China Is Running Two AI Races at Once, and Winning One of Them

A factory somewhere in China is producing a humanoid robot every thirty minutes.[¹] That figure — roughly ten thousand units a year — circulated widely this week, and the reaction was less alarm than a kind of stunned recalibration. The conversation around China and AI has spent years organized around the question of catching up: Is China closing the gap? How fast? On what timeline? The robotics data suggests that framing is already obsolete, at least in one domain. The question isn't whether China is closing the gap in humanoid robotics — it's whether anyone else is still in the same race.

AGIBOT's X2 Ultra drew crowds in Hong Kong this week, and AI robotics observers on Bluesky noted that Unitree's H1 has now hit sprint speeds approaching Usain Bolt's 2009 world record.[²] These aren't research demos. They're products, and they're being manufactured at scale. The Stanford AI Index drew a similar observation: China has caught up in robotics and patents, and the US lead — once treated as structural — has effectively vanished in this category.[³] What's interesting about how this lands in the conversation is that the celebratory posts and the anxious ones are describing the same facts. The people cheering the robot half-marathon and the people warning about geopolitical implications are reading the same press releases.

The military picture is more contested — and the discourse reflects that ambiguity. US intelligence chiefs called China the biggest military and cyber threat the country faces.[⁴] An Air Force general argued China is leaving the US behind on AI.[⁵] And yet the US and China both opted out of a joint declaration on military AI use,[⁶] which is either a sign of principled sovereignty or a sign that neither side wants constraints. Analysts writing for outlets like the BBC and South China Morning Post have landed on a more careful read: China is winning certain AI races, the US is winning others, and the competition has enough mutual dependency baked in that the cold war framing may be doing more rhetorical than analytical work.[⁷] The phrase "co-opetition" appeared in serious coverage this week — not as a dismissal of the rivalry but as a description of its actual structure.

The education angle adds a different texture to how China shows up in the conversation. China's National Data Administration published an action plan for AI in schools that would have AI writing lesson plans and grading homework.[⁸] One observer put it more starkly: "The student, the teacher, and the grader are now one system talking to itself." Meanwhile, students in Chinese exam halls are reportedly renting AI-equipped glasses for six dollars a day to scan questions and receive answers in real time.[⁹] These two stories — top-down AI integration into curriculum design, bottom-up AI cheating on exams — sit in odd tension with each other, and neither the state nor the critics seem to have fully reckoned with what happens when both are true simultaneously.

What the discourse keeps circling without quite landing on is that China functions as several different arguments at once depending on who's making the case. For military hawks, it's a threat that demands accelerated US investment. For robotics enthusiasts, it's proof that AI hardware ambitions can be industrialized at speed. For AI ethics critics, its education and surveillance applications are the cautionary tale. For open-source advocates, Chinese labs releasing competitive models complicate the narrative that openness is a Western democratic virtue. China isn't one thing in this conversation — it's the surface onto which everyone projects the AI future they're most worried about, or most excited by. That's why it keeps appearing everywhere. Not because China is uniquely central to AI development, but because it's uniquely useful as a mirror.