A packed "event calendar" is now materializing. From August 19 to 23, the 2026 World Robot Conference will be held in Yizhuang, Beijing; Unitree Robotics—the first humanoid robot company on the A-share market—has concluded its subscription and will be listed in late August; across the ocean, Tesla's mass production line for Optimus is undergoing installation and debugging. These three major events are landing intensively within the same window period, leading the market to dub August the "Robot Super Month."
Underpinning all this is a fundamental shift in the industry narrative: humanoid robots are transitioning from "exhibition prototypes" to the "year of delivery."
01. The Year of Delivery: Data Starts to Speak Over the past few years, the biggest embarrassment for humanoid robots was "exhibiting without selling"—products remained at the concept prototype stage, with manufacturers performing backflips and dancing at exhibitions, but actual commercial deployment was scarce. In2026, this situation has been completely reversed, and the data is now speaking.
Shipments have taken the lead in exploding. According to data from the U.S. market research firm Smart Analytics Global, global humanoid robot shipments reached 19,100 units in the first half of 2026, a year-on-year increase of 272%. More symbolically significant is the structural shift: industrial and commercial applications now account for over 70% of first-half shipments, compared to just 50% a year ago. Humanoid robots are moving from "exhibition demonstrations" to "production line productivity."
Chinese manufacturers dominate this delivery race absolutely. Chinese humanoid robot manufacturers account for over 97% of global shipments, and the top six global shippers are all from China, with a combined market share of 87.87%. Data from the Ministry of Industry and Information Technology (MIIT) shows that China's humanoid robot production has increased from about 20,000 units in 2025 to over 40,000 units in the first half of 2026, and is expected to exceed 100,000 units for the full year. Calculations by the Gaogong Robot Industry Research Institute are even more optimistic: China's humanoid robot production in 2026 is expected to scale across the 100,000 to 200,000 unit level.
The capacity utilization rate ramp-up curves of leading manufacturers are clearly visible. Unitree Robotics has cumulatively produced and rolled out about 12,500 humanoid robots in 2026; the monthly capacity of its G1 at the Hangzhou factory has reached 200 units, with plans to increase it to 500 units per month in the fourth quarter; UBTECH's Walker S produces 150 units per month, with a 2026 shipment target of 3,000 units; Agibot's A2 produces 100 units per month, and Fourier Intelligence's GR-2 produces 80 units per month.
More critical than production volume is the "factory entry" validation of real machines. UBTECH's Walker S series passed a three-month field validation at NIO's Hefei factory, officially obtaining "mass production access qualification." The first batch of 20 robots has been integrated into the production sequence to perform door lock detection and seatbelt installation tasks; BYD's self-developed "Yao Shun Yu" robots have deployed about 150 prototypes for practical training, and its Shenzhen factory already has 50 Unitree G1s in operation, with plans to expand to 200 units by the end of the year; UBTECH's Walker S has entered Foxconn's production lines to conduct process tests such as component sorting and visual inspection; JD.com and Cainiao have begun pilot programs in the last-mile delivery and warehouse sorting stages for humanoid robots.
Large-scale procurement from government and enterprise clients is building the "foundation" of the industry. State Grid plans to purchase about 8,500 special robots this year, with an amount of approximately 6.8 billion RMB; SF Express and China Post are deploying logistics sorting robots in large quantities. The industry's judgment on this turning point is quite incisive—Tian Feng, founder of Songyan Dynamics, pointed out: "2026 is the watershed from 'high-traffic videos' to 'high-hour billing.' The real inflection point does not lie in whether robots can perform perfect backflips, but in whether they can endure night shifts, handle tedious tasks steadily, keep errors controllable, make deliveries replicable, and ensure accountability is clearly calculated."
Huayuan Securities characterizes this stage as: the humanoid robot industry is moving from the "0 to 1" validation period into the new stage of "1 to 10" mass production ramp-up, and 2026 is expected to usher in exponential growth.
02. Unitree's Listing: Valuation Anchor and Industrial Chain Re-evaluation The most direct catalyst for the sector is the continued advancement of Unitree Robotics' IPO.
On the evening of August 6, Unitree Robotics announced an offering price of 150.80 RMB per share, with a calculated offering market capitalization of about 61 billion RMB, exceeding market expectations. The oversubscription ratio was about 45%, and the diluted offering price-to-sales ratio was 35.89 times (based on 2025 revenue). On August 10, the online and offline subscription was completed, and the strategic placement list was nothing short of luxurious: the National Social Security Fund, DeepSeek (allocated about 141 million RMB), China National Petroleum Corporation, Tencent, and other institutions and industrial capital were prominently listed.
The frenzy on the subscription side has broken records. Unitree Robotics' valid online subscription accounts reached 9.7846 million, setting a new historical record for the STAR Market; the subscription multiple was as high as 8,288 times, with a final winning rate of only 0.0181%. The company plans to raise 4.2 billion RMB, half of which will be invested in the R&D of embodied AI large models, and the remaining funds will be used to expand humanoid robot production capacity—from 20,000 units per year to 100,000 units.
As the first domestic legged robot OEM to achieve scaled profitability, the significance of Unitree's listing goes far beyond a single company entering the capital market. It provides a "valuation anchor" for the entire humanoid robot industrial chain—the market generally compares it to the market paradigm where CATL's listing drove the re-evaluation of the entire lithium battery industrial chain back in the day. Zheshang Securities pointed out that Unitree's IPO pricing of 150.80 RMB per share, coupled with the accelerated diffusion of Xiaomi's open-source robots, indicates that the sector is switching from thematic narratives to commercial validation.
The wave of capitalization is far from over. Dobot's ChiNext IPO passed the review on July 22, taking only 86 days from application to approval; Leju Robotics and Deep Robotics have entered the "inquired" stage at the end of May, respectively; manufacturers such as Agibot, Galbot, and ZQ Robotics also have plans to update their capitalization progress. Data from Goldman Sachs shows that the median P/E ratio of the Asia-Pacific robot basket is 22 times, while similar assets in the U.S. are at 28 times, with the former about 21% lower than the latter—Asian robot and automation enterprises currently have relatively low valuations, but their earnings growth expectations are no weaker than similar assets in the U.S.
03. August Event Calendar: Conference, Mass Production, and Policy If Unitree's listing is the catalyst for the "domestic chain," then the World Robot Conference and Tesla's mass production serve as the dual engines of "events + overseas."
World Robot Conference: The largest in history. From August 19 to 23, the 2026 World Robot Conference will be held in Yizhuang, Beijing, concurrently hosting the 2nd World Humanoid Robot Games. According to disclosures by the Chinese Institute of Electronics, the exhibition this year has over 300 participating enterprises, a 36% increase from last year; there are more than 2,000 exhibits, up 27%; and the number of newly launched products exceeds 150. The conference will also create a robot consumption street for the first time and set up a dedicated zone for the complete humanoid robot industrial chain. More importantly, the State-owned Assets Supervision and Administration Commission (SASAC) will establish a central enterprise robot innovation consortium on-site during the conference, and orders for industrial humanoids and factory robots are expected to accelerate their landing.
Tesla Optimus: Countdown to mass production. According to supply chain research by Soochow Securities, Optimus V3 is expected to debut from late August to early September. Current weekly production orders have reached over 200 units, and are expected to ramp up to 1,000 units in September; the first Optimus production line at the Fremont factory has been installed on-site, with a long-term planned annual capacity of about 1 million units at this base; the construction of the Optimus plant at the Texas Gigafactory is accelerating, with a long-term planned annual capacity of about 10 million units, and the company expects production to start around the summer of 2027—the parallel advancement of the two bases is not a revision of a single target. In the second quarter of 2026, Optimus Gen-3 had already started mass production, priced at about $49,000; Elon Musk stated that costs are expected to drop below $20,000 after long-term scaling. Securities China views point out that starting in August 2026, Optimus will end its demonstration attributes and be embedded into existing vehicle production lines to complete complex assembly.
Policy combination: From encouragement to regulation. On August 7, the MIIT, jointly with the State Administration for Market Regulation (SAMR), released the "Safety Requirements and Test Methods for Industrial Application Scenarios of Humanoid Robots" (Draft for Comments), proposing for the first time a three-level safety certification system: L1 (isolated area operation), L2 (low-speed human-robot collaboration area), and L3 (fully dynamic human-robot integration). It requires scenarios of L2 and above to pass 18 mandatory tests, including force-controlled collision limits and emergency stop response times—the dimension of industry competition is switching from "whose demo is more stunning" to "who can reliably create value on real production lines." In addition, as early as June 8, the MIIT and SASAC launched a special action for real-scene practical training, promoting large-scale real-scene validation of humanoid robots in fields such as NEV (New Energy Vehicle) manufacturing, 3C electronics, and heavy industry inspection, setting a target of landing 10,000-level equipment within the year; Hangzhou promulgated the country's first local regulation focusing on the field of embodied intelligent robots and launched the National AI (Artificial Intelligence) Application Pilot Base.
Goldman Sachs pointed out in its latest report that the AI investment narrative is shifting from the infrastructure layer to the application layer, and humanoid robots are expected to become an important growth pole in the next stage. It is estimated that global humanoid robot shipments will reach 76,000 units in 2027 and rise to 502,000 units in 2032. Omdia's forecast is even more long-term: global humanoid robot shipments will grow from 18,000 units in 2025 to 2.6 million units in 2035.
04. Cost Avalanche: From 650,000 to 26,900 Another undercurrent of the year of delivery is the avalanche-like drop in prices—which directly determines whether demand can shift from "trying out the new" to "scaled procurement."
Unitree's price reduction curve is highly representative: the H1 released in August 2023 had an initial launch price of about 650,000 RMB, and the average unit price of humanoid robots that year was about 590,000 RMB; by 2026, the current selling price of the G1 basic version has dropped to 85,000 RMB; and the latest R1 series dual-arm humanoid robot has a starting price of only 26,900 RMB—a drop of over 95% in three years. Tesla's Gen-3 is priced at about $49,000, and Musk stated that costs are expected to drop below $20,000 after long-term scaling.
Behind the price avalanche is the deep restructuring of the entire industrial chain. China Post Securities' teardown of Unitree G1 shows that the Bill of Materials (BOM) cost of this complete machine totals about 41,600 RMB, of which the core joint cost is 27,500 RMB—even so, the gross margin of the basic version remains as high as 40.7%, and the gross margin of the EDU version even breaks through 60%. This means that the price reduction is not "selling at a loss to gain attention," but rather a proactive concession driven by cost reductions.
The four major engines for cost reduction are indispensable. First, comprehensive domestic substitution of the supply chain: the three core components—reducers, servo systems, and controllers—account for over 70% of the total cost. As of May this year, the localization rate of core components has risen to 75% to 90%, and the prices of domestic components are 30% to 50% lower than similar products from Japan and Germany. Second, the architecture shifts from "material stacking" to "pragmatism": cutting redundant configurations and focusing on real scenario needs. Third, "cross-boundary feedback" from the NEV (New Energy Vehicle) industrial chain: motors, reducers, sensors, and other links are highly reused with the EV supply chain, and the manufacturing base is directly transformed into a cost advantage. Fourth, the explosion of economies of scale: Gaogong data shows that China's humanoid robot shipments in 2025 were 14,400 units, and the production in 2026 is expected to scale across the 100,000 to 200,000 unit level.
The cost gap between the Chinese and U.S. supply chains is clear at a glance. In 2025, relying on China's mature local supply chain, the BOM cost of Optimus was about $46,000, while completely bypassing the Chinese supply chain would be as high as $131,000—a gap of nearly three times, with costs instantly dropping by nearly 70%.
Cost reductions are rewriting the "economic account" of robots. A humanoid robot can work 20 hours a day, with an effective cost of about $3 to $5 per hour including depreciation and maintenance, while the labor cost in first-tier cities is $8 to $12 per hour—the critical point where "robots are cheaper than humans" has arrived. Industry surveys show that 40% of the surveyed enterprises consider 100,000 to 200,000 RMB as the widely adopted price range, and the era of "20,000 RMB-level" civilian standard products for humanoid robots is approaching.
05. Industrial Chain Teardown: Where is the Value? The industrial dividends brought by the year of delivery are not evenly distributed, but are transmitted in layers from top to bottom along the industrial chain.
From the perspective of the complete machine cost structure, the execution system is the absolute core of value. According to Wanlian Securities' cost teardown of Optimus V3: rotary actuators account for 16.1%, linear actuators for 23.0%, dexterous hands for 19.7%, and multimodal sensors for 13.2%—actuators and dexterous hands together account for nearly 60% of the complete machine cost, representing the dual core of value and technology. From the component perspective, bearings account for 24%, cameras for 22%, force/tactile sensors for 15%, reducers for 15%, motors for 8%, and ball screws for 5%.
Technology routes are converging. The electric drive solution of motor + reducer + ball screw has become the mainstream consensus. Rotary actuators, linear actuators (planetary roller screws), and dexterous hands have become the most certain positioning links; dexterous hands, electronic skin, and cooling modules are still in the iteration period. Against the backdrop of route convergence, domestic manufacturers, relying on their manufacturing base and cost advantages, are achieving the leap from "substitution" to "leadership"—the fact that the top six global shippers in 2025 were all Chinese enterprises is proof of this.
The benefit logic and certainty gradient of the upstream, midstream, and downstream of the industrial chain are clear:
Core components (strongest certainty). In the reducer field, harmonic reducers (Leaderdrive), RV reducers (Shuanghuan Drive, Greatoo Intelligent), and planetary reducers (ZD Leader, the largest reducer supplier for Unitree, signed a 3.2 billion RMB three-year fixed-point order) constitute the three major routes; in the ball screw field, planetary roller screws, as the core of linear actuators, are being rapidly laid out by Wuzhou Xinchun, Hengli Hydraulic, Best Group, etc.; in the motor field, frameless torque motors (Wolong Electric provides 60% of Unitree's frameless joint motors) and coreless motors are scaling up in parallel; in the sensor field, six-axis force sensors (Keli Sensing), flexible electronic skin (Hanwei Electronics), and 3D vision (Orbbec, the top supplier accounting for 72% of Unitree's vision procurement) are blooming at multiple points.
Complete machine manufacturing (most direct valuation re-evaluation). OEMs such as Unitree Robotics, UBTECH, Agibot, and ZQ Robotics are the creators of the valuation anchor, but the competitive landscape in the complete machine segment is not yet settled, and the pace of profit realization depends on the delivery ramp-up.
Software and embodied AI (highest elasticity, lowest certainty). iFLYTEK's Spark large model is adapted for Unitree's humanoid robots, and the value of embodied AI software partners is yet to be re-evaluated; however, the Gaogong Industry Research Institute points out that software remains the biggest bottleneck at present—embodied large models require massive industrial task training data, while datasets for most scenarios have not yet been established, and cross-task generalization capabilities are still ramping up.
Order data has begun to validate the prosperity of components. Shenzhen Lingqiao Drive Control's on-hand orders in the first half of 2026 exceeded 1 million units, an increase of about 8.3 times compared to the 120,000 units ordered throughout 2025; Estun's net profit in the first half increased by over 20 times year-on-year, Topstar's net profit increased by 213% to 300% year-on-year, and Leadshine's net profit increased by over 55% year-on-year—the performance improvement of the entire industrial chain is completely transmitted from top to bottom.
06. Bull-Bear Divergence and Risk Warning Although the narrative of the "year of delivery" is materializing, the bull-bear divergence surrounding humanoid robots is equally intense.
Zheshang Securities data shows that, according to its compiled humanoid robot index, the PE percentile over the past 5 years is only about 7%, at a historical bottom; the sector's turnover has rebounded to 4.63% of the entire market, reversing the downward trend in July. Institutions generally judge that August to October remains the best trading window for the robot sector: in the short term, looking at the event catalysts of the World Robot Conference and Unitree's listing; in the medium term, looking at the validation of Optimus' mass production ramp-up. On the demand side, Zheshang Securities proposes the resonance of four scenarios: companionship for elderly living alone, children's companionship, emotional support for urban young people living alone, and pet substitution, citing Huatai Securities' calculations that the global AI companionship market has a CAGR of 31.24%, and the business model is leaping from hardware to "emotional SaaS."
The bearish and cautious camps, however, focus on the "narrowing of validation metrics." Production, delivery, paid deployment, and customer repurchase are sequentially narrowing metrics: currently, the proportion of productive paid deployments truly used in manufacturing terminals remains low, and the project's economics need to be continuously validated by operation and repurchase.
More critical are capability shortcomings—although manufacturers report near-perfect success rates for dedicated tasks, the real-machine success rate for third-party general task tests remains low, and cross-task generalization capabilities need to be improved. The internal divergence within the sector is also striking: the total market capitalization of 258 A-share humanoid robot-related stocks is about 5.56 trillion RMB, with only 55 stocks rising and 203 falling since the beginning of the year; since July, multiple stocks such as Zhimingda, Ouke Yi, and Worldia have fallen by over 50% within the month. Valuations are undergoing severe divergence between "hot-topic chasing" targets and targets with real orders during the theme diffusion period.
From an industrial perspective, risk points are mainly concentrated in four areas. First, mass production ramp-up falls short of expectations: if the capacity expansion pace of complete machine and component manufacturers is slower than planned, delivery data may fall short. Second, scenario landing and data bottlenecks: embodied AI large models lack industrial task training data, and cross-task generalization capabilities are insufficient, constraining the scaling of paid deployments. Third, hardware cost reduction falls short of expectations: if cost reductions in links such as ball screws and sensors stall, it will delay the arrival of the critical point where "robots are cheaper than humans." Fourth, valuation volatility risk: the sector has already seen large short-term gains, and there is a possibility of a "sell the news" adjustment after events are realized; if overseas Optimus demonstrates effects or production progress below expectations, it may also constitute an emotional shock.
However, judging from current industrial signals, the underlying color of this round of market trend is fundamentally different from past theme speculation—it is built on real shipments, real orders, and real capacity ramp-ups. In the first half of 2026, the total financing in the embodied AI track reached 93.5 billion RMB, a 5-fold increase compared to the same period in 2025; in the first four months of 2026, global humanoid robot venture capital reached $3.08 billion, already exceeding the total for 2025. The multiple resonance of capital, policy, orders, and capacity is turning the "year of delivery" from a concept into a report card that can be verified quarter by quarter.
Overall, the investment logic for humanoid robots is shifting from "telling stories" to "calculating the details": whether shipments can be realized, whether orders can be repurchased, whether costs can continue to drop, and whether robots can endure night shifts on production lines. The true meaning of the year of delivery is not how sexy the concept is, but that every robot rolling off the production line must start creating real economic value. This long-distance race from "demonstration" to "delivery" has just fired the starting gun.
- End -