Industry Brief, September 6, 2026: China's quiet robot build-out, humanoids, sensing

By MillBrief Editorial · September 6, 2026 · Retrospective edition, published September 23, 2026

This is a retrospective edition. The brief did not run on September 6, 2026; we wrote it on September 23, 2026 using only reporting that was published on or shortly before September 6, 2026, and it makes no forecasts.

This is a retrospective edition covering September 6, 2026, using only reporting published on or up to two days before that date. A Sunday edition with three items that pull in slightly different directions: a columnist arguing that China’s real robot story is specialised industrial machines, a humanoid maker announcing a customer count, and a large funding round for the sensors robots use to see. Our own words, links to the original outlet, vendor numbers labelled as vendor numbers.

An FT columnist: China’s robot revolution is industrial arms and AMRs, not humanoids

In a Financial Times commentary republished by CNA, Zijing Wu argues that while humanoid games and Unitree’s listing drew the attention, China’s larger robot shift is happening quietly in industry. Wu cites International Federation of Robotics figures: 295,000 new industrial robots installed in China in 2024, 54% of the world total. Citing China’s industry ministry, Wu puts the country’s robotics industry revenue at more than US$44 billion last year, with average annual growth above 20% over five years, and writes that industrial robot exports grew 49% in 2025, the first year China was a net exporter of industrial robots. Two years ago, the column says, China already operated more than 2 million industrial robots, about 4.5 times as many as second-placed Japan.

The column adds that Chinese makers now hold the majority of their home market for industrial robots, which the “Big Four” (Fanuc, ABB, Yaskawa and KUKA) dominated before 2024, and that Guangdong province has approved 48 “Robot+” deployment scenarios. The examples are purpose-built: thousands of robot arms on EV lines at BYD’s Xi’an hub, and an automated terminal at Shanghai’s Yangshan Port run on automated guided vehicles and remote-controlled gantry cranes. Wu’s view on humanoids is that environmental fit and learning from human video are serious but distant advantages, and that for most industrial tasks specialised machines are the better answer for now.

Why it matters: This is opinion, but the numbers behind it describe your supplier landscape. More Chinese-built industrial arms and AMRs in export markets can mean more choice and price pressure on quotes, alongside questions about service depth, spare parts and trade rules. It is also a useful counterweight to humanoid headlines: most of the automation a small plant can buy and justify today is the purpose-built kind covered in what to automate first and cobot vs industrial robot.

Source: CNA, republishing the Financial Times, September 4, 2026.

Wandercraft says 12 customers have signed for its Calvin-40 humanoid

In a press release on GlobeNewswire, Wandercraft, which spent more than a decade building self-balancing medical exoskeletons, says it has secured 12 customers for Calvin-40, its industrial humanoid, including automotive, logistics, retail, healthcare and defense organisations. It does not name them. The company says the robot is being developed to handle boxes, parcels, tires, panels and carts, and that picking and sorting are being developed and tested as next steps.

Its stated strategy is to prove reliability on defined task clusters first and widen from there; CEO Matthieu Masselin puts it as humanoids becoming general as they scale, not scaling because they are general. The release also points to its manufacturing partnership with Renault Group and says Renault has announced plans to deploy 350 humanoid robots over the next 18 months.

Why it matters: Everything here is the company’s own account: no customer names, no deployment counts, no uptime or success-rate figures. The task-first approach is the sensible one, and it is the same question to put to any humanoid seller: which specific task, at what cycle time and success rate, measured where. Until those numbers are published, budget a humanoid pilot as a pilot, and read our engineers’ notes on what they wish buyers knew.

Source: Wandercraft press release, via GlobeNewswire, September 4, 2026.

Lyte raises $165M to put its own sensing chips into robots

The Robot Report reports that Lyte, a Sunnyvale, Calif., company founded in 2021, has raised a $165 million Series C at a $1.6 billion post-money valuation, led by Maverick Silicon. Lyte designs its own processors, multimodal sensors and software; its LyteVision platform fuses 4D coherent vision, high-resolution imaging and inertial sensing on one synchronised timeline. The company says it is in production and shipping to robotics customers in inspection, logistics and manufacturing.

CEO Alexander Shpunt argues that physical AI has a sensing problem before it has a model problem. Lyte says it has now raised $272 million in total, which squares with the $107 million it announced on leaving stealth in January plus this round ($107M + $165M = $272M). The money is earmarked for scaling production, perception capability and deployments.

Why it matters: Shpunt’s own line, that a robot cannot act safely on data that does not faithfully describe the world, applies to every picking or mobile robot you might buy, which makes sensing investment more relevant to buyers than another humanoid demo. You are unlikely to buy Lyte directly; its chips would arrive inside someone else’s robot. That makes one question worth adding to your integrator checklist: whose sensors and perception software sit in the system, and who supports them if the sensor maker changes course.

Source: The Robot Report, September 4, 2026.