Over the past year, embodied AI has emerged as one of the most closely watched sectors in global technology. From Figure AI and Physical Intelligence in the United States to AgiBot and Galbot in China, investors, researchers, and industry leaders have all been pursuing the same question: Who will build the intelligence layer that powers the next generation of robots?
For decades, robots have largely operated through predefined rules, carefully engineered workflows, and highly structured environments. The vision of truly intelligent machines—robots capable of understanding their surroundings, adapting to unfamiliar situations, predicting outcomes, and making autonomous decisions—has remained elusive. Today, however, advances in foundation models and embodied intelligence are bringing that vision closer to reality.
Against this backdrop, a relatively young Chinese company called Daxiao Robotics has rapidly attracted attention. In the first half of 2026 alone, the company reportedly raised hundreds of millions of dollars and reached unicorn status. At the same time, its proprietary world model, Kairos, has achieved strong results across several influential embodied AI benchmarks, while the company continues to promote its belief that world models—not traditional robot control systems—will become the foundation of future robotic intelligence.
The combination of technical ambition, heavyweight investors, and a high-profile leadership team has made Daxiao Robotics one of the most closely watched companies in China’s embodied AI ecosystem. The key question now is whether it can evolve from a promising startup into a foundational platform for the robotics industry.
Why Has Daxiao Robotics Suddenly Become a Major Story?
At first glance, Daxiao Robotics might appear to be just another company entering the increasingly crowded robotics market. However, a closer look reveals that its focus differs significantly from that of many of its peers.
Most robotics companies are centered around hardware. Their competitive advantage comes from building better humanoid robots, more capable robotic arms, or more agile quadruped systems. Public attention tends to focus on physical performance: how fast a robot can move, how much weight it can carry, or how human-like its appearance may be.
Daxiao Robotics is taking a different approach.
Rather than positioning itself primarily as a hardware company, it is attempting to build what it describes as the “brain” of the robot era. The company’s central product is not a robot body but a world model called Kairos, designed to help machines understand, predict, and interact with the physical world.
In other words, Daxiao is not primarily trying to answer the question, “What should a robot look like?” Instead, it is focused on a much deeper challenge: “How can a robot understand reality well enough to act intelligently within it?”
This distinction is important because it reflects a broader shift happening across the robotics industry. Increasingly, the bottleneck is no longer hardware. The real challenge lies in creating systems that can reason about the world, generalize across environments, and operate safely in unpredictable situations.
Why Are Investors Betting So Aggressively?
One of the most intriguing aspects of Daxiao Robotics is not the amount of capital it has raised but the composition of its investor base.
The company has attracted support from an unusual combination of internet giants, industrial corporations, state-backed funds, and top-tier venture capital firms. Such a coalition rarely forms around an ordinary startup.
This suggests that investors see Daxiao as more than a robotics company. Many appear to view it as a potential provider of critical infrastructure for the future robotics economy.
Among the most notable investors is Ant Group, whose involvement initially surprised many observers. After all, Ant is best known for financial technology and digital services rather than robotics.
Yet from a long-term perspective, the investment makes strategic sense. During the mobile internet era, companies like Ant built platforms that connected people to digital services. In a future where robots become widespread in hotels, shopping centers, office buildings, warehouses, and eventually homes, robots themselves may become a new interface between digital systems and the physical world. From this perspective, Ant is not investing in robots as hardware products; it is investing in a potential platform for real-world intelligence.
Geely Capital represents a different strategic logic. Modern autonomous vehicles and future robots share many underlying technologies, including environmental perception, world modeling, decision-making, and edge computing. In many ways, an advanced robot can be viewed as an autonomous vehicle that operates in three-dimensional human environments rather than on roads. Geely’s investment therefore reflects a belief that robotics may become the next major frontier for technologies originally developed in autonomous driving.
The participation of MetaX, a leading Chinese GPU company, adds another layer to the story. World models require substantial computational resources for both training and inference. If embodied AI becomes a major industry, demand for robotics-oriented AI infrastructure could grow dramatically. MetaX is effectively positioning itself within that future ecosystem.
Why Are State-Backed Funds Getting Involved?
Equally significant is the participation of government-backed investment funds, including the Shanghai Science and Technology Innovation Fund, the Lingang New Area Fund, and university-affiliated investment platforms.
Their involvement signals that embodied AI is increasingly being viewed not simply as a promising startup category but as a strategically important technology sector.
Over the past two decades, China achieved remarkable success in industries such as mobile internet, digital payments, and electric vehicles. Looking ahead, many policymakers and industry leaders see robotics as one of the next major platforms capable of reshaping economic productivity and industrial competitiveness.
From this perspective, the intelligence layer that powers robots may ultimately prove as important as semiconductors, operating systems, or cloud infrastructure. State-backed investors tend to prioritize long-term strategic technologies rather than short-term market trends. Their presence suggests a belief that foundational robotics intelligence could become a critical national capability over the coming decades.
What Exactly Is a World Model?
Understanding Daxiao Robotics requires understanding the concept of a world model.
Most current robotics systems rely on what is commonly known as a Vision-Language-Action (VLA) architecture. In this framework, a robot observes its environment through sensors, interprets instructions through language models, and then generates actions.
This approach has produced impressive results, but it also has limitations. In many cases, the system learns correlations rather than developing a deeper understanding of how the world works. As a result, performance can deteriorate when robots encounter unfamiliar environments, unexpected objects, or unusual conditions.
World models attempt to address this problem by introducing an internal representation of reality.
Instead of directly mapping observations to actions, a robot first constructs a predictive model of the environment. It uses that model to simulate future outcomes before deciding how to act.
Humans operate in a similar way. When we see a glass sitting precariously near the edge of a table, we instinctively anticipate what could happen if it falls. We understand that the glass may break, water may spill, and the floor may become slippery—even before any of those events occur.
A world model seeks to provide robots with a comparable ability to reason about cause and effect within the physical world.
The ultimate goal is not merely better task execution. It is to create systems capable of adapting to new situations, transferring knowledge across environments, and operating effectively without exhaustive retraining.
Why Is Kairos Receiving So Much Attention?
Among the many claims surrounding Kairos, perhaps the most noteworthy is its reported efficiency.
According to publicly available information, Kairos-4B contains approximately four billion parameters, significantly smaller than several competing systems that range from sixteen to twenty-eight billion parameters. Yet in a number of world-model-related evaluations, Kairos has reportedly achieved competitive or superior performance.
This matters because robotics imposes very different constraints than cloud-based AI systems.
Large language models can run in massive data centers with virtually unlimited computing resources. Robots, by contrast, must operate within strict limitations involving power consumption, hardware costs, latency, thermal management, and onboard computing capacity.
If a relatively compact model can deliver strong performance while running directly on robotic hardware, it may prove far more valuable than a much larger model that requires extensive infrastructure.
For this reason, Kairos is attracting attention not simply because of benchmark results but because it represents a potential alternative path toward scalable robotic intelligence.
The Most Important Milestone: Edge Deployment
While benchmark rankings often dominate headlines, one of Daxiao Robotics’ most significant achievements may be its focus on edge deployment.
Historically, many robotics systems have depended heavily on cloud computing. Robots collect information from their environment, send it to remote servers for processing, and then receive instructions in return.
Although this approach provides access to powerful models, it also introduces latency, network dependence, operational costs, and reliability concerns.
Daxiao claims that Kairos can run directly on robotic hardware, enabling local perception, reasoning, and decision-making without continuous reliance on cloud infrastructure.
If this capability proves robust in real-world environments, it could represent a major step forward. Robots that operate independently and respond in real time are essential for large-scale deployment across homes, factories, public spaces, and industrial settings.
How Far Has Commercialization Progressed?
Despite the excitement surrounding the technology, commercialization remains the ultimate test.
Daxiao Robotics has publicly discussed applications in retail, security patrols, hospitality, tourism, and intelligent facility management. The company has also highlighted pilot programs involving robotic patrol systems.
However, it is important to maintain perspective. The entire embodied AI industry remains in its early stages.
Neither Daxiao Robotics nor most of its international peers have yet demonstrated deployment at truly massive scale. Large recurring revenue streams, widespread adoption, and proven business models remain largely works in progress.
As a result, Daxiao’s next challenge may not be technological innovation but rather translating technological leadership into sustainable commercial value.
The Real Competitive Advantage: The Team
Ultimately, technology companies succeed because of people, and this may be Daxiao Robotics’ strongest asset.
The company is led by Wang Xiaogang, co-founder of SenseTime and a highly respected figure in computer vision and artificial intelligence. Educated at the University of Science and Technology of China and MIT, Wang combines world-class research credentials with extensive experience in industrial deployment. Unlike many researchers who remain focused on academia, he has successfully scaled AI technologies into commercial products, including large-scale automotive applications.
Alongside him is Professor Dacheng Tao, one of the most influential AI researchers in the Chinese-speaking world. A Fellow of the Australian Academy of Science and former founding dean of JD Explore Academy, Tao brings deep expertise in both academic research and applied AI development.
Together, they represent a rare combination of scientific leadership and commercialization experience, providing Daxiao with a significant strategic advantage.
What Is Daxiao Robotics Really Building?
Although Daxiao Robotics is often described as a robotics company, that label may actually be too narrow.
Viewed through the lens of its technology, investors, and long-term vision, the company appears to be pursuing something much larger: a foundational intelligence platform for robots.
If the future robotics industry evolves in a way that resembles the smartphone industry, robot manufacturers may eventually resemble smartphone makers, while world models function as the equivalent of Android or iOS—a shared intelligence layer that powers an entire ecosystem.
From this perspective, Daxiao’s long-term value may not come from selling robots themselves. It may come from becoming the platform upon which many future robots depend.
Whether that vision ultimately succeeds remains uncertain. But it is increasingly clear that this is the opportunity investors are betting on.
Conclusion
It is still far too early to declare any company the winner of the embodied AI race.
Global competitors such as Figure AI, Physical Intelligence, NVIDIA Cosmos, and Google DeepMind are all advancing rapidly, and the industry remains highly fluid. The technologies involved are still evolving, and commercialization challenges remain substantial.
What does seem increasingly clear, however, is that the future of robotics will be determined less by hardware and more by intelligence. The industry’s center of gravity is shifting from mechanical engineering toward world modeling, reasoning, and generalization.
In that context, Daxiao Robotics has positioned itself at one of the most important intersections in the field. Its commitment to world models, its exceptional investor base, and its leadership team have made it one of the most compelling companies to watch in China’s emerging embodied AI ecosystem.
The most important question over the next five years may not be when robots enter everyday life, but rather who succeeds in building the cognitive architecture that makes widespread robotic intelligence possible.
Daxiao Robotics is attempting to become part of that answer.