On June 1, 2026, Hangzhou-based Unitree Robotics cleared the Shanghai Stock Exchange's listing review in a lightning 73 days, becoming the first humanoid-robot company listed on China's A-share market. But a deep-dive analysis by Beijing Daily reveals that the humanoid-robot industry is standing on a "structural fracture" between technological myth and commercial reality — revenue is surging even as net profit plunges. The essence of the IPO wave is not that "the industry has matured" but that "the industry has realized it is burning cash faster than it imagined."
The Contradiction Written in the Numbers
Unitree's prospectus presents a set of data divided against itself.
On the revenue side, the company booked RMB 159 million, RMB 393 million and RMB 1.699 billion respectively from 2023 to 2025 — a compound growth rate above 226%. Net profit swung from –RMB 11 million to RMB 95 million and then RMB 278 million; the company really is moving from losing money to making it.
But in the first quarter of 2026 the picture changed abruptly: revenue of RMB 423 million, up 68.49% year on year, while net profit fell 52.55% year on year. More sold, less earned — in manufacturing, that combination usually signals systemic margin compression rather than seasonal fluctuation.
Humanoid-robot shipments exceeding 5,500 units, running speed raised from 5 meters per second to 10 — these headline highlights stand in stark contrast to the financial data. Xue Hongyan, vice president of Xingtu Financial Research Institute, put his finger on the crux: "The market's reaction may simply be that expectations were stretched too high." Two years of dazzling performances by Unitree, compounded with the grand narrative of embodied AI, led the public to sketch in advance a straight, steep growth curve. The cost structure and the falling gross margin in the earnings report broke that curve.
"Between the public perception at the peak moments and the cost structure in the earnings reports, there is a fracture." — Xue Hongyan
Listing Is Not the Finish Line — It Is a Resupply Station
Xue Hongyan offered one precise judgment in the report: "For Unitree at this moment, going public is more like a crucial 'resupply.'"
The humanoid-robot sector remains in the "early stage of technological breakthroughs and commercial exploration." 2025 was defined as Year One of mass production; 2026 enters the grind of scaling that production up — meaning products must move from laboratories and demonstration projects to real production lines and real markets. That demands enormous capital: technology iterates at extreme speed and R&D spending only grows; building the supply chain requires covering precision reducers, sensors, motors, control systems and many other links simultaneously; and market education requires proving to potential customers, over and over, that the robots are usable and that the investment pays back.
Once the primary market starts asking hard questions about cost-effectiveness, pushing open the door to the secondary market becomes an inevitable choice for any company trying to survive and stay in the race. Unitree plans to raise RMB 4.202 billion, earmarked for four projects: intelligent robot model R&D, robot-body R&D, new-product development, and a manufacturing base. Deep Robotics plans to raise RMB 2.503 billion; Leju Robotics plans to raise RMB 2.6 billion while still unprofitable, using the ChiNext board's fourth listing standard — meaning that even companies that have not turned a profit can now finance long-horizon R&D.
The collective listing of three humanoid-robot companies is better described not as "the industry has matured" but as "the industry has realized it is burning cash faster than it imagined." Judging at least by the stated uses of the proceeds, R&D and manufacturing are the absolute center of gravity — not market expansion.
The Other Side of the "Supply-Chain War"
The competitive logic of humanoid robots is fundamentally different from consumer electronics: consumer electronics first had a mature supply chain (China-made global electronics production chains) and then the end products; humanoid robots had the product vision first, and the need to rebuild the supply chain afterward. The cost curves of core components — motors, reducers, dexterous hands — have not yet flattened, and economies of scale still need time to materialize.
This is also the structural reason behind Unitree's Q1 profit decline: gross margins were cut to seize market share in shipments, while the supply-chain side has not yet reached the amortization inflection point of economies of scale. Technology in the robotics industry iterates at extreme speed — today's optimal solution may be overtaken by a competitor three months from now, and that tempo forces companies to constantly reweigh R&D investment against gross-margin management.
The Structural Meaning of the IPO Wave
Deep Robotics and Leju Robotics "taking the baton" after Unitree to charge at the capital markets reveals that the financing model of the humanoid-robot sector is undergoing a systematic shift.
From the private-market path of "angel round → A/B/C rounds → strategic financing" to the public-market path of "private → IPO → continuous financing." Behind this is a repricing of the sector's commercial prospects by the primary market: early investors are starting to seek exits, transferring subsequent growth risk onto the secondary market. This is not an idiosyncrasy of any single company but a capital-structure transformation across the whole sector: humanoid robotics is passing from "frontier-technology exploration" into the stage of "public-market validation."
The First Post-Listing Validation Data — 11,000 G1 Units Produced and the First Experience Center
On June 2, the day after its IPO cleared review, Unitree released the latest data on G1 humanoid-robot production: cumulative production had reached 11,000 units. Meanwhile, the first embodied-AI experience center opened in Shanghai, with exhibits including the Go2 quadruped robot (clad in a realistic simulated panda costume) among other products in the lineup.
This data provides an initial direction for post-listing validation of the "mass-production-versus-profit contradiction" at the heart of the IPO analysis. The G1's leap from Year One of mass production in 2025 to 11,000 units cumulative means shipments are climbing along a scaling trajectory, edging closer to the scale range envisioned by the investment projects in the prospectus. But at the same time, the number carries its own validation pressure —
The announcement of 11,000 units shipped came after a first-quarter report in which net profit plunged 52.55%. The timing is coincidental, but logically the two form a natural contrast: the analysis above argued that the structural cause of the Q1 margin compression was "cutting gross margins to seize market share in shipments," and the disclosure of 11,000 units is itself direct evidence that this strategy is being executed. Shipments are expanding — but whether gross margins can rebound as economies of scale kick in is the core metric to watch after the IPO.
The opening of the first experience center points to the other leg of commercialization: market education aimed at end consumers. With a base of 11,000 robots already out there, the company needs to let potential customers, investors and partners feel the landing form of its products' capability in a more intuitive way — not just parameters on paper and numbers in earnings reports.
This new data has not yet changed the core judgment of the IPO analysis: the humanoid-robot industry is standing on the structural fracture between "shipment ramp-up" and "the profit inflection point." Eleven thousand units prove that Unitree is ramping up on schedule — but reaching the profit inflection point still awaits validation from the next quarterly report.
The Listing Window — Must List in Mid-August (increment, 2026-08-05)
In a livestream, Guyè (a finance-focused Weibo commentator) offered the latest reading: Unitree must list between mid- and late August, with the subscription opening on the 10th, no more delays.
The reasoning: any further delay risks being dragged down by US developments — across the Pacific, the US keeps poking the AI tiger with a stick, and no one knows when the tiger might turn and bite back. While domestic market liquidity remains ample and policy support is in place (the STAR Market's fifth listing standard), companies that qualify should list while they can and satisfy standards while they can — once listed, deal with the next set of problems. Miss this window and a US tantrum could trigger a reflexive sell-off in China, and the market would crack.
This is part of his recurring thesis: high-volatility boards (the Beijing Stock Exchange, STAR Market, ChiNext) exist in essence to finance and relieve the debt burden of issuers, and frontier-technology companies should list precisely while they are commercially early-stage — slightly stuck but with products on the ground. His first piece on the STAR Market was about Ziguang Storage — an optical-storage company that listed and then ran into trouble, so the lesson is not to mimic the indiscriminate listings of the "Easy-Master" era, but that headline companies that qualify must absolutely seize the window.
2026-07-31 Guyè livestream, "Boldness and Luck, Pick One" — "Miss this window, and a US tantrum triggers a reflexive sell-off at home, and the market cracks."
The Issue Price Lands — RMB 104 Estimate and RMB 4.2 Billion Raise (increment, 2026-08-06)
On August 5, Unitree's IPO timetable firmed up: August 5 was the STAR Market's preliminary inquiry day; August 7, 14:00–17:00, online roadshow; August 10, online and offline subscription both open; payment deadline August 12. The planned new-share offering: 40.4464 million shares, representing 10% of post-issuance total share capital, putting total share capital after issuance at roughly 404 million shares. Planned proceeds: RMB 4.202 billion, mainly for polishing the robot "brain" and forging the robot "body." Market estimates put post-IPO market capitalization above RMB 40 billion. At the STAR Market's subscription unit of 500 shares and the estimated market cap, the implied issue price is roughly RMB 104 per share, with one subscription lot costing about RMB 52,000 to subscribe. On the earnings side, the company projects RMB 1.052–1.128 billion in H1 2026 revenue, up 35.62–45.41% year on year. After listing, Unitree becomes the first A-share company focused on the humanoid-robot field.
This aligns perfectly with the "must list in mid-August" judgment above: subscription on the 10th, listing in mid-to-late August, slipping neatly into the slot where liquidity is ample and the policy window (the STAR Market's fifth listing standard) remains open. The bet is not on earnings — RMB 1.1 billion in half-year revenue against a RMB 40 billion market cap puts the P/E far above traditional manufacturing valuations — but on the scarcity pricing of "the first humanoid-robot stock."
This section adds execution-level IPO details — issue price, proceeds size, subscription timetable — and does not change the prior core judgment of "commercialization ahead of profit realization, window-period financing." It simply moves the listing from a forecast to a specific date on the calendar.
Issue Price Lands at RMB 150.80 — The FCC Response and the DeepSeek Partnership in the Roadshow's 300-Plus Q&A (increment, 2026-08-08)
On August 7, Unitree held its first online investor-exchange session ahead of the STAR Market IPO. Chairman Wang Xingxing answered questions from 14:00 until 19:40 — more than five hours of roadshow and over 300 questions answered. The company filed an announcement on August 6: issue price RMB 150.80 per share; subscription online and offline both open on August 10; payment deadline August 12. This is far from the "roughly RMB 104 estimate" recorded earlier — the actual pricing is about 45% higher, and the implied market-cap expectation has lifted accordingly.
The two most informative threads of the roadshow revolved around two distinct risk lines. The first was the FCC's China-related restrictions: the U.S. Federal Communications Commission recently rolled out China-related restrictive measures on advanced robotics equipment, power inverters and the like. Wang Xingxing responded that all of the company's currently selling flagship products — humanoid robots G1, H2 and R1, and quadruped robots Go2, B2 and A2 — have already obtained FCC certification, which places them within the "advanced robotics equipment" scope of the announcement. The policy change therefore does not currently affect the continued sale of existing flagship products in the U.S. market — what the new restrictions block is the entry of future models, not the continued sale of existing ones. The second was the DeepSeek partnership: the two sides have signed a Strategic Cooperation Memorandum, with the priority areas being R&D collaboration targeting artificial general intelligence, in-depth collaboration on high-performance general-purpose robots, and deep collaboration on AI foundation models.
Overseas-revenue data was also disclosed at the roadshow, in the prospectus caliber, for the first time: in 2025, overseas revenue accounted for 43.65% of total revenue, with overseas revenue of RMB 731.6553 million, and Unitree ranked first globally in humanoid-robot shipments. Wang Xingxing characterized the industry as "still in a stage similar to the early days of the home computer," with generalization capabilities still to be lifted across the whole industry.
The Arithmetic of Subscription Eve — RMB 60,000 Threshold and a 0.02% Allotment Rate (increment, 2026-08-09)
On the eve of the August 10 subscription, every subscription parameter for Unitree was locked in. The issue price is RMB 150.80 per share; the offering is 40.4464 million shares, raising a total of RMB 6.099 billion; post-issuance market capitalization is roughly RMB 60.99 billion; and the implied P/E is 219.23x (calculated against 2025 non-recurring-net-profit of RMB 590.7528 million). Compared with Changxin Memory, Unitree's biggest difference is the ceiling-application threshold: only RMB 60,000 in Shanghai-exchange market-cap is required (ceiling subscription 6,000 shares), versus Changxin's RMB 16.72 million — a threshold reduction of two orders of magnitude. But a low ceiling-application threshold does not mean an easy allotment: only 6.47 million shares are issued online initially; at 500 shares per subscription lot, that is roughly 12,900 lot numbers up for grabs, and institutional estimates put the allotment rate in only the 0.02%–0.05% range — possibly lower if subscription demand runs hot.
The per-lot arithmetic is equally striking. Winning one lot of 500 shares requires a payment of RMB 75,400, roughly 17 times the winning-payment amount for Changxin Memory (RMB 4,330). Even if the allotment rate is low, the amount of capital that needs to be set aside in the event of winning a lot is not small. Brokers therefore flag four reminders: eligibility and market-cap threshold (2 years of trading experience + 20-day daily-average assets ≥ RMB 500,000; the T-2 daily-average Shanghai-exchange market-cap over the prior 20 days ≥ RMB 10,000; only stocks with codes starting 60 or 688 count); reserved funds (turn off auto money-market funds and reverse repos; ensure sufficient balance in margin accounts); the abandonment red line (3 winning lots in 12 months not fully paid bars the investor for 6 months); and post-listing volatility risk (Q1 non-recurring net profit of RMB 40.2536 million, down 52.55% YoY; H1 projected RMB 236–283 million, down 6%–22% YoY).
The profit forecasts are split at the extremes: CCB International projects Unitree's post-listing market cap could reach RMB 109 billion (implied target P/S of 32x), implying ~79% upside on day one and RMB 59,000 profit per winning lot. If it rises 500%, the market cap hits RMB 365.9 billion and per-lot profit reaches RMB 377,000 (this year's second-largest "fat lot" on A-shares, behind only Lianxun Instrument's RMB 389,000). But some institutions argue the fair-value market-cap center sits at RMB 30–45 billion, in which case a day-one break is possible. Against the same issue price, one calculation produces a RMB 59,000 profit per lot, another produces a possible break — the scarcity pricing of "the first humanoid-robot stock" is undergoing its first bull-bear collision.
This section advances the listing from "a landed price" to "the subscription parameters": ceiling-application threshold, allotment rate, per-lot payment, abandonment red line and the profit-forecast split. The prior judgment of "scarcity pricing versus commercialization ahead of profit realization" stands, but the magnitude of the market's disagreement now has concrete numbers — between RMB 109 billion and RMB 30–45 billion lies an entire order-of-magnitude gap of market-cap imagination.