Futures
Access hundreds of perpetual contracts
TradFi
Gold
One platform for global traditional assets
Options
Hot
Trade European-style vanilla options
Unified Account
Maximize your capital efficiency
Demo Trading
Introduction to Futures Trading
Learn the basics of futures trading
Futures Events
Join events to earn rewards
Demo Trading
Use virtual funds to practice risk-free trading
Launch
CandyDrop
Collect candies to earn airdrops
Launchpool
Quick staking, earn potential new tokens
HODLer Airdrop
Hold GT and get massive airdrops for free
Launchpad
Be early to the next big token project
Alpha Points
Trade on-chain assets and earn airdrops
Futures Points
Earn futures points and claim airdrop rewards
A trillion-dollar industry’s cost reduction challenge: robot component manufacturing process difficulties, bearing prices halved but still have a 30% gross profit margin
(Source: Finance News)
On March 20, the Shanghai Stock Exchange accepted Yushu Technology’s IPO application for the Sci-Tech Innovation Board, with the latter officially aiming to become the “first stock of embodied intelligence” on the board.
This company, which is projected to have the highest shipment volume of humanoid robots globally by 2025, despite its gross margin increasing from 44.22% to 60.27% over the past three years, still lists “the risk of high revenue growth and declining gross margin” as one of the risks in its prospectus.
As a core component of the robotics industry, the price level of parts directly determines the overall cost, with joints valued at 60,000-80,000 yuan, dexterous hands priced close to 100,000 yuan, and electronic skin that only covers one hand exceeding 10,000 yuan… all of which are questioning the cost reduction capabilities of this trillion-dollar industry.
Craftsmanship and scale both constrain robot costs
Upstream in the robotics supply chain, bearings are the core components that determine joint flexibility and durability.
On March 18, at the AI Show 2026 Beijing International Artificial Intelligence Application and Robotics Innovation Expo, a product manager from Lanyu Bearings introduced to Finance News that a reducer requires four to five bearings, including input ball bearings, flexible bearings, and output crossed roller bearings, with a complete set costing about 200-300 yuan.
According to the aforementioned product manager, the production of crossed roller bearings at Lanyu Bearings is entirely manual, without using mold pressing or automation processes. Due to the complex processing of structural parts, which requires dozens of machines to complete over thirty processes, the costs remain high.
This means that there is a significant gap between high-precision, small-batch customized production models and the industry’s expectations for large-scale cost reduction.
“If humanoid robots reach a high level of maturity, then their prices will definitely not be high,” he candidly stated, noting that currently, robots bought can only be used as exhibits and have no practical application. Cheaper robots cost around 100,000 yuan but are inconvenient to walk, while robots with some autonomy can cost four to five hundred thousand yuan per unit, with little increase in hardware, only stronger AI capabilities.
If bearings are the “muscles” of humanoid robots, then controllers are their “small brains.”
Pan Haobo, the robotics business line manager at Xunbu Technology, compared that while a home computer motherboard only needs to connect fewer than ten peripherals like monitors, power supplies, memory, and mouse-keyboard, industrial environments, especially controllers for robots, require a large number of motors, sensors, and other peripherals, with some controller motherboards needing waterproof functionality, making the price naturally incomparable.
It is reported that Xunbu Technology has 34 years of experience in industrial control motherboard R&D, covering solutions for various platforms including X86, ARM, and domestic products, with the highest shipment volume in China. However, humanoid robots are currently in their initial stage, characterized by a multitude of varieties, small batches, and rapid iterations, with a single company selling only a few hundred units a year.
“The company can leverage the massive shipment volumes from other industries to backfill, and there are also some standardized industrial control motherboards that can be directly used for the ‘small brain’ part of robots, but it mainly depends on whether the installation space for the robot’s head is limited and whether the specifications meet the requirements; essentially, there is not much difference, each taking what they need,” Pan Haobo said, noting that from a cost perspective, the rising prices of memory chips have become a trend, and the company does not have the capability to produce memory chips and can only purchase from the market, thus being greatly affected.
On the other hand, Pan Haobo mentioned that for robots to run AI, the higher the functionality and the more complex the algorithms, the larger the memory required at the start. Although the memory demand for robots is far less than that of AI servers, the small demand becomes a disadvantage—fewer robots in the market naturally lead to weaker bargaining power.
Yushu Technology’s procurement of mechanical components has increased
It is noteworthy that the lack of convergence in the technical routes for robotics is also one of the reasons for the high costs of components.
From Yushu Technology’s prospectus, one can glimpse the pace of technological iteration in robotics. The company disclosed that in 2023, it launched its first self-developed humanoid robot H1, followed by the release of the medium-sized humanoid robot G1 in 2024, and then R1 and H2 in 2025, indicating a rapid product iteration speed.
This rapid iteration enhances product performance but also puts pressure on the supply chain. The prospectus shows that the decline in the average selling price of humanoid robots since 2024 mainly comes from two factors: product structure and technological advancements.
Among them, G1, as a medium-sized humanoid robot, has a cost and product price lower than the full-sized H1, with its sales growth driving the overall average selling price of the company’s humanoid robots down. Regarding technological advancements, continuous improvements in process technology and optimization of cost structures have also led to a steady decline in the unit cost of the company’s humanoid robots.
However, from the cost composition perspective, the proportion of direct materials remains high. From January to September 2025, the proportion of direct materials in Yushu Technology’s main business costs was 80.12%, a significant increase from 72.62% in 2024. The company explained that the rapid expansion of the humanoid robot business and the increase in its revenue share, due to the high material cost proportion of this business, has raised the overall proportion of direct materials in the company’s main business costs.
Regarding suppliers, the prospectus shows that the procurement amount from the top five suppliers accounted for 18.77%, 22.07%, 26.44%, and 21.72% of the total raw material procurement amount for the period, indicating a relatively low concentration. The procurement mainly includes mechanical components, electronic components, and electrical materials. Among these, the proportion of mechanical component procurement has risen from 40.13% in 2022 to 47.88% in January-September 2025.
According to the aforementioned product manager from Lanyu Bearings, since the beginning of this year, the price of reducers has halved, bringing down bearing prices by more than half as well, but the processing difficulty has not decreased, nor has processing efficiency improved. Currently, for bearings used in humanoid robots, it takes over two months to process 1,000 sets, with a gross margin around 30%, while ordinary bearings can be processed in just one month with a gross margin of only 10%-20%.
In this regard, Guolian Minsheng Securities pointed out in a research report on March 13 that the current technical routes for dexterous hands in humanoid robots have not yet fully converged, with multiple technical routes still running in parallel in the four core areas of degrees of freedom design, drive systems, transmission methods, and perception technologies, further exacerbating uncertainties in the upstream supply chain.