Gasgoo Munich – RealMan recently announced that it has secured nearly 500 million yuan in funding. Guided by strategic investments from several well-known public companies, the funds will be funneled into product research, development and iteration, the construction of the AUTRON superfactory, and the deepening of a global ecosystem built on “hardware, data and remote control networks.”
As one of the earliest domestic players focused on embodied AI infrastructure, RealMan aims to build a system-level platform for the next era. The company aims to make robots truly practical by combining “reliable hardware, real-world robot data, and remote control networks.”
With this goal in mind, the latest round not only represents industry validation of RealMan’s strategic flywheel; It also provides new weapons for end-to-end layouts, from core components to superfactories, and from real-world data to global networks.
Image source: RealMan
Deploying robots requires reliable hardware
RealMan’s story begins with a series of numbers.
By 2025, the annual production capacity of RealMan’s self-developed integrated joint modules will exceed 100,000 units. The company’s robotic arms are CR L3 certified and have a mean time between failures of 50,000 hours.
What does 50,000 hours actually mean? Assuming an 8-hour workday, that means over 17 years of continuous operation without failure. For robots entering homes, factories, and commercial spaces, hardware reliability is perhaps a more important prerequisite than algorithmic power. No matter how smart your “brain” may be, if your “body” malfunctions, you will not be able to truly develop your brain.
Behind this is 8 years of accumulation by RealMan.
Since its founding in 2018, RealMan has driven innovation around dexterous robotic arms and precision, durable joints. The company’s position is clear. For robots to enter everyday life, the hardware must be reliable enough.
To achieve this goal, RealMan is designing a robotic arm to benchmark the arm length, girth, flexibility, and weight capacity of an adult male. The aim is to achieve seamless integration between people and their working and living environments, both in form and function.
It is worth noting that in the field of embodied AI, many companies are now sourcing collaborative modules externally and focusing on system integration. RealMan chose a different path, starting with the underlying core components and achieving full-stack self-development.
This path may be heavier and slower, but once established, it means you have complete control over the reliability of your product.
Beyond hardware, data becomes a moat
If reliable hardware is the first layer of the foundation, data is the second layer.
There is industry consensus that data matters when it comes to embodied intelligence. However, data collection approaches vary. Some rely on synthetic simulations, others rely on remote control, and still others learn from internet videos.
RealMan’s solution is the GLN Remote Operations Network.
The logic of the system is simple. An immersive remote operation interface allows operators to instruct the robot in real-time to perform tasks in real-world scenarios. This ranges from everyday chores like folding towels or moving boxes to specialized tasks that require specialized knowledge like inspecting pipe galleries or controlling power grids.
All remote operations have two purposes: performing tasks and collecting real-world robot data.
Image source: RealMan
At CES earlier this year, RealMan demonstrated transoceanic operations from Beijing to Las Vegas. During Beijing TV’s Lunar New Year celebration, the company’s robots participated in making dumplings, while accumulating data from real-world settings.
The value of these scenarios is that the data is “captured” through real-world tasks, rather than “done” in the lab.
These capabilities are primarily due to RealMan’s deep accumulation of high-quality data in real-world scenarios. At the Beijing Humanoid Robot Data Training Center, leveraging 108 embodied robot bodies across 10 major application scenarios, the company has accumulated real-world data covering more than 1,000 tasks and tens of millions of trajectory segments.
Building on this, the company recently announced that it will open source the world’s first high-quality real robot dataset with most modalities.
The move likely aims to go beyond simple philanthropy. In the early stages of an industry, open sourcing data, establishing standards, and forcing developers to iterate on algorithms on hardware and data is itself a way to build ecosystem barriers.
From 100,000 to 1 million: Breaking through the production bottleneck
The third layer of foundation is mass production capacity.
Since taking its first steps toward globalization in 2024, RealMan has expanded its operations to Asia, Europe, North America, and South America. We currently serve more than 8,000 enterprise customers worldwide, spanning industrial automation, research and education, and commercial services.
Our customer base of 8,000 is unusual among embodied AI startups. A closer look suggests several impacts, including products enabling standardized delivery capabilities, expanding the customer base from “early adopters” to “practical users,” and covering a greater variety of scenarios.
To better overcome industry production bottlenecks and ensure large-scale deliveries, RealMan plans to leverage the AUTRON Superfactory in 2026. The facility will host large-scale manufacturing capacity for flexible, mixing lines, with a targeted annual production capacity of 1 million bonded modules.
Scaling up from 100,000 units to 1 million units will take RealMan’s supply chain bargaining power, delivery capabilities, and cost control to a new level.
Production capacity is emerging as a new bottleneck, especially in the field of body-based AI. As more robotics companies move from R&D to mass production, the availability of core components will determine which companies can actually deliver and which are left with only prototype demonstrations.
RealMan’s decision to double down on superfactories at this point suggests that the company is clearly aware of this trend.
In particular, RealMan’s operating cash flow entered a virtuous cycle in 2025, driven by full-stack in-house development of core components and large-scale delivery.
It’s important to note that while most startups in the embodied AI race are still wasting money, this sign shows that RealMan’s business model is achieving some degree of self-sustainability. Against this background, the simultaneous introduction of capital from multiple listed companies in this round is further proof that the company’s technology roadmap and product definition have won the support of the industrial chain.
About Seeds Discovery:
Gasgoo’s Seeds Discovery column aims to build a service platform that connects startups, industrial ecosystem partners, investment firms, and local governments to deeply empower upstream and downstream supply chains. Since its inception, this column has been dedicated to providing inspiration and leadership in the wave of intelligent transformation, thereby uncovering exemplary companies, technologies, and business models that fuel the growth of the automotive industry’s innovative forces. According to Gasgoo statistics, almost all startups featured in “Seeds Discovery” have successfully connected with industrial ecosystem resources.
