Embodied AI's ChatGPT Moment: Hype or Reality?

At the 2026 Inclusion · Bund Conference in Shanghai, the morning of September 11 saw a packed room of industry veterans, investors, and young entrepreneurs grappling with a question that's equal parts tantalizing and thorny: when will embodied intelligence have its ChatGPT moment? The discussion, however, quickly turned to a more immediate concern—is the sector already inflating into a bubble?

The Bubble Debate: Pre-Dawn or Overheated?

With a surge of capital and a wave of startups, embodied AI has become a magnet for both excitement and skepticism. Jia Peng, co-founder and CEO of Zhijian Dynamics, doesn't mince words. "The bubble people feel now is more about a mismatch between stages and valuations," he says. In the long run, he argues, embodied intelligence could spawn an industry larger than mobile phones or cars, with a supply chain stretching from sensors to final applications. But right now, "we haven't created much value yet; it's a phase mismatch." The sector, he adds, needs far more human talent and GPU power—overall investment is still lagging.

Xiao Wanggang, founder and chairman of Daxiao Robotics, echoes that sentiment. "I feel that we are in the pre-dawn," he says. Compared to last year's reliance on real-machine data and small models, this year has brought ego data and world models, hinting at the Scaling Law's manifestation. Yet the pieces—data, components, models, physical bodies, scenarios—remain scattered. "Different types of companies may further recombine in the future," Xiao notes.

Han Zheng, co-founder and CEO of Sudo Technology, points to China's rich supply chain and testing grounds as a breeding ground for exploration. He sees a convergence in technical routes, with a few players likely to dominate complete underlying solutions, while vertical applications and global markets still hold opportunities.

The ChatGPT Moment: When and How?

So when will embodied AI hit its stride? Han Zheng compares it to large language models in 2018-2019: key frameworks exist, but the challenge is integrating them into a reliable, commercially viable system. "Integrating this big system is the most important issue right now," he says.

Jia Peng, however, bristles at the term "embodied ChatGPT." The physical world demands far higher reliability than digital AI. "If we only look at model capabilities, embodied intelligence has already shown some in-context learning abilities," he says. "But for real productivity, it may still take three to four years, or even four to five." His biggest bottleneck? GPU computing power—especially on the edge, where many factory scenarios can't connect to the internet.

Xiao Wanggang zeros in on data. "If we always need to re-train for new tasks, we won't achieve large-scale promotion," he warns. Yet recent zero-shot capabilities in general models give him hope: intermediate data could accelerate the evolution of embodied foundation models.

Young Entrepreneurs: World Models and Global Competition

The forum also spotlighted a new generation of founders born in the 90s and 00s, bringing fresh perspectives. Li Yiming, founder of Licheng Intelligent, recalls, "When we started, we said that in the next six months, all embodied intelligence companies would claim to be world model companies." The key, he insists, is solving specific problems. Startups should define the end goal first, then work backward to the technical routes.

Ding Ning, founder of Natural Will, believes the GPT-3 moment for embodied AI has already arrived—just unnoticed. Since 2025, large models have solved fundamental issues, but operating in the physical world remains a hurdle. He describes the industry's pace as "unusual," with a "very strong pushing force" from code volume to experimental results. Facing overseas giants with massive computing resources, he admits to feeling "very anxious," but sees China's edge in hardware and supply chains. "High-precision data requires better sensors, and better sensors require intelligent fusion technology. This can only be produced in Shenzhen globally right now."

Chen Boyuan, a 22-year-old founder of Nix Matrix Technology, grew up after the Scaling Law emerged. "We've experienced the rise and peak of large models, but they still face insurmountable challenges in the physical world," he says. For embodied AI to truly enter the physical world, physical AI foundations must understand causality. "Edge cases are not the tail, but the norm." He rejects the idea that a few geniuses can crack it: "People won't easily trust your chosen path just because you used to work at a big company; nor will they dismiss your ideas because you're an intern." The key is respecting first principles and validating through small-scale tests and large-scale scaling.

Key Points

Bubble or not? Industry leaders see a mismatch between valuations and current value, but remain optimistic about long-term potential.
Main bottlenecks: GPU computing power (especially edge-side) and data scarcity hinder progress toward a ChatGPT moment.
Timeline: Some predict 3-5 years for real productivity; others argue the moment is already here but unnoticed.
Young entrepreneurs: Emphasize world models, physical AI, and China's supply chain advantages in global competition.
System integration: The biggest challenge is combining disparate technologies into a reliable, commercially viable system.

Embodied AI's ChatGPT Moment: Hype or Reality?