In late March 2026, a key engineer who spent years perfecting the L9 SUV's autonomous features quietly left Li Auto to start an embodied AI company. Within weeks, Li Auto made an unusual move: it invested in that very startup. Alibaba's CEO followed, participating in the round. Three separate signals—investment, open-source tools, and industry declaration—now converge on a single conclusion: 2026 is when physical AI stops being theoretical.
The Li Auto investment carries weight beyond capital. The EV maker has long guarded its engineering talent; that it allowed a core developer to leave and then co-invested suggests urgency. Industry observers quoted by 量子位 QbitAI note that Li Auto may already be developing its first humanoid robot internally. If true, the company joins BYD, Xiaomi, and a dozen other Chinese firms racing to translate automotive AI into physical machines.
The second signal came from SentiPulse and Renmin University's Gaogl Institute, which open-sourced SentiAvatar on April 8th. The interactive 3D digital human framework reportedly outperforms existing industry models in real-time responsiveness. Unlike proprietary systems, this open release lowers the barrier for any company building digital assistants, telepresence robots, or customer-facing avatars. If SentiAvatar gains adoption, it could become the foundational layer for how humans interact with embodied systems at scale.
The third signal is editorial: 量子位 QbitAI published a lengthy analysis declaring 2026 the "Year of Physical AI" (具身原生元年), tracing momentum across investment rounds, research breakthroughs, and corporate pivots. The publication identified a pattern—companies that once treated robotics as a side project now treat it as a core competency. The analysis argues that 2026 marks the inflection point where embodied AI graduates from demos to deployment.
What ties these three events together is timing and intent. Li Auto's investment targets humanoids. SentiAvatar enables the digital interfaces those humanoids will need. And the industry declaration legitimizes the entire ecosystem. Separately, each is notable. Together, they suggest a coordinated acceleration rather than scattered experimentation.
The stakes are asymmetric. Chinese EV makers possess manufacturing advantages—batteries, sensors, supply chains—that Western robotics firms lack. But the talent gap remains. Li Auto's willingness to co-invest in its former engineer reveals how desperate the competition has become. Whoever controls embodied AI talent controls the next platform shift. The L9 engineer did not simply start a company; he triggered a bidding war that will reshape how Chinese tech allocates capital, talent, and strategic focus through the end of this decade.