On Tuesday, July 28, 2026, the humanoid robot industry split in two. The FCC added “foreign-produced advanced robotic devices” — humanoids and quadrupeds — to its Covered List, the same regulatory weapon that shut out Huawei, grounded DJI, and blocked Chinese consumer routers. The mechanism is simple and brutal: equipment on the Covered List cannot receive an FCC equipment authorization, and without one, a connected device cannot legally be imported, marketed, or sold in the United States.
No separate import ban. No customs seizure list. Just a wall around the one piece of paper every wireless device needs.
The legal definition is narrower than the headlines: a mobile robot with locomotion, a network connection of at least 200 kbps, environment sensors, autonomous software, and a combined weight over 4.4 pounds including its dock. Road vehicles, drones, surgical robots, and fixed industrial arms are explicitly excluded — this is aimed squarely at walking, rolling, networked machines.
Two things most coverage got wrong. First, the notice names no company and no country — it says “foreign-produced,” defined by the Buy American component test, which means it technically covers Korean and European robots too. Second, it’s new models only. Everything already authorized is grandfathered: the Unitree G1 (FCC-cleared March 2025), R1, H2, Go2, B2, and A2 all remain legal to buy and use, and even keep getting firmware updates through at least January 2029 under a limited waiver. Your existing Unitree is not contraband.
But every new model — every successor, every refresh, every new entrant like BYD’s Xiao Di showroom humanoid unveiled this month — starts on the outside of the wall. UBTech’s Walker S2, notably, has no FCC authorization on file at all.
This didn’t come from vibes. In March 2025, researchers documented CVE-2025-2894 — a hidden service (“CloudSail”) in Unitree’s Go1 firmware that auto-connected to a China-operated cloud tunnel and handed full remote control to anyone with the API key. Vulnerable units were confirmed inside networks at MIT, Princeton, Carnegie Mellon, and Waterloo. Six months later came UniPwn: a wormable Bluetooth exploit giving root access to Unitree’s Go2, B2, G1, and H1 — an infected robot can compromise nearby Unitree robots with no human involvement. Neither has a confirmed public patch.
Layer on the paper trail — Unitree’s self-disclosed PLA research funding, the Pentagon’s Section 1260H Chinese-military-company designation in June, a bipartisan House letter — and the July 27 national-security determination that triggered the FCC was a when, not an if.
China holds roughly 85% of the global humanoid market. Of ~15,000 humanoids shipped in 2025, AgiBot and Unitree each moved more than 5,000 — while Tesla, Figure, and Apptronik shipped a few hundred apiece. The ban doesn’t change that scoreboard; it changes where the game is played. Unitree launched commercially in Europe on July 22 — six days before the ban — and its planned August North American push is dead on arrival. Expect every Chinese maker to follow AgiBot’s Munich playbook: Europe, Southeast Asia, the Gulf — anywhere but America.
Meanwhile, US makers just received the most valuable gift in the industry’s short history: a protected home market. Figure 03, Optimus Gen 3, Apollo 2/3, Digit v5, and NEO no longer have to beat a $16,000 Unitree G1 successor on price — the successor simply won’t be for sale. Whether that protection breeds better robots or complacency is the billion-dollar question; there is a Conditional Approval escape hatch (applications to the FCC, decided by the Department of War, until January 2028), but no Chinese maker has received one yet.
One subtlety worth watching: because the rule says “foreign-produced,” not “Chinese,” it nominally covers allied nations’ robots too — including South Korea’s open-source AI Sapiens K1. If the K-Humanoid Alliance wants the US market, it’ll need Conditional Approvals, US assembly, or a domestic-component audit. The ban’s authors clearly weren’t aiming at Seoul. But the letter of the rule makes “built in America” the only fully safe answer — and in 2026, almost nobody’s humanoid is.
China’s Foreign Ministry called it protectionism that “will only hurt the interests of US companies and consumers.” Both things can be true: the backdoors were documented, and the wall is real. The humanoid race is now run on two separate tracks — and the finish lines are diverging.
Humanoid robotics has its first public-market bellwether. Unitree Robotics listed on Shanghai’s tech-focused STAR Market on Wednesday, August 19, and the debut was a stampede: shares priced at 150.8 yuan (about $22), spiked as much as 629% intraday to 1,100 yuan, and closed at 845 yuan — up 460% on day one, valuing the Hangzhou company at roughly $50 billion.
Demand was absurd even by STAR Market standards. Retail investors oversubscribed the roughly $900 million offering more than 8,000 times — the odds of getting an allocation were about 0.018%. Unitree sold 40.45 million shares, 10% of its enlarged capital, and plans to pour the proceeds into robot AI models, new hardware, and a smart manufacturing base.
Unitree is the first humanoid robot maker to list on a mainland Chinese exchange (UBTECH went public in Hong Kong back in 2023), and it did something almost no robotics startup manages: it showed up profitable. Revenue hit 1.7 billion yuan (~$250 million) in 2025, up 335% year over year, with gross margins near 60% and a net profit of about 278 million yuan. The company shipped roughly 5,500 humanoids in 2025 — more than any other manufacturer on Earth — and together with AgiBot accounted for over 70% of global humanoid shipments last year, per Omdia.
It also hands the industry its first real valuation yardstick. Figure 03’s parent Figure AI carries a reported $39 billion private mark that has never been tested by actual trading. Unitree’s $50 billion is a public number with a tape behind it. Every humanoid pitch deck from San Jose to Shenzhen just got repriced.
Two days before the bell, Unitree unveiled a new humanoid it calls Superman — a handless locomotion demonstrator that does a 2-meter standing vertical jump and hits a claimed 12.66 m/s (about 28 mph) top speed, faster than Usain Bolt’s peak. Unitree says the machine went from blank page to record claims in just over three months. It can’t pick up a box; it exists to prove that the company’s joint motors and control stack can out-muscle biology. As IPO marketing goes, it beats a roadshow video.
The listing landed weeks after Washington effectively banned imports of new foreign-made robots — the FCC added humanoid and quadruped robots to its Covered List in late July, and the Pentagon put Unitree on its Chinese-military-companies list in June (Unitree insists its robots are civilian-only). The US was about 13% of revenue last year, so the ban stings but doesn’t kneecap. Meanwhile the debut coincided with the opening of the World Robot Conference in Beijing, where hundreds of Chinese firms are showing hardware — Chinese vendors shipped more than 97% of the world’s humanoids in the first half of 2026.
This is the first of a wave. Leju Robotics and DEEP Robotics have IPO applications accepted in Shenzhen and Shanghai, AgiBot is going public via a backdoor stake in a listed company, and Chery is reportedly planning to list its robotics business separately. Whether Unitree’s stock holds these levels will decide how friendly that window stays — and whether Optimus Gen 3-era promises from Tesla ever face the same public-market test.
Pop or no pop from here, the floor facts are real: Unitree is profitable, vertically integrated (90%+ domestic components), and sells actual robots at listed prices — the $16,000 Unitree G1, the $5,900 R1, the H1 and H2 platforms — to anyone with a credit card outside the US ban. That makes it the index’s most important supplier to watch this year. Track it on the Unitree Robotics supplier page, and watch the AgiBot G1 entry for the next listing in the queue.
Sources: NBC News and Reuters debut-day reporting (Aug 19, 2026); Quartz and Business Times IPO terms (8,000x retail oversubscription, $904M raise); Unitree prospectus figures via SCMP and Global Times; Omdia and Smart Analytics Global shipment data; Interesting Engineering on the Superman reveal.
Two stories landed within weeks of each other this summer, and together they mark the moment the robot workforce stopped being a keynote demo and became a labor-market fact.
In Seoul, Hyundai Motor Group moved to take full ownership of Boston Dynamics, buying out SoftBank’s remaining ~9.65% stake for roughly $325 million after SoftBank exercised a put option from the original 2020 deal. In Seattle, Amazon confirmed its warehouse robot fleet has passed one million units — closing fast on its 1.56 million human employees, with robots now touching about 75% of all deliveries.
The Boston Dynamics price tag implies a valuation around $3.3 billion. That number is wild, because Hyundai’s own logistics arm bought in last year at terms implying $20 billion, HSBC models $26 billion, and Morgan Stanley has estimated a Boston Dynamics IPO — reportedly under consideration for as early as next year — could raise $15 billion. SoftBank needed the exit; Hyundai got the world’s most famous robotics lab for what analysts consider a deep discount.
What Hyundai is buying is a deployment pipeline. The new Electric Atlas — fresh off a “best robot” showing at CES 2026 and a very public ball-delivery gig at the FIFA World Cup — enters Hyundai’s Georgia Metaplant in 2028 for parts sequencing, moves to component assembly by 2030, and the group is building manufacturing capacity for up to 30,000 humanoids a year. Alongside Atlas come the proven sellers: Spot for inspection and Stretch for warehouse unloading.
Hyundai’s Korean union responded to the Atlas timeline with the auto industry’s first factory stoppage over humanoid robots — partial strikes demanding job-protection guarantees and a fixed-salary structure before machines join the line. It will not be the last. Every legacy manufacturer now has to answer the question Hyundai’s workers just asked out loud.
If Hyundai is the future, Amazon is the present tense. One million warehouse robots — Hercules haulers, Pegasus sorters, the fully autonomous Proteus — now run across 300+ facilities, coordinated by DeepFleet, a generative-AI foundation model that cut fleet travel time 10% on deployment. The newest addition, Vulcan, is the first warehouse robot with a sense of touch, stowing and picking 75% of item types at human-comparable speeds in beta sites in Spokane and Hamburg.
The labor math is no longer subtle. Leaked internal documents reported by The Verge and the New York Times describe plans to avoid roughly 600,000 US hires through automation. Average headcount per Amazon facility is already at a 16-year low, and CEO Andy Jassy told employees directly: “We will need fewer people doing some of the jobs that are being done today.” Even the consumer side is robotic — the Astro home robot keeps finding new tricks.
A comment under the viral clip of the Hyundai news put it bluntly: “this will make building factories in countries with cheap labour unnecessary.” That is the actual headline. A carmaker just bet its manufacturing future on humanoids at home, and the world’s largest robot operator proved the economics already work at million-unit scale. The question for every other manufacturer is no longer if — it is which robot, from which supplier, on what timeline.
Watch three things from here: whether the Boston Dynamics IPO materializes in 2027 (a $15B raise would reprice the entire humanoid sector), whether Hyundai’s union deal becomes the template for human-robot factory contracts, and whether Amazon’s Vulcan beta graduates to general deployment — touch-sensing manipulation is the unlock for the next million robots. The index entries for Electric Atlas and the Boston Dynamics supplier page will track both.
Sources: Korea JoongAng Daily and Boston Dynamics reporting on the SoftBank put-option sale (July 2026); Amazon official newsroom (1M robots, DeepFleet, Vulcan); WSJ robot-vs-employee counts; The Verge/NYT leaked automation documents; Morgan Stanley/HSBC valuation estimates via Korean press.
NEURA 4NE1: Germany’s 100-kg Humanoid Is Open for Business
August 19, 2026 · Biohammer
While American and Chinese humanoids hog the headlines, a German company out of Metzingen has quietly assembled the best-funded robotics play in Europe — and its flagship humanoid is now taking real orders at real prices.
NEURA Robotics’ 4NE1 / 4NE1 Mini line covers both ends of the market: a full-size industrial humanoid and a compact research platform. What makes NEURA different from most of the field isn’t a viral dance video — it’s a €1 billion-plus order book, transparent pricing, and a June 2026 Series C of up to $1.4 billion backed by Tether, Qualcomm, Amazon, NVIDIA, Bosch, and Schaeffler. That is the largest raise ever by a full-stack robotics company.
The Gen 3.5, styled with Studio F.A. Porsche, is a 180 cm, ~80 kg humanoid built for factory floors. Its headline number is payload: 100 kg maximum lift — the highest of any general-purpose humanoid on the market, and four times what the Optimus Gen 3 carries. Dual hot-swappable batteries give 6–8 hours per pack, which means true 24/7 shift coverage if you swap on the fly.
Under the skin: an NVIDIA Thor T5000 running the Isaac GR00T foundation model, seven cameras for 360° perception, force-torque sensing in every joint, and NEURA’s patented artificial skin that detects proximity before contact happens. The whole machine is water-cooled for sustained industrial duty.
Pricing is public — a rarity in this industry. €98,000 for single units, dropping to €60,000 each for fleets of 20 or more, with a fully refundable €100 reservation. First deliveries are expected at the end of 2026.
The Mini shrinks the same cognitive stack into a 132 cm, 36 kg package: 25 DOF, 3 kg payload, about 2.5 hours of runtime, and a €19,999 price tag. It has been shipping since April 2026 and is squarely aimed at university labs and developers who want a Western alternative to budget Chinese platforms. Both variants ship with Python and C++ SDKs plus a ROS 2 interface.
NEURA’s real bet is software. The Neuraverse is a shared-intelligence platform: when one NEURA robot learns a task, the skill propagates to every connected robot. The company is also rolling out “NEURA Gyms” — real-world training environments for cognitive robots — and targets serial production in the millions of units by 2030.
The competitive frame the figure shows is telling — pardon the pun. Against the current field:
NEURA is doing three things almost nobody else in humanoids does: publishing prices, taking refundable reservations, and naming its industrial partners. The 4NE1 Gen 3.5 is not the cheapest way to put a humanoid on your floor — it is currently the strongest one, and the only European option with a billion-dollar order book behind it. If deliveries land on schedule at the end of 2026, the humanoid race stops being a two-country contest.
Specs and pricing: NEURA Robotics public product documentation and June 2026 Series C announcement. See the full index entry for 4NE1 / 4NE1 Mini for links and updates.
← All News & Analysis
© 2026 HCHQ.COM — Humanoid Compute & Hardware HQ
Links may use affiliate tags. All trademarks belong to their respective owners.