Gasgoo Munich- "If you enter through a narrow niche and the business model is scalable, the odds of success are high—from the perspective of capital, industry, and long-term sustainability, it's a massive opportunity."
At the 2026 Embodied AI Industry Scenario Fusion Conference on July 3, hosted by Gasgoo Embodied AI, Gasgoo Group CEO Zhou Xiaoying opened with a clear thesis for the sector: the goal isn't to chase grand, all-encompassing narratives, but to find precise, replicable entry points that close the loop.
Gasgoo Group CEO Zhou Xiaoying
In Zhou's view, AI is accelerating its migration from the digital realm to the physical one. While digital assistants have become standard, the physical AI landscape in China is characterized by a uniquely fragmented boom. If you count just the robot makers, there are already over 500 domestic companies; factor in the entire upstream and downstream supply chain, and Gasgoo's internal data puts the number in the thousands. That scale far surpasses the map of "new forces" in the new energy sector from a decade ago, which listed just over 200 players.
Why such density? Because robotics is a "scenario-driven" industry, spanning companionship, care, and industrial applications—where the industrial segment itself allows for infinite subdivision. "Often, simply calling yourself a robotics company makes it hard to stand out," Zhou noted. "But if you define the conditions, the closed loop, and the specific type of robot, you become far more recognizable." She emphasized that scenarios define form, and form determines positioning.
From Spectacle to Substance: Mass Delivery and Commercial Viability Take Center Stage
By 2025, the industry had completed its market education and the first round of concept validation, with two "Tech Spring Festival Galas" playing a key role in popularization. But by 2026, the sector's shared keywords have shifted to mass delivery, financial metrics, and commercial viability. "You can't just keep burning venture capital; you need a strong commercial story," Zhou stated.
She cited her visit to the Munich Electronics Show in Pudong two days prior, where robots roamed every exhibition hall. "People might view it as a marketing stunt, but nobody is surprised anymore. When a robot suddenly appears next to you, it doesn't elicit the shock it did a year or two ago." Coexisting with robots has become mundane in daily life.
Zhou predicts the next two to three years will be a watershed for the industry. China's industrial iteration is blisteringly fast, meaning "technological leadership might last only three months." In her view, the gap between companies is no longer defined by pitch decks, but by execution in three dimensions: application in concrete scenarios, stable interaction, and closed-loop data. "In two or three years, given China's rapid iteration cycle, the path should be clear and the pitfalls identified."
Take humanoid robots, for example. While shipments are surging, applications remain concentrated in research, education, exhibitions, data collection, and pilot programs; large-scale industrial deployment is still limited. Yet, globally, China now accounts for over 80% of shipments. "That figure is staggering."
Zhou also highlighted a paradox: some European embodied AI companies appear "mediocre" in terms of business, yet command astronomical valuations. It reminds her of the path Chinese EVs took globally over the last decade: "In the coming decade, China likely holds a strong winning hand in unstructured AI. It's a space where policy, capital, and enterprise are all reinforcing each other."
Hardware: From General to Specialized; Software: From Layered to Integrated
The physical AI landscape currently features multiple technological paths running in parallel. Scenarios dictate form—whether fixed-base, wheeled, legged, or dual-armed. Notably, a wave of robotics companies is entering the automotive supply chain to find their fit. From perception and decision-making to thermal management and range, embodied AI and smart electrification share "more than 65% of the same DNA." The next trend, however, is for core hardware to shift from general-purpose to robot-specific components, mirroring the shift from consumer electronics to automotive-grade parts.
Data presented by Zhou shows that for categories like joint modules, ball screws, motors, and cameras, China already has a deep bench of existing suppliers, the vast majority of which have expanded into robotics.
She shared a telling anecdote: at an investor meeting for an A-share listed parts manufacturer, "they insisted on saving the robotics business update for last. If they put it first, people would walk out. They had to cover the other businesses first, and only then did investors show interest in the robotics segment."
This reflects the duality of China's current industrial landscape: the auto sector is in a survival-of-the-fittest phase with waning capital interest, while the robotics track is a mix of bubbles and competition. Valuations are "ridiculously high," yet the sector is drawing more participants, effectively "using a race-to-survive mechanism to filter out the truly strong companies."
On the software front, the short term remains dominated by layered, controllable systems; in the long run, enhanced VLA (Vision-Language-Action) world models will improve multimodal perception and modeling. Data is becoming the core moat—encompassing teleoperation, synthetic data, and data scraping—where quality, cost, and usage methods define competitiveness. "The logic is exactly the same as it was for autonomous driving," Zhou noted.
The Auto Plant: Complex Enough, Yet Simple Enough
The running, yoga, dancing, and simple material handling showcased today are essentially single-machine demos. "Actually going into factories to tighten bolts, achieving large-scale coordination and application—that will likely happen in the next two or three years." Zhou is looking forward to seeing not just individual unit training, but multi-machine collaboration, with swarms of robots working simultaneously on the factory floor.
Why is the auto factory one of the most critical early testing grounds for embodied AI? "Because it is complex enough, yet simple enough." On one hand, automotive plants are highly intelligent with complex production lines and numerous processes, yet the boundaries are relatively safe and controllable. On the other, a vast number of processes, material pacing, and line management tasks can be standardized.
Crucially, this transformation isn't a solo endeavor. OEMs are becoming increasingly open, with the entire supply chain moving in lockstep. Traditional automotive suppliers like Schaeffler and INOVANCE are entering the robotics space, opening up their own scenarios for validation, seeking reusable capabilities, and building a second growth curve. "It's a mutual convergence that delivers results."
Safety, of course, is the non-negotiable baseline—defining safety boundaries, trusted systems, and standardized engineering, combined with "human-centric design" and sufficiently low costs. Each of these hurdles presents a different path that must be overcome.
Meanwhile, going global is emerging as another growth curve for embodied AI. In the past, overseas expansion by companies like UBTECH was a "separate line of business." Now, the global footprint of OEMs is opening up new physical spaces for robotics.
Zhou pointed to Chery, whose self-developed robots are already demonstrating in overseas sales showrooms. Future applications in after-sales and community services could soon follow. More importantly, "tariffs on exported robots are far lower than for EVs—in many places, they enter at zero duty."
In her view, the next two to three years will not only see the convergence of the automotive and embodied AI industries but also witness them expanding from China to the world together. "We have the opportunity to participate in a genuine, historic era of industrial innovation. It is truly a once-in-a-century moment."
