SemiAnalysis’s analysis of Unitree’s G1 robot revealed that, at a selling price of $27,300, its material cost is only $9,000, resulting in a gross margin of 67%. Unitree achieves a price advantage 3 to 5 times lower than comparable Western products by developing its own actuators, vertically integrating its supply chain, and leveraging China’s electric vehicle and drone industry ecosystem. Its development path mirrors that of BYD (self-developed batteries) and DJI (self-developed flight controllers): first mastering core components, building actuator and supply chain capabilities through quadruped robots, and then transferring these advantages to humanoid robots. Currently, approximately 250 G1 units have been deployed in real production environments, with an hourly operating cost of about $26.40 in light material handling scenarios—below the U.S. labor cost threshold of $30 per hour.

Article author and source: Silicon Base Observation Pro

Unitree is just one step away from a breakthrough.

In the humanoid robotics industry, Unitree has always been a controversial presence. Previously, SiliconBase published a series of articles: "Unitree Technology: The Strangest IPO Company in History—Don’t Be Fooled by Its 600 Million Profit! The Most Profitable Hardware Is Also Unitree’s Greatest Risk."

Supporters view it as the BYD and DJI of robotics, leveraging extreme cost control to bring technologies once confined to laboratories to the mass market.

Critics argue that Unitree’s low price comes at the cost of performance. After all, while humanoid robots from overseas manufacturers like Figure and Agility typically sell for over $100,000, Unitree has already reduced the G1’s price to $27,300—a price difference of as much as three to five times.

This raises a core question: Is Unitree’s “affordability” a compromise on quality, or an unreplicable structural advantage?

Recently, SemiAnalysis did something: they meticulously broke down the BOM for each component of the G1, quoted prices from each part’s manufacturer, and cross-verified with multiple supply chain buyers and sellers to try to answer this question.

After analyzing Unitree, SemiAnalysis found that Unitree’s greatest advantage may not be the robot itself, but rather an industrial ecosystem mirroring the rise paths of BYD and DJI.

Please note that while SemiAnalysis holds significant influence in the semiconductor, AI infrastructure, and hardware supply chain sectors, it has also faced controversy, including allegations and legal actions regarding potential conflicts of interest.

Why is the gross profit margin still 67% when it's sold so cheaply?

Affordability is Unitree’s most important competitive advantage.

Over the past year and a half, Unitree has reduced the pre-tax price of the G1 from over $50,000 to approximately $27,300.

More surprisingly, even at this price, SemiAnalysis estimates that the G1’s gross profit margin still reaches 67%.

To verify this, SemiAnalysis dismantled the full bill of materials (BOM) for the G1 EDU Advanced version, obtained itemized quotes from suppliers, and cross-verified with multiple buyers and vendors.

The final results show that the core material cost of this humanoid robot, priced at $27,300, is only about $9,000.

In other words, Unitree is not only one of the most affordable humanoid robot manufacturers globally today but has also achieved considerable profitability.

With increased production, Unitree has already lowered the price of the G1 below $20,000 on some orders, while similar Western products continue to sell for over $100,000.

Why can Unitree achieve this? SemiAnalysis provides three reasons.

First, it is the scale advantage of China's supply chain.

Over the past two decades, China has built the world’s largest new energy vehicle and consumer electronics supply chain. In 2024 alone, China’s new energy vehicle production exceeded 30 million units. Meanwhile, the DJI ecosystem has spurred the emergence of thousands of drone component suppliers.

Large quantities of brushless motors, drivers, encoders, batteries, and precision manufacturing processes, originally developed for the automotive and drone industries, can now be directly reused by the robotics industry.

For Unitree, building humanoid robots isn't about creating a supply chain from scratch—it's about assembling components on top of an established industrial ecosystem.

Second, it is Unitree's own highly advanced vertical integration capability.

Many robotics companies opt to purchase off-the-shelf components for assembly. However, Unitree has consistently focused on in-house research and development of core components since its founding, having independently designed and manufactured key parts such as brushless DC motors, planetary gearboxes, LiDAR, and depth cameras.

This means Unitree not only has lower costs but also faster iteration capabilities. For example, its in-house developed motors cost only 30%-40% of some Western counterparts, and its planetary gearboxes are also among the lowest in the industry.

Even within China’s robotics industry, Unitree’s level of vertical integration is clearly ahead. Many competitors are still catching up to work it has already completed.

Third, Unitree is reverse-engineering the supply chain.

As sales grow, an increasing number of suppliers are developing complementary products based on Unitree’s technology roadmap and specifications. China has now established a new ecosystem comprising hundreds of robotics-related companies and numerous core component manufacturers.

These suppliers serve not only Unitree but the entire robotics industry. As the industry scales up, supply chain costs will continue to decrease.

The victory of the "BYD and DJI" model

In SemiAnalysis’s view, Unitree’s growth trajectory resembles a hybrid of BYD and DJI.

These two companies appear to belong to entirely different industries, but their rise follows a highly consistent logic:

First, master the most expensive and challenging core components in the cost structure, then enter a market that seemed small at the time but was willing to pay, leverage economies of scale to continuously reduce costs, and finally use each generation of products to open new markets.

BYD controls the battery.

Batteries account for 30%-40% of an electric vehicle's cost and are the most critical core component. BYD has continuously expanded upstream and downstream around battery technology, gradually integrating key elements such as motors, electronic controls, IGBTs, and SiC power modules into its in-house R&D system.

This vertical integration brings not only cost advantages but also continuous innovation capabilities.

After the launch of the Blade Battery in 2020, lithium iron phosphate batteries gained the ability to directly compete with ternary lithium batteries for the first time. In the following years, BYD's sales entered a period of explosive growth, ultimately becoming one of the world's largest new energy vehicle manufacturers.

DJI controlled the flight control system, which served as the drone's brain and was one of the most expensive and technologically demanding core components at the time. Third-party flight controllers cost as much as $200 to $400, and they were precisely what determined whether a drone could fly stably.

DJI chose to tackle this toughest challenge first.

After mastering flight control, DJI gradually expanded into core components such as gimbals, motors, and electronic speed controllers, continuously reducing costs by leveraging the economies of scale of Shenzhen’s consumer electronics supply chain.

This capability was quickly reflected in the product. When the Phantom 1 was released in 2013, it wasn’t perfect: it had no camera, no gimbal, and only 10 minutes of battery life.

At the time, users either had to spend thousands of dollars on professional equipment or build their own drones. In comparison, the Phantom 1, priced at $679, significantly lowered the barrier to entry despite its limited performance.

It initially attracted geeks, researchers, and drone enthusiasts. DJI then continuously iterated its products: the Phantom 2 opened the commercial aerial photography market, and the Phantom 4 entered enterprise markets such as surveying and inspection.

Each generation of products unlocks new demand, and continuously growing sales further reduce costs, ultimately creating the strongest scale advantage in the drone industry.

SemiAnalysis believes that Unitree is replicating the same path today.

Just as BYD chose batteries and DJI chose flight controllers, Unitree set its sights from the start on the most critical component of robotics—actuators.

Actuators serve as the joints and muscles of robots, accounting for 50%-70% of the cost structure of humanoid robots and representing one of the most technically challenging components.

But Yutu did not start with humanoid robots; instead, it chose to begin with quadruped robots.

The reason is simple: at the time, university labs, research institutions, and robotics developers consistently lacked access to affordable legged robot platforms, while similar overseas products typically cost over $70,000 to $100,000.

Unitree captured this seemingly small market through low pricing.

In 2018, the Laikago was priced at approximately $45,000;

In 2020, A1 dropped to $15,000;

In 2021, Go1 further declined to the thousands of dollars range.

In just a few years, the entry price of quadruped robots has dropped by more than 90%, enabling robots that were once confined to laboratories to enter the developer and consumer markets at scale for the first time.

More importantly, during this process, Unitree accumulated expertise in actuators, motion control, supply chain management, manufacturing processes, and real user feedback.

These capabilities were later almost entirely transferred to humanoid robots.

The product that truly established Unitree's foothold in the humanoid robotics field is the G1, released in 2024.

Before G1, there were virtually no humanoid robots on the market that could be purchased in bulk. Companies like Figure, Apptronik, and Agility are still in early deployment stages, with high product prices and low shipment volumes.

G1 was the first to bring humanoid robots into the tens of thousands of dollars range and achieve mass production, directly opening up a market that previously barely existed—the humanoid robotics research market.

For university labs, AI research institutions, and robotics developers, a humanoid robot priced between $30,000 and $50,000 that is immediately available for purchase means the barrier to entry for research has been significantly lowered.

As a result, G1 quickly became one of the most widely used humanoid robot research platforms globally.

The goal of the next-generation product is to truly enter industrial deployment and commercial applications.

From this perspective, Unitree is following the same business path that BYD and DJI once took:

Master the core components, start in a niche market, continuously reduce costs through economies of scale, and use each generation of products to enter larger markets.

High-stakes bet on QDD

Unitree's development was not without challenges. When BYD's Blade Battery and DJI's flight control systems entered the market, they already had clear performance advantages. However, Unitree's early humanoid robots did not.

Both the H1 and the original G1 model have a noticeable issue— they tend to overheat.

For example, after fully extending, the G1 robotic arm can only lift approximately 2 kilograms for a few seconds before requiring throttling or even forced cooling. After completing short bursts of work, the robot often needs tens of minutes to return to normal operation.

This means that the robot's actual efficiency in performing real tasks is not very high.

The root of the issue lies in Unitree's choice of core executor roadmap.

Unlike most humanoid robot companies, Unitree has opted for quasi-direct drive (QDD) instead of traditional high-reduction-ratio actuators.

In simple terms, traditional robots are more like “small motor + large gearbox.” The motor itself produces limited output, and force is amplified through a complex reducer, resulting in stable performance but at the cost of higher price, greater weight, and more complex structure.

QDD, on the other hand, takes the opposite approach. It uses a "large motor + small gearbox" design, delivering power directly through a more powerful motor and reducing reliance on complex reducers.

The biggest advantage of this solution is its low cost. Actuator costs can be significantly reduced, and efficiency is also improved. However, the drawbacks are equally apparent: the motor bears a greater load, is more prone to heating, and demands extremely high standards for control algorithms and thermal design.

Many industry professionals initially believed this approach could not support true humanoid robot applications.

To some extent, Unitree bet on a technology path in the robotics industry that had not yet been proven, and the overheating issues in its early products confirmed that market concerns were not unfounded.

But Yu Shu did not look back.

Over the past few years, the company has continuously iterated on QDD, enhancing actuator performance through optimizations such as improving motor structure, increasing copper fill factor, reducing energy losses, and enhancing the cooling system.

Over time, the advantages of the QDD route began to emerge.

First is cost.

The low reduction ratio planetary gearbox used by QDD is essentially a mature industrial standard component that can be manufactured using standard hobbing equipment, with numerous suppliers available.

In contrast, harmonic drives still commonly used in humanoid robots are typical high-precision components requiring more than ten complex processes, including heat treatment, precision hobbing, and micron-level machining.

The global leader Harmonic Drive spent decades building a complete manufacturing system, and even China’s leading player, Leader Harmonic, is still actively catching up.

This process difference is directly reflected in cost. The QDD solution achieves a transmission efficiency of 95%-98%, higher than the 85%-90% of harmonic drives, while reducing costs by up to 80%.

For humanoid robots, where actuators account for 50% to 70% of the total cost, this cost difference can determine whether the product has the potential for large-scale adoption.

In addition to cost advantages, QDD also enables faster iteration speeds.

Traditional high-precision reducers rely on complex supply chains, often requiring months from design modifications to sample delivery. In contrast, QDD extensively uses mature industrial components, resulting in a more open supply chain and simpler structure.

This means Unitree can complete an actuator iteration in just a few weeks, while many competitors may need months. In a new industry still rapidly evolving, the speed of iteration itself is a competitive advantage.

The results show that Unitree bet on the right direction.

As the product continues to iterate, the reliability of the QDD actuator has steadily improved, leading an increasing number of Chinese humanoid robotics companies to adopt similar technical approaches. Unitree, leveraging years of accumulated expertise, has established a first-mover advantage in actuators, supply chains, and manufacturing systems.

Has surpassed the break-even point for labor costs

The SemiAnalysis report concludes with an important judgment:

Unitree is approaching the tipping point for the commercialization of humanoid robots.

Although G1 still has many issues, it can still handle certain scenarios, such as light material handling.

This job involves moving boxes weighing 2-5 kilograms from point A to point B in a warehouse, factory, or logistics center.

This type of work has several characteristics: low workload, repetitive motions, high tolerance for errors, and minimal requirement for dexterity—all of which fall precisely within the current capabilities of the G1.

SemiAnalysis's research found that, in addition to the research and developer markets, Unitree may have delivered approximately 250 humanoid robots to real-world production scenarios by 2025.

Some companies have already deployed dozens of G1 units and have begun handling real-world logistics tasks.

This means that humanoid robots are transitioning from the laboratory validation phase to the commercial validation phase.

To evaluate its economic feasibility, SemiAnalysis conducted calculations using Agility Robotics' Digit robot as a reference.

Digit currently serves as a bridge between automated systems in logistics scenarios, such as transporting bins delivered by AMRs (autonomous mobile robots) onto conveyor belts.

Previously, these tasks required human workers to wait long periods for automated equipment to be installed before performing simple搬运. Now, with robots taking over, these repetitive tasks can be performed continuously.

Based on field deployment feedback, the actual throughput of the G1 for tasks involving moving pallets weighing 2-4 kg has approached that of the Digit. Although the G1 typically requires a 5-10 minute cooldown after 10-15 minutes of continuous operation, overall equipment utilization still reaches 50%-67%.

On this basis, SemiAnalysis conducted a rather conservative calculation:

Assume the robot is 100% dependent on remote operation; remote operators earn $11 per hour; annual service fees amount to 15% of the equipment's value; the lifespan is only two years; salvage value is zero; the robot operates only two shifts per day.

Even so, at a price of $35,000 and a utilization rate of 67%, the combined hourly cost of the G1 is approximately $26.40.

This figure is now below the U.S. labor cost of approximately $30 per hour.

If future reliability improves further, such as achieving a mean time to failure (MTTF) of 33 minutes and increasing equipment utilization to 80%, the hourly cost could decrease further to $24.60.

SemiAnalysis repeatedly emphasizes that these figures are still in a very early stage. In terms of software systems, warehouse management integration, security infrastructure, and autonomous capabilities, Agility is clearly ahead of Unitree.

But that’s not important. What matters most is that Unitree has begun approaching a critical threshold:

Even under the most conservative assumptions, the economic viability of humanoid robots in certain warehouse scenarios is beginning to approach or even cross the break-even point with human labor costs.

For the entire industry, this could be the first real step toward humanoid robots performing “useful work.”