Hangzhou is accelerating its shift toward artificial intelligence focused on physical applications, with major companies and startups developing chips, robotics, and brain computer interfaces alongside consumer-facing AI tools, as China positions the city as a key hub in its next phase of AI development.
Embodied Intelligence Becomes Policy Priority
The city’s transformation comes as China and the United States pursue what many researchers describe as the next stage of AI, where systems interact with the physical world. Companies including Meta and Tencent are building so-called world models designed to guide robots, autonomous vehicles, and simulations of real-world events.
Beijing has identified “embodied intelligence” as a priority area in its upcoming Five-Year Plan. In November, the U.S.-China Economic and Security Review Commission warned that China is advancing quickly in physical AI and urged Washington to increase investment and speed up regulatory approvals for robotics and autonomous systems.
Hangzhou Startups Prepare For Public Listings
In Hangzhou, spatial intelligence company Manycore, along with robotics companies Unitree and Deep Robotics, are preparing for initial public offerings in Hong Kong or mainland China. The firms are part of a group of local startups known as the “six little dragons,” amid a broader wave of AI-related listings.
Unlike large language models trained mainly on internet text and images, AI systems designed for physical environments require data on weight, texture, temperature, structure, and object interaction. This includes understanding grip strength and force needed to handle items.
Energy And Computing Trade-Offs
Manycore cofounder Victor Huang, a former Nvidia software engineer, said AI-based methods can calculate many of these physical variables faster than manual data collection. He said Manycore uses Nvidia chips because of their performance per watt but believes China’s lower electricity costs can offset the use of less advanced, more energy-intensive processors.
Huang said a three-nanometer chip can reduce power use by about 30% compared with five- or seven-nanometer chips, but lower energy prices of 40% to 50% can help maintain competitiveness. He added that computing performance depends on data quality, power supply, and operating conditions, with colder data center locations reducing energy use and cleaner datasets lowering training demands.
Open Models And Investor Pressure
Manycore has open-sourced its spatial AI model, reflecting a broader Chinese preference for open approaches compared with subscription-based models offered by US companies such as OpenAI and Anthropic. Huang said open sourcing allows faster feedback and adoption but limits direct revenue, increasing pressure from investors.
Practical Focus Shapes China’s AI Apps
China’s AI development has largely centered on practical consumer uses rather than abstract intelligence goals. Applications range from personalized navigation in Baidu Maps to chatbots such as Doubao, developed by ByteDance.
According to QuestMobile, Doubao recorded 155 million weekly active users in December, nearly double the user base of DeepSeek’s chatbot. The figures suggest usability and experience are driving adoption.
Liangzhu Emerges As Experimental Subculture
Alongside Hangzhou’s major firms, a suburban district called Liangzhu has developed into a testing ground for experimental AI projects. The area hosts independent developers building products such as AI pets, fitness tools, and productivity apps.
Entrepreneur Alex Wei, who moved to Liangzhu in 2025, is developing an AI application based on traditional Chinese divination practices, focusing on emotional support use cases. He said the area’s low costs and limited commercial pressure allow developers to prototype quickly with minimal funding.
Investor Interest And Overseas Ambitions
Liangzhu’s informal innovation scene has attracted investors through a monthly Demo Day that has grown from a private gathering into a regional event. Exposure to backers is influencing how startups approach scale, with many targeting overseas markets.
Observers say intense competition at home and reluctance among Chinese consumers to pay for apps are pushing founders to look abroad, often leveraging China’s hardware supply chains to compete on price.
Individual Builders And Market Saturation
Writer Afra Wang, who publishes the Concurrent newsletter, said some developers are using AI to bypass traditional career paths and build small, independent businesses. She cautioned that some products rely on superficial AI features for marketing purposes, applying the term “physical AI slop” to low-quality implementations.
Hangzhou’s ecosystem continues to test a wide range of ideas, from industrial efficiency to personal comfort, as companies and individuals navigate a crowded and rapidly shifting AI market.
Featured image credits: Flickr
For more stories like it, click the +Follow button at the top of this page to follow us.