In the late 19th century, Standard Oil’s founders quietly bought stakes in railways to secure cheap, safe transport for their oil and a captive customer for lubricant. The same move is playing out today, with CATL buying seats at the AI table before demand explodes.
Battery money meets AI
On 6 October, Bloomberg reported DeepSeek’s new round at RMB 80 billion to RMB 100 billion, with Tencent and CATL again among the largest backers, only four months after CATL first appeared on DeepSeek’s cap table. CATL, the power-battery leader, once aimed its capital at mines, materials and carmakers. Now its reach extends to large models, data centres, compute operation and embodied intelligence, through names like Qianxun, Galaxy General, Songyan, Zhongqing and Vita, plus compute plays and energy-AI startups. In September it committed RMB 1.62 billion to two funds explicitly focused on AI and embodied intelligence.
An energy operating system
The International Energy Agency estimates data-centre power at 415 TWh in 2024, near doubling to 945 TWh by 2030, mostly AI-driven. Humanoid runtimes of two to four hours need denser, longer-cycle cells, and AI data centres are adding lithium backup power. Founder Zeng Yuqun says CATL will not compete with AI firms but contribute to their energy systems, aiming at an intelligent-age energy platform: green power and chips upstream, supply and storage midstream, compute operation and robots downstream, plus a battery bank and recycle loop.
The pick-and-shovel limit
But a stake is not a purchase order. Robot firms still compare weight, power, safety, life and cost; data centres still weigh reliability and lifetime cost. The auto industry already showed how a customer can become a rival. And demand timing is unsure: embodied intelligence in 2026 is still more prototype than volume, and AI capital spending is cyclical. The old question remains, for batteries and for AI: should the shovel be bound to the gold mine, and how.
Editor’s note: This is an adapted translation of the original Sohu report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://www.sohu.com/a/1087742644_120816185.