OpenAI Is Getting Into Robotics. China Already Got There First.
OpenAI Is Getting Into Robotics. China Already Got There First.
When OpenAI posted on May 31 that it was hiring engineers to build humanoid robots, the market read it as a competitive threat to Tesla. Tesla stock fell 4.6% that day, according to Investor's Business Daily. But the premise of that trade — that Tesla's Optimus robot faces a two-company race — does not survive contact with the actual shipment data.
China sold roughly 90% of all humanoid robots globally in 2025. Unitree, based in Hangzhou, moved 5,500 units, making it the world's top seller by volume. Shanghai-based Agibot moved 5,168. Tesla's production target for Optimus that same year was 5,000 units. Tesla did not meet it. Two Chinese companies each outsold the production target of the world's most recognizable robotics brand before Tesla shipped a single commercial unit.
The competitive framing Tesla investors are betting on reverses the actual order of operations. OpenAI has no product, no prototype, and no timeline beyond a recruiting post. Tesla has a product that does not yet ship at commercial scale. China has both: volume manufacturing and a growing installed base accumulating real-world deployment hours.
Musk said at the World Economic Forum in 2026 that China is very good at AI and very good at manufacturing, and will definitely be the toughest competition for Tesla. That was not a throwaway line. It was the most accurate short summary of the humanoid robot market available.
Tesla's Shanghai factory delivered 85,982 cars in May 2026, up 39% year-over-year, providing the revenue base that funds Optimus development. The Optimus 3 production timeline calls for initial output starting Summer 2026 with high-volume production by Summer 2027, according to Musk. Cantor Fitzgerald projects first deliveries in Q3 2027. OpenAI has announced no comparable schedule.
The market projections are large and speculative: a projected $38 billion by 2035 and $5 trillion by 2050, estimates that depend on both technical execution and whether humanoid robots find durable commercial applications at scale. Those projections are directionally consistent across analysts but vary widely in their assumptions about adoption speed.
The honest version of this story is not a three-way horse race. It is a two-layered bet: whether humanoid robots become a real product category at all, and whether the first movers in that category, Chinese manufacturers building and shipping today, will be displaced or complemented by whatever Tesla and OpenAI eventually bring to market. The 5,500 Unitree robots sold in 2025 are not prototypes in a lab. They are units in the field, accumulating hours.
What to watch: whether Tesla's 2027 production timeline holds, whether OpenAI makes a hardware acquisition to close the gap between its software capability and physical product, and whether Chinese exporters begin targeting the same warehouse and logistics customers Tesla has named as its initial Optimus market. The race is not between Tesla and OpenAI. It is between everyone and the question of whether a humanoid robot market exists at scale, a question Unitree and Agibot are already answering with shipment data, not projections.