The first humanoid robot company to face public market discipline will not be American, and it will not be losing money. Unitree Robotics begins book-building for its Shanghai STAR Market listing on August 5, 2026, at a base valuation of roughly 42 billion yuan, about $5.9 billion. Figure AI, which has never disclosed meaningful product revenue, carries a private mark near $39 billion. One of those two numbers is about to be tested by strangers with money at risk.
What Actually Happened
Unitree Robotics formally launched its STAR Market issuance process, setting book-building to open on August 5, with the offer price to be determined the following day, public subscription opening August 10, and final allocation results published August 14. The company plans to issue 40.4464 million shares to raise 4.202 billion yuan, roughly $622.7 million, with 85% of proceeds earmarked for research and development. CITIC Securities is acting as sponsor and lead underwriter. The issuance schedule was reported by Gasgoo.
The prospectus contains the numbers that make this listing different from every other humanoid robotics story of the past three years. Unitree forecast first-half 2026 revenue of 1.052 billion to 1.128 billion yuan, growth of 35.62% to 45.41% year over year, and net profit attributable to shareholders of 258 million to 306 million yuan. A humanoid robot manufacturer disclosing positive net income, audited and filed, has no precedent. The revenue and profit ranges were detailed in coverage from BigGo Finance, and the broader listing context was covered by the South China Morning Post.
The comparison that frames the listing is unavoidable. Unitree shipped more than 5,500 humanoid units in 2025 and is targeting 10,000 to 20,000 in 2026, selling its G1 platform from roughly $16,000. Figure AI, the best-capitalized American entrant, manufactured its 1,000th Figure 03 at its BotQ facility on July 23, 2026, running production at roughly one robot per hour, with a 40-unit commercial fleet at a BMW assembly plant billed at approximately $25 per robot-operating-hour. AgiBot, another Chinese manufacturer, sits at 15,000 cumulative units. The unit gap between the Chinese manufacturers and the American ones is now more than an order of magnitude, a point made directly in TechTimes coverage of the subscription opening.
Why This Matters More Than People Think
Private markets have been valuing humanoid robotics on a narrative that says the category is pre-revenue because the technology is not ready, and that whoever solves general-purpose manipulation first captures a market measured in trillions. Unitree's prospectus proposes a different theory: the category already has revenue, the revenue comes from research labs, universities, entertainment, inspection, and light industrial work rather than from replacing factory labor, and a company can be profitably built serving that demand today while the general-purpose problem remains unsolved. Those are two incompatible theses about the same industry, and public market pricing is about to arbitrate between them.
The valuation gap makes the stakes concrete. Figure AI at $39 billion against Unitree at $5.9 billion means the market is currently paying roughly 6.6 times more for the company with 1,000 cumulative units and no disclosed profit than for the company with more than 5,500 annual units and audited net income. That is a defensible position only if you believe Figure's approach, a high-cost humanoid aimed at replacing labor in Western industrial settings, converges on a much larger addressable market than Unitree's approach of selling capable hardware cheaply to anyone who will buy it. It is the same argument made about Tesla versus BYD in 2019, and that comparison did not resolve the way most Western investors expected.
For anyone deploying capital or planning a robotics strategy, the more actionable implication is about component supply. Unitree's cost position at $16,000 per unit is not primarily a software achievement; it reflects a domestic Chinese supply chain for actuators, harmonic drives, reducers, and rare-earth magnets that has been built out over a decade for the electric vehicle and drone industries. An American or European competitor targeting a comparable price point has to either source from that same supply chain, with the geopolitical exposure that implies, or accept a bill of materials several times higher. The humanoid cost curve is therefore an industrial policy question before it is an engineering question.
There is a fourth implication that most coverage will miss because it sits in the use of proceeds rather than in the valuation. Allocating 85% of a $622.7 million raise to research and development is a declaration that Unitree intends to attack the harder problem it has so far avoided, general-purpose manipulation and the foundation models that drive it, funded by cash flows from a hardware business that already works. That is a materially different risk profile from a company attempting the same research on venture capital with no revenue underneath it. If the strategy works, Unitree arrives at the hard problem with a manufacturing base, a supply chain, and a fleet generating real-world interaction data. If it fails, the company still has a profitable hardware business. Very few competitors in this category have that second branch available to them.
The Competitive Landscape
The field splits cleanly into two camps that are not actually competing for the same customer yet. Unitree, AgiBot, and the broader Chinese cohort are selling hardware at volume to a diffuse buyer base that wants a capable platform now. Figure, Agility Robotics, Apptronik, and Tesla Optimus are pursuing contracted deployment into Western industrial operations, where the buyer is a Fortune 500 operations leader and the sale requires safety certification, uptime guarantees, and integration into existing processes. Figure's BMW deployment and its roughly $25 per robot-hour billing model is the clearest expression of that second strategy. Tesla has deployed Optimus internally but has not confirmed third-party commercial availability, and Fremont production had not started as of mid-July 2026.
The historical parallel that fits best is the drone industry between 2013 and 2018. Western entrants raised heavily on enterprise and defense narratives while DJI shipped volume into consumer and prosumer channels, compounded manufacturing learning, drove unit cost down, and then moved up into the enterprise segment from a cost position nobody could match. By the time the Western firms had certified enterprise products, the incumbent had a decade of supply chain advantage and roughly 70% of the global market. The mechanism was not superior research. It was that shipping volume teaches you things that prototyping does not, and those lessons compound.
However, the bear case on Unitree is sitting in its own prospectus, and it deserves equal weight. First-half revenue growth of 35.62% to 45.41% is a sharp deceleration for a company in what is supposed to be the opening phase of a category, and it raises the question of whether the addressable market for a $16,000 research and entertainment platform is closer to saturation than the narrative allows. Net profit of 258 million to 306 million yuan on roughly 1.1 billion yuan of revenue implies healthy margin, but margin on a decelerating base is a different asset than margin on an accelerating one. Skeptics point out that allocating 85% of proceeds to research and development is what a company does when its current product line is approaching its ceiling and the next one is not ready.
Hidden Insight: Profitability Is Evidence of a Small Market, Not a Large One
The reflexive read on Unitree's numbers is that profitability validates the humanoid category. The more careful read is closer to the opposite. A company is profitable at this stage because it chose a market small enough and technically undemanding enough to serve with today's capability: robots for research labs, developer platforms, entertainment installations, inspection routes, and demonstration units. Those buyers do not require the thing that makes humanoids economically transformative, which is reliable autonomous manipulation in unstructured environments over long horizons. Unitree is profitable precisely because it is not attempting the hard problem.
Figure's losses, by the same logic, are not evidence of failure. They are the cost of attempting the version of the problem that actually matters. The $39 billion mark is a bet that solving general-purpose manipulation creates a market so large that today's revenue is noise, and that bet is either wrong by an order of magnitude or right by two. What the public listing does is force these two theories to be priced against each other daily, in liquid markets, with disclosure requirements. That has never happened in this category, and the resulting price series will become the reference input for every private humanoid round negotiated over the next three years.
The consequence for private valuations is where this gets uncomfortable. Every humanoid startup currently raising has been marked against private comparables, which are set by the last round rather than by any external test. Once Unitree trades, there is a public multiple on humanoid revenue and a public multiple on humanoid units shipped. If Unitree trades at, say, 20 times revenue, then a private company with no revenue has to justify its mark on something other than comparables, and the "we are the Western Unitree" pitch acquires a number attached to it. Public listings do not just price the company that lists. They reprice everything that looked like it.
There is a second-order effect on the talent and capital flowing into the sector. A liquid Chinese humanoid stock gives global investors an instrument to express a view on the category without waiting for a Western IPO, and it will attract index and thematic fund flows, particularly through STAR Market and humanoid robotics ETFs. That flow is a subsidy to the Chinese cohort's cost of capital at exactly the moment the Western cohort is burning the most. If the listing trades well, expect AgiBot and others to accelerate their own listing timelines within twelve months, compounding the advantage. Capital markets access is itself a competitive weapon, and one side of this race is about to get it first.
The data argument cuts against the profitable-but-narrow reading, and it should be stated fairly. Every G1 sold into a research lab or an entertainment venue is a sensor package operating in an unstructured environment, and the aggregate interaction hours across a fleet approaching five figures dwarf what a 40-unit deployment can produce, regardless of how well instrumented that deployment is. If robot foundation models follow the scaling behavior that language models did, fleet size becomes the input that matters most and the volume leader inherits a compounding advantage that capital cannot short-circuit. Whether that analogy holds is the central open question in robotics right now, and Unitree's listing effectively lets public investors take a position on it for the first time.
What to Watch Next
Over the next 30 days, the single most informative number is where the offer price lands relative to the 42 billion yuan base valuation after book-building closes on August 6, and then how the stock trades through the August 14 allocation results and the first sessions after listing. STAR Market debuts have historically opened well above issue price, so the useful signal is not day one but weeks two through six, once the allocation flip has cleared. A stock that holds above issue after the initial float rotation is a real market verdict on humanoid economics; a stock that fades to issue is a verdict on the deceleration in the prospectus.
Over 90 days, watch Unitree's first reported quarter as a public company and specifically whether unit shipments track toward the 10,000 to 20,000 target for 2026. Half-year revenue growth in the 35% to 45% range needs to reaccelerate in the second half for that target to be credible. Watch also for any disclosure of the revenue split between hardware sales and services, and between domestic and export customers. Export concentration would introduce a policy risk the prospectus valuation does not obviously carry.
Over 180 days, the markers are competitive rather than financial. Watch whether Figure discloses a second named industrial customer beyond BMW, whether Tesla confirms external Optimus availability after Fremont production starts, and whether AgiBot files for its own listing. Watch also for the first public teardown comparing a G1 bill of materials against a Figure 03, because that document, whenever it appears, will settle the cost-position argument more decisively than any earnings release. If the gap is as wide as unit prices imply, the strategic conversation in Western robotics shifts from capability to supply chain within a single quarter.
Unitree is profitable because it declined to attempt the hard version of the problem, and Figure is unprofitable because it did. The public market is about to tell us which choice was the business.
Key Takeaways
- Book-building opens August 5, subscription August 10, with allocation results published August 14 on Shanghai's STAR Market.
- 4.202 billion yuan raise, roughly $622.7 million, from 40.4464 million shares at a base valuation near 42 billion yuan, about $5.9 billion.
- First-half 2026 net profit of 258 million to 306 million yuan makes Unitree the first humanoid manufacturer to list with audited profitability.
- Revenue growth of 35.62% to 45.41% is a sharp deceleration and the strongest argument against the listing valuation.
- Unitree shipped 5,500-plus units in 2025 against Figure's 1,000 cumulative, an order-of-magnitude volume gap that compounds manufacturing learning.
Questions Worth Asking
- If profitability at this stage signals a small addressable market rather than a validated one, how should that change the way you read every early-revenue AI hardware company?
- Once a public humanoid multiple exists, what happens to the private marks that were justified entirely by comparables?
- If the humanoid cost curve is set by a supply chain rather than by algorithms, is your robotics strategy actually a procurement strategy?
