Elon Musk's robot plan can no longer be hidden.
During Tesla's Q3 earnings call, Elon Musk hardly talked about cars and instead shifted the narrative focus entirely to the humanoid robot Optimus. He announced that Optimus V3 will be released in Q1 2026, and plans to launch a production line with an annual capacity of 1 million units by the end of next year.
Behind Musk's vision of the robot future lies a fierce global race in the field of embodied intelligence. In the first half of 2025, there were 141 investment and financing events in the Chinese market, with 51 single - deals exceeding 100 million yuan. The total financing amount has more than tripled that of the whole last year.
Previously, the outside world's perception of Optimus remained at the level of "cutting - edge technology exploration". Even though the second - generation model demonstrated more flexible movement capabilities at the World Robot Conference in early 2025, most analysts still set the mass - production schedule after 2026.
However, Musk's statement broke this wait - and - see attitude. The annual production capacity target of 1 million units is backed by Tesla's experience in supply - chain management and large - scale production in the electric - vehicle field, as well as the improved AI training efficiency brought by the Dojo supercomputer.
More importantly, when the production volume exceeds one million units, the unit price of Optimus is expected to drop to $20,000. This price will completely break through the entry barriers of the consumer - grade and industrial - grade markets, just as the Model 3 reshaped the electric - vehicle market back then, and kick off the popularization cycle of humanoid robots.
In other words, the underlying logic of this upheaval is driven by both technological breakthroughs and market demand.
Among international giants, Tesla's Optimus has completed the iteration from the prototype to the mass - production model. Figure AI has implemented core processes in the BMW factory. OpenAI has achieved the scenario transformation of cognitive - layer technology through investing in hardware carriers. In the domestic market, the competition is also fierce. Unitree has released the new - generation humanoid robot R1, and Zhipu has launched an embodied - intelligence solution leveraging its large - model advantage. It has raised funds in 11 rounds in the two years since its establishment, with a total of billions of yuan.
01
Why are humanoid robots so popular?
"In the past three years, the enthusiasm for embodied - intelligence startups has been rising almost every six months. Now, it has reached a critical turning point from the technical concept to commercial value." An analyst who has been continuously tracking the robot industry told Xinmou.
If we take Tesla's first release of Optimus in 2022 as the starting point, the embodied - intelligence transformation in the past three years can be clearly divided into three stages. The technological breakthroughs and player layouts in each stage have jointly promoted the "breakout" of humanoid robots.
Specifically, from 2022 to 2023 was the period of breakthrough in motion control, with the core feature of "making the robot stand and walk steadily". During this stage, the industry's technological focus was on optimizing the mechanical structure and realizing basic movement capabilities. The players in this stage were mainly technology - based startups and tech giants. Most of the financing was concentrated in the seed and angel rounds. The industry consensus was to "first achieve technological feasibility and then talk about commercialization".
From 2023 to 2024, it entered the stage of multi - modal perception fusion, with the core breakthrough of "making the robot see and understand". With the maturity of multi - sensor technology and the iteration of AI algorithms, humanoid robots began to have the ability to interact with the environment. During this stage, the number of industry players increased significantly, and the financing rounds were concentrated in Series A. Enterprises began to conduct small - scale scenario tests, and industrial assembly and logistics handling became the preferred fields.
From 2024 to 2025, it was the critical period of cognitive awakening, with the core change of "making the robot think and make decisions". Especially the commercialization of ultra - large - scale models such as GPT - 4.5 provided strong cognitive support for embodied intelligence. The typical feature of this stage was the explosive growth of the industry's financing scale, with frequent large - scale financings. International giants and domestic leading enterprises began to layout mass - production plans, and the competition in the field shifted from "technological competition" to "commercialization race".
In this global race, there are significant differences between the domestic and international embodied - intelligence markets. In terms of technological routes, international players pay more attention to "full - stack self - research" and "in - depth scenario exploration". Tesla relies on its technological accumulation in the automotive field to vertically integrate the FSD chip, vision system, and motion algorithms. Figure AI focuses on industrial scenarios and has reached in - depth cooperation with BMW to iterate products through practical applications.
In contrast, domestic players show the characteristics of "single - point breakthrough + ecological collaboration". They have continuously achieved technological breakthroughs in core components such as dexterous hands and sensors, but the overall - machine integration ability still needs to be improved. Most enterprises choose to cooperate with downstream scenario providers to open up the market through "customized solutions".
Behind this difference are the multiple effects of technological accumulation, industrial foundation, and market demand.
International giants, with years of experience in AI, automobile manufacturing, and industrial automation, have the ability to integrate the entire industrial chain. Moreover, overseas industrial scenarios have a more urgent demand for high - precision robots and are willing to pay a premium for new technologies. The domestic market, however, is limited by issues such as dependence on imported core components and inconsistent standards, resulting in high costs for the whole machine. At the same time, the consumer - grade market is price - sensitive, which prompts enterprises to prioritize technological breakthroughs in niche scenarios.
In addition, policy orientation also affects the market trend. Domestic policies to support embodied intelligence focus more on "industrial - chain cultivation", while overseas policies emphasize "technological innovation and free commercial competition". This difference further shapes the layout logic of domestic and international players. Whether it is the full - stack layout of international giants or the single - point breakthrough of domestic enterprises, they all point to the same conclusion: the technological threshold of humanoid robots is rapidly decreasing, and their commercial value is gradually emerging, which is the core reason for their continuous popularity.
02
What kind of grand plan is Musk plotting?
Every move of Tesla's Optimus affects the industry's nerves. Behind this is not only the product's technological breakthrough but also Musk's strategic restructuring of Tesla's future.
Peeling off Optimus's technological shell, its uniqueness lies in that it is not an isolated robot product but the core carrier of Tesla's "AI + manufacturing" ecosystem and a crucial move for Musk to realize his vision of "bringing artificial intelligence into the physical world". Reviewing Optimus's development process, every key node hides strategic decisions.
When it was first released in 2022, Optimus's prototype moved stiffly and was even questioned by the outside world for being "technologically immature". However, Musk insisted on introducing it to the public. The core of this decision was not product display but "seizing the position". By locking in industry attention in advance, it attracted supply - chain resources and technical talents, and at the same time sent a signal to the market that Tesla was transforming from an "automobile manufacturer" to an "intelligent - agent company".
In 2023, Tesla migrated its FSD autonomous - driving technology to Optimus. This strategic reuse was like a "game - changing move". The powerful computing power provided by the FSD chip supports the robot to model the real - time environment, and the Autopilot vision system solves the perception problem. It not only reduces R & D costs but also accelerates technological maturity. This "technology migration" strategy has become Optimus's core competitiveness.
In 2024, Musk launched an expansion plan for the Dojo supercomputer, investing heavily to achieve parallel training of robots and compressing the new - skill learning cycle to 24 hours. This decision addresses the core pain point of mass production - AI training efficiency. Only through supercomputer empowerment can robots quickly adapt to different scenarios and lay the foundation for large - scale commercial use.
During the Q3 2025 earnings call, Musk proposed an annual production capacity target of 1 million units and a release plan for Optimus V3, marking a shift of the strategic focus from "technological iteration" to "commercialization". All this confidence comes from Tesla's experience in large - scale production and supply - chain management in the electric - vehicle field. In other words, Musk's expectations for Optimus have long exceeded the scope of "a single product".
He has publicly stated many times that in the future, about 80% of Tesla's value will come from the Optimus robot. Its ultimate goal is to create a "general - purpose humanoid robot" covering multiple scenarios such as industrial production, home services, and medical care. In Musk's vision, Optimus will become the next - generation intelligent terminal after personal computers and smartphones, reshaping the way humans interact with machines. Tesla will build a business closed - loop of "hardware + AI services" through large - scale deployment of robots, completely getting rid of its dependence on the electric - vehicle business.
It is worth noting that this expectation is not a castle in the air. A report from Morgan Stanley predicts that by 2050, the global humanoid - robot market will reach $5 trillion. If Tesla can capture 10% of the market share, the Optimus business alone can generate $500 billion in revenue, far exceeding the peak of the current electric - vehicle business.
As Optimus accelerates mass production, the embodied - intelligence market is at a "divergence point", and the path differences between different players are becoming clearer. Musk has chosen the path of "scale + generalization". Relying on Tesla's manufacturing advantages, it first achieves mass - production ramp - up in industrial scenarios to reduce unit costs, and then gradually penetrates the consumer - grade market, ultimately creating a general - purpose robot. The core of this path is "winning by scale", which essentially replicates the successful logic of the Model 3 in the electric - vehicle field.
In contrast, most domestic players choose the path of "niche scenarios + customization". Unitree focuses on industrial logistics and special - operation scenarios, optimizing the robot's load - carrying capacity and environmental adaptability to meet the rigid needs of specific industries. Zhipu takes the embodied large model as the core and provides AI solutions for traditional robot enterprises, avoiding the heavy - asset trap of whole - machine manufacturing.
The root cause of this path difference lies in resource endowments. Tesla has the world's top - notch manufacturing capabilities, supercomputer resources, and brand influence, and is capable of bearing the R & D and mass - production risks of general - purpose robots. Domestic enterprises, limited by issues such as high costs of core components and lack of manufacturing experience, find it more practical to make breakthroughs in niche scenarios.
03
Beware of the embodied - intelligence bubble
The popularity of humanoid robots reminds us of many past technology - industry trends. However, historical experience tells us that there are often bubbles behind the enthusiasm. From the hype of industrial - robot concepts around 2010, to the capital frenzy of service robots in 2018, and then to the speculation of metaverse robots in 2021, each trend was accompanied by a surge in financing and an influx of enterprises, but only a few players could survive the cycle.
Compared with previous trends, this wave of embodied - intelligence enthusiasm is larger in scale and faster in technological iteration, but the exposed problems are also more prominent.
Data shows that in 2025, 74% of the financing of domestic embodied - intelligence enterprises was concentrated in Series A and earlier stages. More than 80% of the enterprises have not achieved large - scale revenue. Most products are still at the stage of "prototype display" or "small - scale testing", and there is still a long way to go before a real commercial closed - loop is established.
Looking at the progress of embodied intelligence at home and abroad, the industry generally shows the characteristic of "technology leading business". Internationally, although Tesla's Optimus has set an annual production capacity target of 1 million units, as of Q3 2025, the trial - production scale is still less than 1,000 units. The pressure on the motor supply chain caused by export controls on rare - earth materials and the problem of AI algorithm adaptation in complex scenarios may delay the mass - production progress. Although Figure AI has made breakthroughs in the BMW factory, it only covers 4 core processes and relies on customized environmental transformation, making it difficult to quickly replicate to other scenarios.
The progress of domestic enterprises is even more in the early stage. Although Unitree's H1 model has improved its movement ability, it still needs to rely on large - model capabilities in terms of cognitive decision - making. Although Zhipu's embodied solution has been implemented in some scenarios, its commercialization ability still needs to be verified.
This contrast between "hot technology and cold business" stems from the misalignment of supply and demand.
From the demand side, the market's demand for humanoid robots is "high cost - performance + strong adaptability". Industrial scenarios require robots to reduce production costs, and consumer scenarios require robots to be affordable and easy to operate. However, from the supply side, current robot products generally have the problems of "high cost and weak adaptability". The cost of a humanoid robot is quite high, far exceeding the affordability of industrial enterprises, and the functions of consumer - grade products are difficult to meet diverse needs.
More importantly, there are currently no unified standards for key technological indicators such as the degree of freedom, perception accuracy, and AI training framework of robots in the industry. Enterprises working independently lead to a waste of R & D resources. Domestic robots have a high degree of dependence on imported core components such as high - end motors and precision sensors, which not only drives up costs but also poses supply - chain security risks. The blind influx of capital further amplifies the bubble. Some enterprises raise funds based on concepts alone and lack core technologies and implementation capabilities. Once the capital tide recedes, they will surely be eliminated by the market.
Beware of the bubble does not mean denying the industry's value but maintaining rationality. After all, the development of humanoid robots is a long - distance race, not a sprint. For enterprises, instead of chasing trends and hyping concepts, it is better to focus on product refinement and in - depth scenario exploration.
This article is from the WeChat official account "Xinmou" (ID: xinmouls), author: Tang Ning, published by 36Kr with authorization.