Unitree Can Build the Body, Can It Build the Mind?

Best-in-class hardware, unproven AI

Unitree robots in a feisty boxing match.

Hi everyone,

We’re actually ramping up our publishing schedule, but this is a surprise update anyways. We’ve been quietly building out our robotics coverage and wanted to share a few things.

First, we’re launching a new site to track the humanoid robotics landscape. It’s live in alpha at robotics.techbuzzchina.com. We’re still uploading and scrubbing data, so it’s rough around the edges, but bookmark it now and check back often.

Second, the most buzzy humanoid robotics company in the world behind Tesla Optimus just filed for its IPO, so we wanted to send you a quick note on our view. Our take on the Unitree STAR Market filing is below.

And third, we’re leading two trips to China in April with plenty of robotics exposure. One of our stops is Liuzhou, currently the largest humanoid manufacturing base in China, where humanoids are already working on EV factory floors. We’d love to have you join us on future trips. Get a group together or join one of our open enrollment trips.

Unitree Files for IPO: The Hardware Is Real, the Intelligence Isn’t

Unitree Robotics, the robotics company whose humanoids went viral performing martial arts at the Spring Festival Gala, filed to list on Shanghai’s STAR Market on March 20, 2026. Unitree actually built its reputation on quadruped robots, where it still holds dominant market share, and quadrupeds still account for 42% of sales. But it is the humanoid business that has driven the growth story. The numbers behind the filing are striking: revenue surged from 159 million RMB (US$23M) in 2023 to 1.708 billion RMB (US$248M) in 2025, a 335% increase. The company swung from a small loss to an adjusted net profit of 600 million RMB (US$87M), with gross margins around 60%. The IPO seeks to raise 4.2 billion RMB (US$610M) at an implied valuation near 42 billion RMB (US$6.1B). Retail investor excitement has been palpable.

Translated from the Unitree prospectus.

Unitree sold over 5,500 humanoids in 2025, claiming roughly one-third of global shipments. State media declared embodied intelligence a national strategic priority alongside quantum computing and brain-computer interfaces. But a careful reading of the prospectus reveals a company whose real strength is motion control and vertical integration, and whose ultimate challenge, developing autonomous intelligence for meaningful labor substitution, remains unresolved.

The Business: Three Segments, Not Much AI Required

Unitree’s revenue comes from three segments, none of which require much in the way of advanced AI.

Performance robots showcase motion control through scripted or teleoperated sequences. The Spring Festival show was choreography, not autonomy.

Research and developer platforms drive most revenue. The G1, priced at about 99,000 RMB (US$13,500), is an open platform where customers buy hardware and develop their own AI. Think developer kits: premium prices, minimal integration costs, high margins. The top five customers represent just 10.6% of revenue, and overseas sales exceed 35%.

Industrial pilots remain limited. Tour-guide and reception applications account for 50 to 70% of industry revenue. Many demos are teleoperated. Real factories require 95 to 99% uptime; current humanoids manage about 90 minutes.

This is why Unitree’s current business works so well. Developer platforms and performances play to its strengths and avoid its weaknesses. The prospectus allocates 2.022 billion RMB (US$294M), 48% of proceeds, to embodied AI R&D, an acknowledgment that hardware alone won’t sustain leadership.

Why Margins Are So High

Unitree achieves 60% gross margins, nearly double oft-cited rival UBTech’s ~33%, partly through relentless vertical integration. The company makes almost all core components in-house (motors, reducers, sensors, proprietary 4D LiDAR), cutting costs by roughly 30%. Notably, margins have actually expanded as Unitree shifted toward humanoids: gross margins were just 44% in 2023 when the business was mostly quadrupeds. The humanoid pivot has been margin-accretive, though current levels may not be sustainable as competition intensifies.

What looks impressive on stage is not what works on a factory floor. When we visited Kepler, another humanoid maker, in Shanghai last fall, this was clear: robots designed for industrial reliability are built differently from the ground up. The agility that wows in a martial arts demo can work against you when what you need is repetitive, stable task execution for hours on end. A factory worker and a Douyin performer need different things from a robot.

The margin gap with UBTech is mostly a business model story, not an execution story. UBTech sells turnkey solutions requiring integration, training, and after-sales service. Unitree offloads all of that onto customers. If Unitree enters industrial markets with full support, expect that gap to narrow.

For context: UBTech generated 621 million RMB (US$90M) in revenue for the first half of 2025, with an adjusted net loss of 367 million RMB (US$53M). UBTech’s Walker S2 sells for US$240,000+, nearly 18 times the G1. These are adjacent markets, not head-to-head competition.

The Autonomy Gap

Unitree’s robot videos frequently feature teleoperation rather than true autonomy, a distinction Chinese media have noted and the company’s own GitHub repositories confirm. University research labs, not Unitree, have produced many of the most impressive AI demonstrations on its hardware. The prospectus acknowledges that intelligence remains largely customer-supplied.

Industrial customers will not accept teleoperation. The prospectus candidly states that industrial revenue “mainly came from enterprise reception and tour-guide use,” semi-scripted tasks far from real factory work. No company has reliably solved autonomous useful work yet, and Unitree is no exception.

What the IPO Means, and What We’re Watching

Unitree’s IPO proves profitable business models exist in humanoid robotics. Raising US$610M onshore without foreign capital signals maturing Chinese capital markets and Beijing’s commitment to leading this sector.

But the markets driving profitability have natural limits. Universities and labs will saturate. Entertainment gigs will not support millions of units. Mass-production claims of 75,000-unit annual capacity, twelve times actual 2025 shipments, are aspirational. The long-term prize is industrial and consumer applications where humanoids substitute for human labor at scale, and that requires a major advance in embodied intelligence. Competition from Tesla’s Optimus, UBTech’s Walker series, AgiBot’s Yuanzheng and Lingxi lines, and dozens of new entrants will only intensify.

We are watching three things closely. Whether Unitree starts landing orders from Fortune 500 manufacturers and global logistics companies, not just local governments and universities. Whether the embodied AI investments produce real generalization to novel tasks, or just more scripted demos. And whether Unitree can hold margins as it shifts from selling developer kits to providing industrial solutions with real integration and support.

Unitree is not a proven winner yet. But we would not count them out. The growth profile is extraordinary, the cost discipline is real, and the company has captured the public imagination in a way no other humanoid maker has outside of Tesla. When people think of humanoid robots now, they think of Unitree. That kind of brand recognition, combined with profitability and a STAR Market listing, means the valuation is likely headed up, potentially way up. This is the first humanoid-focused robotics IPO that actually makes money, and capital will chase that story.

AI is not Unitree’s natural strength, and they have not proven they can close the autonomy gap. But a company with this kind of cash flow and market position has the resources and the runway to try. The hardware is real. The intelligence isn’t. Not yet.

---

All RMB to USD conversions use an exchange rate of 6.88 RMB/USD.