Servoservobrief.com, Wednesday, October 28, 2026
Capital day for robot learning: two Chinese embodied-AI rounds land while a UK simulation startup posts the country's largest robotics seed.
-
Machenta raises UK's largest robotics seed to train robots in simulation
Update: Forbes adds detail to Machenta's $40M seed, the largest ever for a UK robotics company, led by Imperial College professor Edward Johns and PhD student Vitalis Vosylius. The approach uses in-context learning, a big network trained on simulated data that masters a new task from one human demonstration, shown on packaging, assembly, and tidying across different robots. If simulation alone can teach robots fast, it undercuts the costly real-world data collection most rivals depend on.
Read more on -
China's Zhongke Huiling raises hundreds of millions of yuan for embodied AI
Beijing Zhongke Huiling, founded in 2023 and backed by the Chinese Academy of Sciences' Institute of Automation, raised a Series A extension worth several hundred million yuan, announced October 26. It builds a full-stack system: robot bodies, an embodied brain, and dexterous hands, aimed at industrial manufacturing. The round adds to a steady run of Chinese state-linked capital into vertically integrated humanoid makers.
Read more on -
DEEP Robotics raises over $70M for quadrupeds and humanoids
Chinese firm DEEP Robotics raised more than $70M to scale its quadruped (four-legged) and humanoid systems, per a trade roundup. No breakdown, lead investors, or valuation disclosed, and the figure is vendor-reported. It is another sign of capital concentrating in China's legged-robot makers as they push from four legs toward two.
Read more on -
GXO extends Digit to two more warehouses after 100,000-tote milestone
Update: Techtimes adds operational numbers to the GXO expansion into North Carolina and Tennessee, slated for early 2027. At the original Georgia site, Digit moved 100,000 totes at roughly 98 percent on-task accuracy, and Digit 5 claims a 10-to-1 run-to-charge ratio, enough uptime to avoid interrupting shifts. The published dataset is the largest public evidence that humanoids can do real warehouse work, ahead of Agility's planned November Nasdaq listing.
Read more on -
Boston Dynamics argues humanoid hands should not copy human hands
Boston Dynamics published an engineering argument that cloning the human hand is the wrong default for a working humanoid, prioritizing durability and task fit over anatomical fidelity. No product or numbers disclosed. It lands against a live industry debate, echoed by John Carmack, that hand dexterity is the main unsolved roadblock in humanoids.
Read more on