- Battery giant said to join AI startup’s $73.5 billion financing round
- Deal would deepen convergence between energy infrastructure and large-scale computing
CATL is reportedly preparing to take part in the first financing round of DeepSeek, in what would mark a fresh crossover between the global lithium-battery giant and one of China’s most closely watched AI startups, according to The Information on May 22, citing people familiar with the matter.
The round is targeting about 50 billion yuan ($7.36 billion), with a potential close as early as June. If completed, DeepSeek’s valuation could exceed 350 billion yuan, with some scenarios cited by investors placing it as high as $45 billion depending on the final structure.
Other potential backers include e-commerce heavyweight JD.com and online gaming titan NetEase, while Tencent and Alibaba Group have also been previously reported to be in talks.
China’s state-backed China Integrated Circuit Industry Investment Fund is also said to be discussing a leading role in the injection that could lift the valuation closer to $45 billion.
For CATL, the potential investment comes as it accelerates its push into AI infrastructure.
Growing appetite for ‘AI plus energy’ deals
Over the past six weeks, the company has committed roughly 10.5 billion yuan into “AI plus energy” infrastructure deals, including a 4.1 billion yuan stake in Hangzhou Zhongheng Electric Co.,Ltd. (中恒电气), a domestic leader in high-voltage direct current (HVDC) equipment, in April, and a 6.4 billion yuan acquisition of about 38.1% of data center operator VNet (世纪互联) in May.
The moves point to a broader shift underway at CATL — from battery manufacturing into what it has described as energy infrastructure for the computing era.
Founder Robin Zeng has previously told Reuters that the company aims to build large-scale standalone energy systems capable of powering hyperscale data centers and, in some cases, entire cities.
DeepSeek, founded in 2023 by quantitative trading firm High-Flyer, drew global attention in early 2025 after its R1 model achieved competitive performance at a fraction of the training cost of US peers.
The company remains committed to open-source development, and its latest V4 model has been optimized for inference on Huawei’s Ascend 950PR chips.
The startup is also expanding into heavy infrastructure, including plans for a large data center in Ulanqab, Inner Mongolia.
Analysts expect it to accelerate development of its core “Code Harness” initiative — a system benchmarked against Anthropic’s Claude Code — as it scales compute capacity and seeks to attract and retain AI talent.
CATL, DeepSeek and the other parties involved did not immediately respond to requests for media comment. Deal size and final participants remain subject to change.
