Unitree Signals Humanoid Profitability With $623M IPO

Unitree Robotics started book-building on Shanghai's STAR Market today, August 5. The company is asking investors to value it at a minimum of 42 billion yuan, roughly $6.2 billion, for a business that shipped 5,500 humanoid robots last year and turned a real profit doing it. That last detail is the one that matters, because no US-based humanoid robot company can say the same thing.

What Actually Happened

Unitree launched its IPO issuance process on August 5, 2026, with book-building open to institutional investors today and public subscription set to begin simultaneously for offline and online channels on August 10. The company is issuing 40.4 million new shares, representing 10% of post-IPO capital, to raise approximately 4.2 billion yuan (about $623 million) at a floor valuation of 42 billion yuan. Pricing is scheduled for August 6, with final subscription results due August 14 and the listing itself expected around August 19. CITIC Securities is acting as sponsor and lead underwriter, according to South China Morning Post.

The financials behind the deal are worth pausing on. Unitree reported 1.71 billion yuan in 2025 revenue, a fourfold increase year-on-year, and 287.6 million yuan in net profit, which doubled from the prior year. Gross margins are running at around 60%. Those are software-company margins attached to a hardware business that assembles physical robots and ships them to customers in volume. In 2025, humanoid robots crossed 51.5% of Unitree's total revenue, up from 27.6% in 2024, according to Seoul Economic Daily. The company shipped roughly 5,500 humanoid units over the year, making it the world's volume leader in that category. Founder Wang Xingxing retains 31.29% ownership and 65.31% voting rights after the offering, meaning operational control of the company stays concentrated even as public shareholders acquire their 10% stake.

The timing of the IPO is not accidental. It came two days after the United States Federal Communications Commission added foreign-made humanoid robots and quadrupeds to its Covered List on July 28, effectively banning new imports of Chinese-built robots into the US market. That ban hits Unitree where it has already been shrinking: US revenue fell from 19.54% of total in 2024 to 13.3% in 2025, as the company diversified into European and Southeast Asian markets ahead of expected regulatory pressure. The FCC move arrived, and two days later Unitree formalized its IPO timeline, according to reporting from RoboZaps. The message to the market: US access is not the point anymore. The company is building a capital base on public markets that does not depend on American customers to justify its valuation.

Why This Matters More Than People Think

Every major humanoid robot company in the United States is either pre-revenue or burning cash at a rate that would alarm a venture fund in any other decade. Figure AI is valued at roughly $39 billion as of its most recent funding round but has not disclosed meaningful commercial revenue. Tesla's Optimus program had not begun Fremont production as of mid-July 2026, though Elon Musk confirmed production would start in late July or August. Boston Dynamics has pivoted to enterprise software and services. Against that backdrop, Unitree is asking a simple but devastating question: why is the company actually generating $623 million of new equity capital the one that makes physical AI work at a profit?

The 60% gross margin figure deserves particular scrutiny. At that level, Unitree is structurally closer to a platform business than a traditional robotics manufacturer. The company builds its own actuators, its own sensors, and its own foundation model integration layer. Vertical integration at that depth usually compresses margins because the capital intensity is spread across more steps. Unitree is doing the opposite, suggesting it has either found genuinely lower-cost manufacturing processes that US competitors cannot replicate at current labor rates, or it has reached a volume scale where amortized design costs are dropping faster than unit economics. Probably both, and the margin line in the IPO prospectus is the evidence for that claim.

The broader implication is that the humanoid robotics race is bifurcating. One track is the US venture-funded model: raise billions, promise future deployment, delay production, revise timelines. The other track is the Chinese manufacturing model: reach volume, hit profitability, then access public markets. Unitree's IPO is the first real data point for what track two is actually worth when millions of buyers and sellers can disagree in real time with real money. Secondary market expectations have already anchored above 100 billion yuan, more than $14.8 billion, which would represent a 2.4x premium to the IPO floor valuation. That gap will either close upward, validating the manufacturing model, or not, validating the skeptics who argue Unitree's margins are a China-specific artifact that cannot travel globally.

The Competitive Landscape

Unitree holds approximately 20% of the global humanoid robot market by units shipped. Its nearest domestic rival, AgiBot, is pursuing a Hong Kong listing and has disclosed 15,000 cumulative units built, though production mix between humanoid and non-humanoid units is less clear. DEEP Robotics and LEJU Robotics are also preparing Chinese exchange IPOs. The Chinese wave of humanoid robot public listings in the second half of 2026 is beginning to look like the EV wave of 2020 and 2021, when BYD, Nio, Li Auto, and XPeng all accessed public capital within months of each other. That EV comparison is instructive but also cautionary: a wave of public listings does not guarantee that every company in the wave survives the subsequent commodity price cycle.

On the US side, Boston Dynamics remains the prestige name but operates in a different market segment targeting enterprise logistics rather than general-purpose humanoid deployment. Figure AI's $39 billion valuation assumes a future where its robots are performing broad tasks at scale in BMW factories and elsewhere. The problem is that assumption is priced in before the robots are shipping at volume. Google DeepMind released Gemini Robotics 2 on July 30, adding whole-body control capabilities that could accelerate deployment of robotics foundation models across multiple hardware platforms, blurring the line between software-layer competitors and hardware makers. If DeepMind's models commoditize robot intelligence, the manufacturers with the best hardware cost structures win. That benefits Unitree more than any of its US counterparts.

The FCC Covered List ban also reshapes the competitive map in a specific way. Unitree loses direct US market access for new imports. But the ban does not prevent Unitree from licensing its hardware designs, entering manufacturing joint ventures in third countries, or supplying components to US system integrators who do final assembly domestically. The historical parallel is Huawei's response to the Entity List: initial shock, followed by aggressive pivots to markets where access remained open, combined with investment in intellectual property that could be licensed rather than exported. Unitree has already started that diversification. The question is whether the US market was ever large enough, relative to Southeast Asia, Europe, and Japan, to constitute a true long-term growth driver.

Hidden Insight: Why Profitability at This Stage Changes Everything

The conventional narrative about frontier hardware companies is that profitability is a problem for the next CEO. Burn capital, capture market share, achieve scale, then optimize economics. Tesla ran that playbook for a decade before its first full profitable year. Every major humanoid robot startup in the US is operating on roughly that assumption, with the modification that AI capabilities will be the moat rather than raw manufacturing scale. Unitree breaks the narrative by achieving profitability not at the end of the scaling phase but during it. That is a structurally different company from a risk perspective, and it suggests the economics of humanoid robot manufacturing may not actually require a decade of losses to reach viability.

The key mechanism is actuator cost. The powered joints in a humanoid robot are historically the most expensive single component category, often representing 40-60% of total bill of materials. Unitree has been vertically integrating its actuator manufacturing since 2019, when it was still primarily a quadruped robot company. That early investment in actuator technology, made before anyone was calling humanoid robots a strategic priority, has compounded into a structural cost advantage that is now showing up in the gross margin line. US competitors are generally purchasing actuators from third-party suppliers or have started in-house programs more recently, meaning the learning curve advantage is already priced into Unitree's margins and not yet reflected in US alternatives.

However, the risk is real and worth pricing into any position. Unitree's profitability depends on margins that have been built in a Chinese manufacturing environment, where labor costs, component supply chains, and government industrial policy all tilt in favor of physical AI manufacturers. Those tailwinds cannot necessarily be replicated in other geographies. If Unitree tries to establish non-China production to serve markets where Chinese imports are politically constrained, it may find that the economics that produce 60% gross margins in Hangzhou generate 30-35% gross margins in Vietnam or Mexico. The company has not disclosed how much of its margin structure is structural versus geographic, which is the most important unknown in the IPO prospectus for any long-term investor who needs to model what happens when the manufacturing base shifts.

There is also a second-order effect that the market is underpricing: the Unitree IPO, if it prices at or above the floor valuation and trades well on listing, creates a credibility signal for every Chinese robotics company pursuing capital. AgiBot, DEEP Robotics, and a dozen other companies become fundable at higher valuations the moment Unitree demonstrates that public markets will absorb a profitable humanoid robot company at 15-20x revenue. The Chinese government's 2-trillion-yuan investment target for data centers and AI infrastructure over five years provides the demand-side backdrop for sustained deployment. A wave of cheap, profitable humanoid robots from public companies with low costs of capital would change the deployment timeline for physical AI globally by several years, compressing a transition that Western analysts have been modeling as a 2028-2032 event into a 2027-2029 reality.

What to Watch Next

The most immediate indicator is the IPO clearing price on August 6. The floor is 42 billion yuan. Secondary market expectations have set an informal target above 100 billion yuan. The gap between those two numbers will tell you how much premium investors are willing to pay for a profitable, volume-shipping humanoid robot company as opposed to a pre-revenue one. If Unitree prices at the floor or below, it signals that even demonstrated profitability at this scale is not worth the narrative premium the secondary market was assigning. If it prices above the floor by a wide margin, it will apply direct pressure to US-based humanoid robot companies to either show revenue numbers or accept a valuation step-down in their next private rounds.

Over the next 90 days, watch two specific metrics: Unitree's US revenue percentage in its H1 2026 disclosure, and whether any US system integrator announces a domestic assembly partnership with Unitree or another Chinese humanoid robot manufacturer. The FCC ban applies to new imports of foreign-made robots, but there is no clear prohibition on domestic assembly using Chinese-made components. The legal and regulatory grey zone there is large enough to drive a forklift through, and smart integrators will exploit it. If that happens within the quarter, the FCC's ban will be partially circumvented before it has even been in force for six months, creating a test of whether US regulators will respond with a broader restriction on Chinese robot IP.

Over the next 180 days, the Unitree listing sets a precedent for how AgiBot, which is heading toward a Hong Kong listing, will be valued. AgiBot's 15,000 cumulative units and its Hong Kong domicile give it a different investor profile from Unitree's STAR Market listing. If both companies are trading publicly with solid first-half 2026 numbers by early 2027, the humanoid robot category will have its first sector price-to-earnings range, the same milestone that the EV sector reached in 2020-2021 and the semiconductor sector reached in 2016-2018. Once that happens, capital allocation decisions by every robotics lab, venture capital fund, and corporate research department get recalibrated against a market-set benchmark. That is the moment the industry truly accelerates, and the US will be watching from outside the primary capital markets that are setting those benchmarks.

The first profitable, volume-shipping humanoid robot company just went to market, and it's not American.

---