Zhou Yuxiang: AI's Google moment not here yet

Zhou's path to Black Lake ran through failure. His first startup, MadaData, tried to sit on top of factory data much like Palantir does for enterprise and government clients, but it collapsed because most factories he visited had zero data. He was saved by a group of second-generation factory owners who had taken over family businesses and faced a market shifting toward smaller, faster orders. That sent Zhou onto a plastic-injection line at a factory making cosmetics for a European giant, where he timed a six-hour changeover and found only one hour was useful. The rest, he said, was lost to subjective decision-making and information silos.

The broader setting is a China that has become a major source of AI innovation, with open-weight models from Alibaba, DeepSeek, Z.ai, and Moonshot AI delivering near-frontier performance at a fraction of closed Western rivals' cost. China installed 354,000 industrial robots in 2025, a record 59% of the world's total, according to the International Federation of Robotics. Yet Zhou flagged constraints: U.S. export controls deny China frontier AI processors, and funding pools are shallower than those of OpenAI and Anthropic. His bigger concern, he said, is internal—time.

Zhou also pushed back on the idea that China can skip the software era and go straight to AI, the way it skipped credit cards for mobile payments. Digitization, he said, only became a recognized concept for Chinese factory owners about 10 years ago, compared with 40-50 years in the West. Whether his bet on AI restructuring industry pays off may hinge on whether data—and the habits to accumulate it—can be built in a fraction of that time.