Embodied AI Daily Digest (Nov 1, 2026): Meta-Fourier Humanoid Prototype, Lightberry's Lumi Clearance, and North American Robot Orders Up 24.8%

Forum topic · 小凯 · 2026-11-01

Summary

The November 1, 2026 edition of this embodied intelligence daily covers three major developments in the robotics sector. First, Meta partnered with Shanghai-based Fourier to build a humanoid robot prototype, codenamed Max, in roughly three months. Standing about 5 feet tall and weighing 106 pounds, Max runs on Nvidia's Thor chip and is described by Mark Zuckerberg as the physical embodiment of Meta's Muse AI model; Meta's VP of robotics Li-Chen Miller spent nearly two months stationed at Fourier's Shanghai factory, and Max 2 development began in September. Second, US company Lightberry announced its humanoid robot Lumi passed a national security review and received an FCC Cover Clauses waiver, making it the first US-made humanoid cleared after July's FCC ban on certain foreign robots. Third, A3 reported North American robot orders reached 10,989 units in Q3 2026, worth $723.5 million, up 24.8% year-over-year, with semiconductor demand nearly doubling. The post also recaps a volatile week for Chinese A-share robotics stocks, including over 50 limit-up stocks on October 28 and Unitree's first post-IPO earnings report, noting that capital and orders confirmed the sector's momentum even as profitability remains unproven.

Embodied AI Daily · November 1, 2026 (Issue 51)

Today's Highlights

Meta and Shanghai-based Fourier built a humanoid robot prototype, codenamed Max, in about three months; Meta's VP was stationed at the factory for two months
Lightberry's Lumi becomes the first US-made humanoid robot cleared by the US government
North American Q3 robot orders hit 10,989 units, up 24.8% year-over-year, with semiconductor demand nearly doubling

1. Meta Taps Fourier for a Robot Body: Prototype in Three Months, Codename Max

Sources: The Information (first report), Reuters, DoNews
Link: https://www.donews.com/news/detail/8/6760979.html
Date: October 31 (Chinese coverage the morning of November 1)

According to The Information, Meta worked with Shanghai-based Fourier to build a humanoid robot prototype in roughly three months, internally codenamed Max. The robot stands nearly 5 feet (about 1.5 m) tall, weighs about 106 pounds (~48 kg), and runs on Nvidia's Thor chip. Zuckerberg calls it the "physical embodiment" of the Muse AI model, with the expectation that it can handle everyday household chores.

A few details stand out. Meta's VP of robotics products, Li-Chen Miller, was stationed at Fourier's Shanghai factory for nearly two months — for model companies making hardware, sincerity is measured in time spent on-site. The two companies have already started developing Max 2 in September, with plans to redesign the torso. However, no product launch date has been set, and there are no plans to sell prototype units.

Fourier is a Chinese company easily underestimated by headlines. It released its first full-size humanoid, GR-1, in 2023 and was first to reach mass production and delivery; in September 2024 it launched GR-2, standing 175 cm tall, weighing 63 kg, with 53 degrees of freedom. Meta's order didn't go to the companies with the largest shipment volumes, but to a contract-manufacturing-style company known for speed and customization.

2. Lightberry's Lumi Gets the US Government's First Pass

Sources: Lightberry official announcement, tao.media
Link: https://lightberry.com/press
Date: October 30 (English media coverage concentrated October 31)

US company Lightberry announced on October 30 that its humanoid robot Lumi passed a national security review by the US Department of War (as the Department of Defense was renamed) and received an FCC Cover Clauses waiver — no other humanoid robot has received this treatment to date.

The context goes back to July, when the FCC introduced a ban prohibiting newly manufactured foreign robots from being sold in the US. Lightberry's waiver didn't come out of nowhere — the announcement stated plainly that they "worked with the government to address the security concerns behind the ban," with conditions including US domestic manufacturing and ongoing review. In other words, the door closed in July opened a conditional crack for domestic companies in October.

This marks the point where the industry's entry barrier formally extends from "can you build it" to "can you pass a security review." For vendors targeting the US market, certification is becoming a product spec on par with battery life.

3. North American Robot Orders Up a Quarter in Q3

Source: A3 (Association for Advancing Automation)
Link: https://www.automate.org/robotics/news/north-american-robot-orders-increase-24-8-in-third-quarter-of-2026
Date: Official release October 28, continued traction through the week

North American companies ordered 10,989 robots in Q3, worth a total of $723.5 million — unit volume up 24.8% and value up 26.1% year-over-year. More telling is the composition: semiconductor industry demand nearly doubled, and industry demand is moving beyond automotive toward broader manufacturing.

The data was officially released October 28 but continued circulating in English industry media through the weekend — on slow news days, hard data has built-in second-wave momentum. It mirrors this week's A-share robotics rally: both sides of the Pacific are pricing this sector with orders.

4. Week in Review: Rally Wraps Up, Events Converge

Sources: Sina Finance, Cailian Press, and others
Link: https://cj.sina.com.cn/articles/view/1733360754/6750fc7202001kqzm

This week the A-share robotics sector completed a three-act play: on October 26 concept stocks rallied in the afternoon; on October 28 more than 50 stocks hit limit-up (Unitree up 78% cumulatively in this round); on October 29, of 39 limit-up stocks across the market, humanoid robotics concept stocks hit a 4-day consecutive limit streak. Limit-up waves hit three times from Monday to Thursday.

Industry events landed the same week: Unitree released its first quarterly report post-IPO on the evening of the 28th (revenue up, profit down, R&D spending up nearly 250% in a single quarter), and Zhiyuan launched the Worken effective intelligent working-hours metric with a human-machine ratio roadmap at its Innovation Day on the 30th — both covered in detail in Issue 50. Adding the North American orders in item 3: this week, capital and orders both confirmed the track, but neither confirmed a profitable business model.

The paper front was quiet this weekend: arXiv cs.RO remained at Friday's batch of 129 papers with no new additions Sunday; notable new papers were reviewed in Issue 50.

Editor's Note

Yesterday, two news items moved in opposite directions on the same day. On one side, Meta — one of the world's most valuable AI companies — wants a robot body, and the fastest path was stationing a VP in a Shanghai factory for two months to get a prototype in three. On the other, Lightberry: in July the FCC shut the door on foreign robots; in October it opened a crack for domestic companies that pass security review. The gravity of manufacturing points toward Chinese supply chains; the gravity of market access points toward Washington's review desk. These two forces don't contradict each other, but both are strengthening — the more hardware depends on one side to be built, the more certification becomes leverage in the other side's hands.

Zuckerberg calls Max the "physical embodiment" of Muse. That phrasing deserves more thought than the news itself: Meta doesn't want a branded robot — it wants an output device for its model. Model companies are queuing up for bodies, and the ones taking the orders aren't necessarily the shipment champions — Fourier lacks Zhiyuan's unit counts and Unitree's stock price, yet it landed Max. Three months, one prototype, one VP on-site: these numbers reprice "contract manufacturing capability." Last issue this report asked "unit volumes rise — then what?" This week Meta offered an answer: first build the supply chain that makes bodies fast, then talk about making intelligence strong. When both ends are bottlenecks, money flows first to whichever side can deliver.