History has a habit of concealing its most consequential turning points inside the mundane. The invention of the shipping container did not announce itself as a revolution in global trade. The rise of the mobile phone was initially dismissed as an expensive novelty for businessmen.<\/p>\n
Today, something similarly quiet and similarly transformative is unfolding across the parking lots, highways, and residential streets of China\u2019s sprawling cities. Forty million EVs, each sitting idle for roughly 23 hours a day, are poised to become something the world has not yet named: the most distributed artificial intelligence infrastructure ever assembled.<\/p>\n
The observation did not come from a futurist or a technology journalist. It came from Robin Zeng<\/a>, the founder and chairman of CATL<\/a>, the world\u2019s largest EV battery maker, speaking at the World Economic Forum\u2019s Summer Davos<\/a> in Dalian last week.<\/p>\n China\u2019s vast EV fleet, Zeng argued, could be reimagined as distributed token factories: computing infrastructure that uses onboard batteries and AI chips to produce the outputs feeding large language models at scale.<\/p>\n It was arguably the week\u2019s most underreported remark. It deserves to be the most studied.<\/p>\n Misread story from the start<\/p>\n For the better part of a decade, the global conversation about China\u2019s EV industry has been dominated by a single, reductive framing of it as a trade threat. Western governments have debated tariffs, imposed import duties and commissioned studies on the industrial displacement that Chinese EVs might cause.<\/p>\n The European Union levied additional duties. The United States raised its own. The underlying assumption, almost universally shared, was that China\u2019s EV success was a manufacturing story with trade consequences.<\/p>\n A dispassionate reading of the evidence, however, suggests this framing has always been incomplete. China\u2019s drive to electrify its transport system was not, at its core, a plan to conquer foreign car markets. Rather, it was a national energy security calculation.<\/p>\n A country that imports the majority of its crude oil<\/a> and has watched energy supply chains weaponized amid geopolitical rivalry wanted a transport fleet that ran on electricity it could generate domestically.<\/p>\n What that imperative produced, as an elegant byproduct, was the world\u2019s largest networked fleet of mobile energy storage assets: 40 million batteries on wheels, all connected to an increasingly intelligent charging infrastructure.<\/p>\n Clean energy sectors<\/a> drove more than a third of China\u2019s GDP growth in 2025. The \u201cnew three,\u201d meaning EVs, batteries and solar panels, generated two-thirds of the value added across the entire clean energy sector.<\/p>\n Battery exports grew 41% year-on-year. These numbers have been reported widely, but almost exclusively through a trade-and-competition lens. What has been missed is the deeper infrastructure logic that the fleet represents.<\/p>\n Platform as much as product<\/p>\n Modern electric vehicles in China, whether built by BYD, Nio, Xpeng or dozens of other domestic manufacturers, carry substantial onboard processing capability. Their chips handle navigation, driver assistance, over-the-air software updates, and increasingly sophisticated AI functions.<\/p>\n For roughly 23 hours every day, that capability sits dormant. Zeng\u2019s insight is that this dormant compute capacity, aggregated across tens of millions of vehicles simultaneously, constitutes a distributed AI processing layer that already exists, costs nothing additional to build and requires no new land, no new permits and no new centralized data centers to operate.<\/p>\n The technical mechanism that makes this vision coherent is vehicle-to-grid technology<\/a>, or V2G, which allows parked EVs to discharge stored electricity back into the grid during peak demand. China has been building the regulatory and physical architecture for V2G since 2024.<\/p>\n By the end of 2027, the country plans to have 28 million charging facilities<\/a> and 5,000 bidirectional stations operational. Chinese officials project that a fleet of 100 million EVs<\/a> by 2030, if networked bidirectionally, could unlock one billion kilowatts of flexible energy capacity.<\/p>\n Zeng\u2019s token factory vision extends that logic one critical step further: the vehicle gives back not only electricity but compute.<\/p>\n This is not the first time China has extracted multiple economic functions from the same fixed investment. Its high-speed rail network was built for passengers and became a logistics and regional integration tool.<\/p>\n Its renewable energy infrastructure was designed for energy security and became a globally dominant export industry. The pattern is consistent and deliberate: absorb the capital cost once, then extract value across multiple domains across time.<\/p>\n Capital follows the vision<\/p>\n What elevates Zeng\u2019s Dalian remarks from visionary speculation to credible industrial strategy is the investment trail CATL has left in the preceding months.<\/p>\n In April 2026, CATL invested approximately $600 million<\/a> for a 49% stake in Zhongheng Electric\u2019s controlling shareholder, one of China\u2019s primary providers of high-voltage direct-current power systems for AI data centers.<\/p>\n In May, a CATL-affiliated fund committed up to $942 million<\/a> to acquire 38.1% of VNET Group, a major data center operator. In June, TechNode reported<\/a> that CATL committed approximately $740 million to DeepSeek\u2019s $7.4 billion first external funding round, making the battery giant one of the two largest outside investors in China\u2019s most consequential AI laboratory.<\/p>\n These three transactions, totaling roughly $1.5 billion across two months, form a vertical chain: battery storage upstream, power conversion in the midstream, data center operations downstream and AI model development at the end.<\/p>\n