In the previous article, we analyzed the technological origins of embodied AI companies. If we want to know how these companies will develop subsequently, how they will plan their businesses, and whether their commercial potential is huge, how should we judge it? Money is the most reliable clue.
From algorithms and data to mass production and market promotion, every step forward in embodied AI burns a massive amount of real money. This process often lasts for five to ten years. So, where does the money come from? Whose money is it? This almost determines everything about an embodied AI company.
Unlike the academic faction with a slight ivory tower temperament, there is a type of embodied AI company that takes the smartest money and must venture into the most realistic business world—those backed by industrial capital.
Industrial capital refers to funds from real economy enterprises participating in equity investment. If government funds focus on strategy, and financial capital casts a wide net and bets on IPOs, then industrial capital not only invests funds but also provides business support to the invested companies.
In this wave of embodied AI, if an embodied AI company secures investment from internet giants, automakers, and home appliance giants, it not only signifies strong financial strength but also means it has gained more direct resource support and extensive application scenarios than mere financial investment. Consequently, the probability of successfully commercializing in the later stages is higher.
Through the flow of industrial capital, we can see through the surface and know which embodied AI companies the smartest money is betting on.
Deepening the Flow: Embodied AI Grows Upon Core Businesses Companies whose original businesses highly overlap with physical AI are the most natural industrial capital for investing in embodied AI.
Further extending their core businesses and reusing existing R&D experience, algorithms, and data in the field of embodied AI can improve marginal benefits, dilute R&D costs, and also make the parent company's financial statements look better.
Among them, Dreame, which makes robot vacuums, and Tesla, XPeng, and Chery, which make cars, are typical representatives.
In 2024, Magic Atom was established, and completed an angel round financing of 150 million RMB in December. The lead investor, Zhui Chuang Venture Capital, is exactly the robotics industry venture capital fund (CETA) under Dreame. Through its own investment platform, Dreame quickly built an embodied AI team. The founder, Wu Changzheng, was once responsible for the R&D of Xiaomi's first-generation robot dog CyberDog.
This kind of directly incubated embodied AI company is born with supply chain and mass production experience, coupled with ready-to-use application scenarios. Therefore, in November 2024, Magic Atom's humanoid robot "Xiaomai" entered Dreame's own factory to undertake tasks such as loading and unloading high-speed digital motors and dispensing glue. Producing and selling by itself is like keeping the meat in the pot, meaning both sides can profit.
However, the cost of independent incubation is that the core team is hard to have 100% control. The steering wheel is entirely in the hands of the funding parent company. Whether to go into the factory to screw bolts or go to the street to perform yangko dances, the team has little room for autonomy. Yet innovation highly depends on independent exploration. This creates a tension between the embodied AI company and the industrial capital. If not handled well, it will fall apart.
In September 2024, Yu Chao, the former head of Dreame's humanoid business, left and founded Luming Robotics. At the 2026 CCTV Spring Festival Gala, Magic Atom appeared as a strategic partner for intelligent robots and became an overnight sensation. Right at the critical moment of sprinting for an IPO, on March 5, Wu Changzheng announced his resignation and entrepreneurship. Sources familiar with the matter said the trigger for his resignation was a disagreement with shareholders over the company's development philosophy.
In any case, the tension over control exists to varying degrees in many internally incubated embodied AI projects.
Automakers incubating embodied AI has been a trend in recent years. The logic is not hard to understand: smart cars and robots share many technologies. Extending the layout of embodied AI vertically along their own business is almost zero-friction strategically for automakers. However, in practice, tension still exists.
In 2020, Dogotix was acquired by XPeng. Founder Zhao Tongyang and He Xiaoping jointly established Pengxing Intelligence, focusing on the R&D of humanoid robots. In October 2023, Zhao Tongyang left Pengxing Intelligence and founded ZQ Robotics, still focusing on bipedal humanoids.
Having experienced the trials of the business world, ZQ Robotics, unlike the academic faction and small startups, entered the market with a strong industrial gene, mainly reflected in extremely fast mass production and commercialization. Less than a year after its establishment, it released SA01, the world's first full-size general-purpose humanoid robot with an anthropomorphic straight-knee gait.
In December 2025, the T800 was released. A test video of Zhao Tongyang being kicked down by the robot once went viral on the internet, making people realize that the comprehensive physical performance of humanoid robots has already surpassed that of 90% of adult males. In January 2026, at its global debut at CES, hardcore demo videos exploded on overseas social media; in May, the Shenzhen Honghualing base was put into use, with one humanoid robot rolling off the line every 15 minutes, reaching a delivery capacity of 10,000 units annually. In a recent robot fighting competition, it was even reposted by Elon Musk, who called it interesting.
Whether it is Magic Atom or ZQ Robotics, both have an extremely strong industrial gene. From these two types of companies, we can find that embodied AI companies with deep participation from industrial capital or talents with industrial backgrounds are often more sensitive to engineering implementation and will balance cutting-edge R&D with mass production and commercialization.
This trait will also further attract capital markets during the window period when the commercialization of humanoid robots begins to materialize.
Giant Capital Flows into the Vast Ecological Wilderness Is there a type of company that can mobilize the resources of the investing entity without being as tightly bound as internally incubated projects? Yes, those with a background of internet giants.
Internet giants often do not build robots themselves, nor can they afford to lose the strategic entry point of physical AI. Through equity ties, they can form a community of shared interests with their core business and upstream/downstream partners, sharing resources within the ecosystem: the invested embodied AI companies get reliable supply, and other companies in the ecosystem get orders.
For embodied AI companies, being selected by the industrial capital of giants is a highly valuable ticket. What they get is not just money, but the first batch of real scenarios for deployment and continuous business feedback.
Take LimX Dynamics as an example. This company, founded by Zhang Wei, a professor at the Southern University of Science and Technology, has a type of giant capital that brings its own application scenarios on its shareholder list. In July 2025, JD.com led a new round of financing for LimX Dynamics, deepening collaboration in retail, logistics, and services, and entering China's largest logistics scenarios.
Zibianliang Robotics is similar, having simultaneously received investments from four internet giants: Meituan, Alibaba, ByteDance, and Xiaomi. Among them, Meituan led the Series A, Alibaba led the Series A+, ByteDance led the Series A++ with 1 billion RMB, and Xiaomi led the nearly 2 billion RMB Series B. By the end of June 2026, after the completion of multiple rounds of closing, the company's valuation broke through 20 billion RMB.
What the four giants give is far more than just money; it also includes real scenarios, data, and orders, which are the scarcest resources in the embodied AI industry. In January 2026, with the joint backing of Alibaba Cloud and ByteDance, Zibianliang's Quantum 1, relying on its self-developed WALL-A model, completed the last 100 meters of autonomous food delivery in a real open environment. Xiaomi's smart home ecosystem and home entry points are also highly compatible with Zibianliang's commercialization route of bringing robots into homes.
Looking a bit further, we will find that behind brands such as LimX Dynamics, Unitree Robotics, Yuanli Lingji, Agibot, and Leju Robotics, there are almost always one or more tech and internet giants standing.
You might ask, what is the difference between these companies and those internally incubated by industrial capital? Aren't they all giving resources and scenarios? The answer is that internal incubation is like a biological son, while giant companies will cast a wide net in the embodied AI track, betting on multiple companies at the same time, which is more like a young talent favored by the patriarch of a large family. Therefore, this avoids the problem of limited autonomy for the core team. This also means that they cannot be unconditionally favored like a biological son. Whether to be supported depends on product strength to win the tilt of ecosystem resources.
For example, JD.com made six consecutive investments in the embodied AI track in 2025, continuously taking stakes in Agibot, Qianxun Intelligence, LimX Dynamics, ZQ Robotics, PaXini, and RoboScience. Later, it further increased its bets, pressing down on Deep Robotics, Shouhang Technology, and Wujie Power. Currently, JD Logistics has announced that it will purchase 3 million robots, 1 million unmanned vehicles, and 100,000 drones in the next 5 years. Who will stand out in this ecosystem? Let us wait and see.
Beyond Core Business: The Ideal Second Growth Curve If the independent incubation of embodied AI companies by smart terminal manufacturers is about digging deep into their own potential, and introducing giant investments is about grafting onto a vast tech ecosystem, then there is another type of industrial capital that enters the embodied AI field to build a new growth track, and that is manufacturing enterprises.
They have mass production experience in mechanical manufacturing, and their traditional businesses have already nearly peaked. It is increasingly difficult to achieve breakthroughs through internal innovation, yet they need to recreate new growth engines, especially by laying out industries unrelated to their core businesses to resist cyclical fluctuations. Incubating or investing in an embodied AI company gives them a ticket to participate in this emerging track.
For example, the cookware leader ASD established Zhejiang ASD Humanoid Robot Co., Ltd. in Wenling, Taizhou in April 2024, positioning it to cover humanoid, quadruped, and home service robots. The 100% holding company is exactly Shanghai ASD Artificial Intelligence Technology Co., Ltd. At the same time, in March this year, it made a strategic investment in a leading embodied AI company and was listed as one of the key investment institutions for Magic Atom.
ASD is not an isolated case. Looking around the traditional manufacturing industry, many segmented leaders are doing similar things.
Midea started with complete machines and core components, releasing the industry's first six-arm wheeled-legged humanoid robot "Melo U", which has entered its own washing machine factory for pilot operations, and announced an investment of over 60 billion RMB in the next three years focusing on AI and embodied AI. Haier even released three home robots—humanoid, laundry care, and companion—at the 2026 AWE in one go. Not to be outdone, Gree officially announced in April 2026 that it has the ability to develop complete humanoid robots, with all core components self-developed and self-produced, and its GR-1 prototype focuses on industrial scenarios.
It is not hard to find that the competitive advantages of traditional manufacturing have not become invalid in the era of embodied AI.
They have the engineering capabilities for complete machine optimization and assembly, coupled with years of accumulation in refining products for home scenarios, and mature sales channels. Once the categories of embodied AI mature, they can quickly push products from prototypes to consumer terminals.
However, limited by insufficient resources and talent attraction, the self-developed AI capabilities of these companies are limited, showing a clear gap with the academic faction and the giant faction.
Therefore, making robots on their own while connecting with external forces through investment has become the mainstream choice for these companies to build a second curve.
At present, these embodied AI manufacturers may not be outstanding yet, unable to win the top prize on competitive stages like marathons or the Spring Festival Gala. But perhaps they will be the ones to enter thousands of households earlier, which deserves everyone's attention.
The Fuse of Industrial Capital Strings Together an Ecosystem The flow of money is the subtle trace of an industry, and often the truest truth.
Simply pursuing financing arbitrage is the game of financial capital. But when industrial capital does embodied AI, it demands deterministic returns, which may be in business, or in the ecosystem and strategy. Whichever return it is, there is a certain capability threshold.
Imagine that a robot product of extremely poor quality cannot enter the factory to screw bolts for Chery or deliver express packages for JD.com. And an embodied AI company with insufficient R&D and slow iteration cannot bring usage of cloud and infrastructure, thereby driving upstream and downstream ecosystem companies to share the pie.
Therefore, this smartest money from industrial capital has already filtered the scenarios and resources for everyone, instead becoming a very useful filter for readers: embodied AI companies that can be invested by industrial capital often come with more reliable resources and more deterministic returns, making them key players to focus on in this ecosystem.