Embodied AI vs AGI is a framing advanced by Nita Farahany in Class 25 of her introductory course (November 30, 2025) describing what she characterizes as China's strategic move to redirect the global AI race away from the US-defined finish line of AGI, or disembodied superintelligence, toward AI embedded in physical systems, where she argues China is positioned to dominate. The framing reads China's investment pattern as an attempt to change which race matters rather than to catch up in the frontier-model race led by US labs.
Background
The conventional "AI race" framing assumes the contestants are racing toward AGI, with the United States in the lead through frontier-model capabilities. Farahany cites David Sacks as representative of that view: "China is not years and years behind us in AI. Maybe they're three to six months" (Inside My AI Law & Policy Class 25: China's AI Governance — Are Different Players Running Different Races? (Farahany, November 2025)).
Against that framing, Farahany poses whether China is trying not to catch up in the race the United States is running but to change which race matters. On her reading, while American discourse focuses on AGI, China's investments point to embodied AI as the priority: robots, autonomous vehicles, and smart manufacturing systems, or AI doing things in the physical world (Inside My AI Law & Policy Class 25: China's AI Governance — Are Different Players Running Different Races? (Farahany, November 2025)).
Evidence cited
Farahany draws on several quantitative indicators of China's position in physical automation (Inside My AI Law & Policy Class 25: China's AI Governance — Are Different Players Running Different Races? (Farahany, November 2025)):
- Industrial robot installations, 2024: China installed approximately 295,000 industrial robots; the United States installed approximately 50,100. China accounted for roughly half of all industrial robots installed worldwide that year.
- AI+ Manufacturing integration (China Ministry of Industry and Information Technology): more than 60% of large Chinese manufacturers had some form of AI+ Manufacturing integration by the end of 2025.
- 14th Five-Year Plan targets: comprehensive intelligent transformation, with AI embedded in key sectors reaching 70% by 2027, 90% by 2030, and 100% by 2035.
- CloudMatrix 384 (Huawei): a system clustering 384 domestic Ascend 910C chips for approximately 300 petaflops, which Farahany describes as comparable to roughly 300 H100s, achieved through architectural innovation rather than raw chip parity, illustrating an emphasis on scale and efficiency over per-chip performance.
Relation to world models
Farahany connects the embodied-AI thesis to world models, which under a framing she attributes to NVIDIA are increasingly seen as central to robotics and self-driving cars. On this view, if world models matter more than chatbots for the long-term economic and military deployment of AI, and if the future of AI is less about chatbots than about physical automation across manufacturing, logistics, agriculture, and infrastructure, then China may be building the foundation to lead the deployment that matters regardless of which country reaches AGI first (Inside My AI Law & Policy Class 25: China's AI Governance — Are Different Players Running Different Races? (Farahany, November 2025)).
Place in China's countermoves
Farahany presents the embodied-AI focus as one of three coordinated Chinese countermoves rather than a standalone strategy (Inside My AI Law & Policy Class 25: China's AI Governance — Are Different Players Running Different Races? (Farahany, November 2025)):
- Compensate through efficiency and scale, exemplified by DeepSeek and CloudMatrix 384.
- Control inputs through rare-earth leverage. In October 2025, Announcement No. 61 added rare-earth processing technology to China's export-control list with a foreign-direct-product rule. Chinese permanent magnets, which Farahany cites as 94% of global production, are used in data-center cooling, precision-guided weapons, and EV motors.
- Redirect the race toward embodied AI, the subject of this page.
Farahany argues that the three moves together position China to be difficult to bypass across several dimensions rather than dominant in any single race (see Three Theories of Victory (US / EU / China AI Governance)).
US framing, per Farahany
Farahany argues that the United States has framed the AI race as the AGI race because that is where US frontier labs, including OpenAI, Anthropic, and Google DeepMind, are leading. She contends this framing:
- Underweights physical-automation deployment as the metric that determines economic and military outcomes.
- Treats robotics, autonomous vehicles, and smart manufacturing as separate from AI policy rather than central to it.
- Underweights world-model research relative to LLM scaling.
- Misses an asymmetry she describes: an advance like DeepSeek can be matched by frontier-lab investment within months, whereas embodied-AI infrastructure deployment cannot.
Relationships
- introduced-by: Inside My AI Law & Policy Class 25: China's AI Governance — Are Different Players Running Different Races? (Farahany, November 2025)
- part-of: Three Theories of Victory (US / EU / China AI Governance) (China's theory of victory)
- related: Artificial General Intelligence (AGI), World Models, Huawei — Ascend AI Accelerators, DeepSeek, US-China AI Competition: Different Races, Different Metrics
- instance-of: AI Policy (umbrella)