Will Yushu become the "DJI and BYD" of the robot era? Watch @ SemiAnalysis_ publish this article titled 'China's Unitree Will Dominate Global Robotics' with the subtitle' The Fastest Iteration Cycle of Next Generation Robots Will Bring Unprecedented Acceleration '. The core is that Yushu is replicating the path of Chinese hardware giants BYD and DJI, using low-cost core components, rapid iteration, supply chain depth, and researcher ecology to push humanoid robots from "laboratory toys" to "deployable labor tools". Disassemble this report 1. The biggest conclusion: Yushu may become the DJI/BYD of the robot era SemiAnalysis's main judgment is that Yushu is not an ordinary "robot concept company", but a developing Chinese hardware platform company. Three years ago, it was mainly a quadruped robot dog company, but now it has transferred the actuator, control, supply chain, and production experience accumulated from quadruped robots to humanoid robots. The Yushu G1 has entered the usable deployment phase and there are three new designs on the road, including products that are more directly benchmarked against Western humanoid robots. SemiAnalysis analogizes Yushu's path to two Chinese hardware giants: BYD controls the battery first before taking care of the entire vehicle; DJI controls the flight control first, then switches to the drone; Yushu first controls the actuators and then expands to quadruped robots, humanoid robots, and labor automation. This is what the report calls "self generating ecology, creating new markets, and then swallowing up the market". 2. The price curve is the first main line: Yushu hits the legged robot at an extremely low price SemiAnalysis has sorted out the price decline path of Yushu quadruped robots: Laikago was about $45000 in 2018, A1 was about $15000 in 2020, Go1 started at about $2700 in 2021, and Go2 started at about $1600-2800 today. SemiAnalysis believes that this is equivalent to a price drop of approximately 94% -96% for entry-level quadruped robots within six years, allowing Yushu to expand from the university/laboratory market to the consumer, developer, and industrial deployment markets. This is crucial. Yushu is not waiting for humanoid robots to be perfect before mass production; It first uses quadruped robots to cultivate real sales, real supply chain, and real manufacturing experience, and then migrates the same set of actuators, controllers, motion control, and production processes to humanoid robots. The report suggests that in H1 2024, humanoid robots will essentially be more like "quadruped robots standing up", while G1 will truly open up new markets. 3. The significance of G1: It is not the strongest robot, but a platform that can be bought, cheap enough, and researched Before mid-2024, there will be very few humanoid robots available for purchase, research, and secondary development worldwide. Agility's Digit has only entered the factory in small quantities, Apptronik's Apollo is still relatively commercialized, Figure's cooperation with BMW is still in single digits/early deployment, and Tesla Optimus has not been sold externally. In contrast, the Yushu G1 entered the market at a price range of $30000 to $50000, directly becoming a platform that universities, AI laboratories, and robotics research teams can buy. The result of this is that G1 may not necessarily be the "strongest performing" humanoid robot, but it may be the platform with the strongest researcher ecosystem, the largest number of real machines, and the fastest iteration feedback. The report also mentioned that companies such as Nvidia, Apple, Meta, etc. have purchased hundreds of G1 for AI research on humanoid robots. 4. Core technology point: The actuator is the cost and performance bottleneck of humanoid robots The most important technical analysis of this report falls on the actuator. SemiAnalysis believes that actuators account for 50% -70% of the BOM of humanoid robots, which are the core components responsible for motion, load-bearing, and torque output in robot joints. The route chosen by Yushu is QDD, quasi direct drive, and collimated drive actuator, which can be roughly understood as "larger brushless motor+lower reduction ratio planetary gearbox". Traditional robots commonly use harmonic reducers or high reduction ratio solutions, which have high precision and strong load-bearing capacity, but high cost, difficult manufacturing, and long supply chain. The benefits of QDD are simpler structure, higher efficiency, lower cost, more suppliers, and faster iteration speed; The comparison given in the report is that the efficiency of QDD can reach 95% -98%, while the strain wave/harmonic route is roughly 85% -90%, and the cost of QDD can be reduced by up to 80%. But QDD also comes at a cost: a low reduction ratio means that the motor itself has to bear more torque, higher current, and more severe heating. The problem with the early G1 was severe overheating, with only a few seconds of holding a 2kg object with an extended arm and only a few minutes of holding a 2-3kg object with a bent arm, followed by a long period of cooling. This is also the reason why the market used to think that Yushu was "cheap but unreliable". 5. The report suggests a key turning point: Yushu's QDD is shifting from "cheap but hot" to "cheap and usable" SemiAnalysis believes that in the past two years, Yushu has made multiple rounds of improvements to the actuator, including reducing motor current requirements, lowering winding resistance, optimizing magnet and stator slot shapes to reduce cogging torque and torque ripple, increasing copper filling rate, and adding heat dissipation design to key joints. The report specifically mentioned that Yuki used vapor chambers at the knees and later added active cooling near the pelvis, which increased the heat margin. The result is a significant improvement in the availability of G1. The report states that G1 can now carry 5 kilograms in a bent arm state and last for 10-15 minutes, which is about twice the load and five times the duration compared to earlier models; Reaching 5 kilograms with fully extended arms can last for about 1 minute. For truly lightweight tasks such as warehousing and turnover box transfer, this is approaching the threshold of "deployable". SemiAnalysis also mentioned that in some 2-4 kg tote transfer tasks, the throughput performance of Yushu G1 can be close to that of Agility Digit; However, it is still not as robust as high-end Western robots, typically requiring 5-10 minutes of cooling after 10-15 minutes of operation, corresponding to a utilization rate of approximately 50% -67%. Nevertheless, under conservative assumptions such as full remote operation, 15% service contract, two-year lifespan, zero residual value, and two shift system, it already has economic feasibility in certain commercial scenarios.