figures are not public. "World's first mass-produced cable-driven AI robot" is a company characterization; no third party has rigorously benchmarked the claim against every competitor. Durability of rope transmissions under years of daily use is unproven. The funding band (数亿元) is the company's wording, not a disclosed number, so the US$/HK$ conversion is a range, not a point estimate. I have not included any projection of future revenue or valuation.
What readers can do now
- If you track embodied intelligence (具身智能), watch whether "rigid vs. cable-driven" becomes a real split or a niche — Stardust's batch deliveries and Ant's repeat backing are the signals to monitor.
- For robotics buyers evaluating service or home units, ask vendors directly about force-control and contact-safety design, not just payload and battery life; cable-driven is one answer to that question.
- For investors, treat the order and "world-first" claims as company-stated until an audited filing or independent teardown appears; the repeat lead by one strategic investor is a more useful diligence clue than the headline round size.