Embodied AI's ChatGPT Moment: Are We There Yet?

At the 2026 Bund Conference, the buzz around embodied intelligence was palpable. But amid the excitement, a nagging question lingered: are we in a bubble? Industry veterans and young entrepreneurs gathered to dissect the hype, the tech, and the timeline for that elusive 'ChatGPT moment.'

The Bubble Debate: Pre-Dawn or Overhyped?

Many speakers agreed that the current 'bubble feeling' stems from a mismatch between the industry's early stage and soaring valuations. "We haven't created much value yet," said Jia Peng, CEO of Zhijian Dynamics. "It's a phase mismatch." He believes embodied intelligence could eventually dwarf mobile phones or cars in scale, but it needs more talent, GPUs, and investment.

Xiao Wanggang of Daxiao Robotics echoed this, calling it the "pre-dawn" phase. He noted that while last year relied on real-machine data and small models, this year's focus on ego data and world models is starting to show signs of a Scaling Law. Still, the pieces—data, components, models, and scenarios—remain scattered.

The ChatGPT Moment: When and How?

So when will embodied AI have its ChatGPT moment? Han Zheng of Sudo Technology compares it to large language models in 2018-2019: key frameworks exist, but integrating them into a reliable, commercial system is the big challenge.

Jia Peng dislikes the term 'embodied ChatGPT' because the physical world demands far higher reliability than digital AI. He estimates three to five years for real productivity, with GPU computing power—especially on the edge—as the biggest bottleneck. Many factory scenarios can't rely on the cloud.

Xiao Wanggang pins the bottleneck on data. The key, he says, is achieving intelligent emergence, scenario generalization, and zero-shot capabilities. "If we always need to retrain for new tasks, we can't scale," he warned. Recent zero-shot successes in general models give him hope.

Young Entrepreneurs: World Models and Global Competition

The forum also spotlighted young founders born in the 90s and 00s. Li Yiming of Licheng Intelligent quipped, "In six months, every embodied AI company will claim to be a world model company." He advises startups to define their end goal and work backward.

Ding Ning of Natural Will believes the GPT-3 moment has already arrived—just unnoticed. He describes the pace as "unusual," with rapid code and experiment iterations. Despite anxiety over overseas computing power, he sees China's edge in hardware and supply chains for high-precision data. "High-precision data needs better sensors, and that can only be produced in Shenzhen right now."

Chen Boyuan, 22, founder of Nix Matrix Technology, emphasizes that physical AI must understand causality. "Edge cases are the norm, not the tail." He argues that physical AI is complex system engineering, not a solo genius act. "People won't trust your path just because you worked at a big company, nor dismiss it because you're an intern." Respect first principles, validate small, scale big.

Key Points

Embodied intelligence is in a 'pre-dawn' phase, with valuations outpacing commercial value.
The 'ChatGPT moment' may be 3-5 years away; bottlenecks include GPU computing power and data.
Young entrepreneurs see opportunities in world models, China's hardware supply chain, and physical AI fundamentals.
System integration and zero-shot generalization are critical for scaling.
The industry structure is far from finalized; vertical applications and global markets remain open.

Embodied AI's ChatGPT Moment: Are We There Yet?