HANGZHOU, China — Also known as China’s Silicon Valley, Hangzhou’s tech giants and startups are competing to develop cutting-edge chips, robotics and brain-computer interfaces. Across the city, ambitious founders are developing AI pets and fortune-telling apps. This city on China’s southeast coast is being reinvented as a hub for artificial intelligence. A year after the rise of DeepSeek captured global attention, Hangzhou’s rapidly changing AI scene reflects both its global ambitions and growing competition at home. Racing towards physical AI China and the United States are moving in lockstep toward developing physical applications that many see as the next frontier for AI. From Meta to Tencent, companies are racing to build “world models” that use AI to help robots navigate, pilotless cars, and even simulate real-world events like climate change. The Chinese government has positioned “embodied intelligence” as one of China’s priorities going forward in its next five-year plan. In November, the U.S.-China Economic and Security Review Commission, which makes recommendations to Congress, warned that China was rapidly advancing in physical AI and urged Washington to step up investment and regulatory approvals in autonomous systems and robotics. In Hangzhou, spatial intelligence firm Manycore and robot maker Unitree and Deep Robotics, one of the city’s tech startups known as the “Six Little Dragons,” are preparing to go public in Hong Kong and mainland China, joining a spate of AI-related listings. Large-scale language models are primarily trained on internet data to produce text and images, but training AI for the physical world requires additional inputs such as how objects interact, such as weight, texture, structure, temperature, and the grip and force needed to manipulate objects. Instead of manually collecting all data points, using AI techniques to calculate them makes the process much easier and faster, Manycore co-founder Victor Huang told CNBC. Huang, a former Nvidia software engineer, said his company uses the California-based company’s chips, which he says provide superior computing power per watt and deliver the best performance for the energy it consumes. But he believes China’s cheap energy could provide an advantage and offset the use of less advanced, more power-hungry chips. For example, a 3-nanometer chip can use about 30% less power than a 5-nanometer or 7-nanometer chip, but companies can remain competitive if electricity costs are 40% to 50% lower, he said. He added that computing power cannot be considered in isolation. Depends on data quality, energy supply and operating conditions. Placing data centers in colder regions saves energy, and cleaner data reduces the computing power required for training. Manycore has also open sourced its spatial AI model. This is the approach China is taking to many pay-as-you-go models common in the US, such as those offered by OpenAI and Anthropic. Huang said this approach allows the company to collect feedback, but limits direct revenue because users don’t have to pay for access. “So there will be pressure from investors,” he says. Indie AI scene takes shape Compared to Silicon Valley’s fascination with artificial general intelligence and superintelligence, China’s AI push is focused on practical applications, from Baidu Maps’ personalized recommendations to ByteDance’s chatbot and assistant Doubao. As of December, Doubao was China’s top AI app, with 155 million weekly active users, nearly twice as many as its closest competitor, DeepSeek’s chatbot, according to business intelligence service provider QuestMobile. Doubao’s popularity suggests that user experience and practicality may outweigh technical sophistication. Parallel to these efforts, a looser, more experimental culture is emerging. While Hangzhou’s powerhouses like Alibaba and DeepSeek are hard at work developing cutting-edge AI, a leafy, hipster-meets-hacker suburb called Liangzhu has become ground zero for AI’s more exotic side. Liangzhu residents call themselves “villagers” and are building everything from gamified fitness trackers to ADHD-friendly calendar tools. Alex Wei, an entrepreneur who moved to Liangzhu in 2025, is developing an AI app based on traditional Chinese fortune-telling tools. He is interested in how AI can meet people’s emotional needs. Part of its appeal is that Liangzhu is under less commercial pressure. “If you come to Liangzhu with 1,000 RMB ($143), you can take home a product demo,” Wei said. “This is a very inclusive place. You don’t have to have a unicorn product; even a small app serving 1,000 people will find support,” he added. The wave of innovation in Liangzhu is attracting the attention of investors. Monthly “Demo Days” that started in our living rooms now attract founders and supporters from across the region and beyond. This exposure shapes how startups think about scale. Many are eyeing overseas users, with some founders hoping to leverage China’s hardware supply chain to offer competitive prices around the world. Observers say fierce domestic competition and consumers’ reluctance to pay for apps are also encouraging startups to look overseas. Aphra Wang, who writes Concurrent’s newsletter on China and Silicon Valley, said some developers are leveraging AI to avoid traditional career paths in a changing job market and instead aim to become “hyperindividuals” building profitable businesses independently or with small teams. But Wang warned that some companies are simply patching in AI features for marketing appeal, from air conditioners to mirrors that check sunscreen application. She calls some of these “physics AI slops.” This is a play on the term used to describe low-quality content produced by generative AI. Today, Hangzhou entrepreneurs are experimenting with almost every idea, from efficiency to comfort, serious to frivolous, in a rapidly evolving market.
