This is the very last post of the year. The Daily Planet will be back in 2026, though I am not sure when. I am going to take a long break, and will be back no later than the end of Feb. And the topics are going to change dramatically when I do so. Happy New Year!
When UBTech robots walked onto the factory floor at Zeekr's electric vehicle plant in March 2025, they did something no humanoid robots had done before: they worked as a coordinated team, lifting boxes, assembling car parts, and performing quality checks - all without human supervision. Powered by DeepSeek's reasoning model, these machines represented more than a manufacturing curiosity. They embodied a fundamentally different vision of artificial intelligence, one that may reshape the global technology competition in ways Silicon Valley hasn't fully grasped.
While American tech companies pour hundreds of billions of dollars into data centers and language models, chasing the dream of artificial general intelligence through ever-larger neural networks, China has placed a different bet. Beijing believes that true AI dominance will come not from systems that generate text and images, but from systems capable of autonomous operation in the physical world - AI-powered robotics that can perceive, decide, and act. This is embodied AI, and understanding China's strategic commitment to it reveals a radically different approach to the technology that may define this century.
The Real Economy, the Real Bet
To understand China's AI strategy, you must first understand Xi Jinping's economic philosophy. Since becoming party general secretary in 2012, Xi has repeatedly emphasized that the "real economy" - the production of tangible goods and essential services in the physical world - constitutes the foundation of China's economic strength. This conviction was shaped partly by the 2008 global financial crisis, which exposed the risks of excessive reliance on the "virtual economy" of financial services and digital platforms. For Xi, deep integration of digital technology with the real economy represents the key to China's long-term growth.
When DeepSeek's R1 model shocked global markets in January 2025, demonstrating that Chinese AI capabilities had matured dramatically, the party concluded that the time had come to translate AI into the real economy. This wasn't just about keeping pace with American tech giants. It was about addressing China's most pressing domestic challenges: an economy slowing after decades of rapid growth, a property-market crisis, sluggish domestic consumption, and the beginnings of population decline. Embodied AI - robots that work alongside humans in factories, drones that survey infrastructure, autonomous vehicles that navigate cities - offered a path to revitalize productivity while solving practical problems.
The Fourth Plenary Session of the 20th Communist Party Central Committee, held in October 2025, made this official. The session outlined China's 15th Five-Year Plan for 2026-2030, identifying embodied AI alongside biomanufacturing, quantum technologies, and 6G as core tools for building the industries of the future. Premier Li Qiang's keynote at the World AI Conference in July 2025 highlighted embodied AI alongside large language models as areas experiencing major breakthroughs. The message was unmistakable: Beijing intends embodied AI to be a defining technology of China's future.
A Different Path to Intelligence
The contrast with American AI development is striking. In Silicon Valley, the dominant conviction holds that artificial general intelligence will emerge from scaling up language models - that money converts reliably to compute, which converts to capability, and that AGI is merely a matter of investment. This semi-religious belief shapes investment decisions and market expectations, sustaining valuations for companies betting everything on the next breakthrough in text generation.
Chinese AI communities haven't caught the same fever. The intellectual canon shaping Chinese entrepreneurs differs fundamentally from the one circulating in San Francisco. Where American founders read Peter Thiel alongside rationalist blogs speculating about superintelligence, their Chinese counterparts blend Western business classics with the "Red Canon" of political texts - Mao's selected works, Xi's writings on governance - that provide tactical guidance on organizational mobilization and survival in fiercely competitive markets. This is complemented by the "Grey Canon" of classical Chinese philosophy: Confucius on hierarchy and duty, Laozi on adaptability, Han Feizi on power and incentives. Literary works like Jin Yong's martial-arts novels and Liu Cixin's The Three-Body Problem offer frameworks for thinking about loyalty, strategy, and geopolitics in a hostile universe.
This produces a distinctly different approach to AI development, one less focused on metaphysical speculation about superintelligence and more grounded in practical applications aligned with national strategic goals. Leading Chinese AI scientists like Zhang Bo of the Chinese Academy of Sciences have argued that while large language models laid a crucial foundation for AGI by enabling machines to understand and generate language, embodied AI will ultimately allow AI to replicate the full spectrum of human capabilities. The reasoning is straightforward: systems that interact with and learn from the physical world have access to limitless data, can learn autonomously, and can share knowledge between agents - conditions ideal for intelligence to emerge.
Learning on the Factory Floor
The philosophical divide manifests in how robots are actually being developed. China is taking a bold, fast-paced approach by deploying large numbers of robots directly into real-world environments. This "learn-on-the-job" strategy allows machines to gather vast amounts of real-world data, which is then used to continuously improve their artificial intelligence. Companies like Unitree and Agibot are leading this effort, with Agibot offering an open-source operating system called Lingqu OS to encourage collaboration across the industry. By flooding the market with task-specific robots, China creates a massive, living laboratory that accelerates progress through collective learning and rapid iteration.
American companies, by contrast, are adopting a more cautious and controlled approach. Google and Meta focus on developing robot intelligence in simulated, controlled environments before deploying machines in the real world. This method prioritizes precision, safety, and reliability, aiming to perfect cognitive abilities in the lab to avoid costly failures or public mistrust. Google maintains closed, proprietary systems to protect its innovations, while Meta uses virtual platforms to train robots before any real-world exposure.
The tradeoffs are real. The American approach ensures more refined and dependable robots but limits exposure to the unpredictable challenges of actual environments. The Chinese approach accepts messier performance in exchange for rapid data collection and iterative improvement. Which strategy will prove superior remains to be seen - but the Chinese method aligns with Beijing's broader philosophy of moving fast, learning from mistakes, and scaling aggressively.
The Playbook in Action
China's development strategy follows a familiar pattern: encourage local governments to experiment, then scale the most successful approaches nationally. This "pilot first, scale later" method dates back to Deng Xiaoping's special economic zones in the 1980s and has proven remarkably effective for technologies from electric vehicles to digital currency.
Different provinces are now specializing in different segments of the embodied AI supply chain. Beijing, home to the chip developer Cambricon, has prioritized high-performance AI chips tailored for embodied applications. Shanghai, headquarters of the sensor company Hesai Technology, concentrates on core hardware components. Guangdong and Zhejiang - home to UBTech Robotics and Unitree Robotics respectively - focus on complete humanoid robotic platforms. Hubei province has established a laboratory for embodied intelligence technology in automobiles, leveraging the resources of local manufacturer Dongfeng Motor.
The funding is substantial. Beijing has launched a 100 billion yuan investment fund with a fifteen-year lifespan to support AI and robotics. Shanghai has established an embodied AI fund with initial funding of 560 million yuan. These investments are building an ecosystem of national champions: Unitree and UBTech in humanoid robotics, DJI in drones, Baidu Apollo and XPeng in autonomous vehicles.
Agentic AI and the Lexicon of Action
Beyond robotics, China is advancing rapidly in agentic AI - systems capable of taking autonomous actions rather than simply generating responses. In March 2025, the Singapore-based firm backed by Tencent released Manus, an agentic system that reviewers described as "mind-blowing, redefining what's possible." The same month, Beijing-based Zhipu AI launched AutoGLM-Rumination, claiming state-of-the-art performance on agent benchmarks. Alibaba, ByteDance, and Tencent have all released their own agentic frameworks.
The proliferation of terminology in Chinese discussions reveals a technology still finding its conceptual footing. Some terms emphasize the AI acting on someone's behalf - dàilǐ, meaning agent or proxy. Others emphasize autonomy - zìzhǔ zhìnéngtǐ, autonomous intelligent entity. The China Academy of Information and Communications Technology has released standards focusing on technical capabilities, safety, reliability, and controllability. The government is working to balance rapid innovation with risk management—encouraging development while ensuring systems remain under human control.
Strategic Advantages and Vulnerabilities
China's comparative advantages in embodied AI are significant. Its robust manufacturing base and comprehensive supply chains position it to rapidly scale production once the technology matures. The country leads in LiDAR sensor technology essential for 3D mapping and environmental navigation. Most importantly, China possesses vast real-world data from its hundreds of thousands of factories—the living laboratories where robots are being trained.
But vulnerabilities remain. China still trails in access to advanced AI chips for training and inference. Some critical sensors, like high-precision torque and force sensors, still rely on Western imports. And the devolved, local-government-driven development strategy risks duplication and waste—a pattern seen in the electric vehicle sector, where fierce inter-provincial competition produced staggering overcapacity even as it generated technological breakthroughs.
The Long Game
Beijing's bet on embodied AI serves multiple strategic goals. Domestically, it could boost productivity in manufacturing and logistics, addressing the economic slowdown. It could provide eldercare services for an aging population, easing the burden on working-age individuals. Militarily, embodied AI systems could enable autonomous warfare with unprecedented resilience to electronic interference - not mere order-executors, but commander-fighters capable of real-time tactical decisions.
Geoeconomically, if China becomes the world's leading supplier of embodied AI systems, it could create dependence on Chinese technology that surpasses reliance on 5G networks or solar panels. Beijing has already begun laying institutional groundwork for global diffusion through its Global AI Governance Action Plan and the newly proposed World AI Cooperation Organization.
And if embodied AI ultimately holds the key to AGI - intelligence emerging from grounded interaction with reality rather than statistical patterns in text—China could gain a decisive edge in the frontier AI competition with the United States.
What the Shock Means
The China shock in AI is not primarily about who builds the best chatbot. It's about two civilizations making fundamentally different bets on what intelligence is and how it should be developed. Silicon Valley is betting on scale and abstraction - that enough compute and enough data will produce general intelligence in digital form. Beijing is betting on embodiment and integration - that intelligence emerges from acting in the world, and that AI's value lies in transforming the real economy.
The outcome will shape not only the trajectory of artificial intelligence but also the balance of economic and military power in the decades ahead. China's vision may prove correct: that the path to AGI runs through factory floors and delivery routes rather than data centers and language models. Or the American bet on scaling may eventually pay off, producing the breakthrough that renders embodied approaches obsolete.
Either way, we are witnessing a genuine civilizational divergence in how to approach the most consequential technology of our time. The China shock is not just about competition. It's about two different answers to the question of what artificial intelligence is for—and what kind of future it should build.