CATL Considers Stake in DeepSeek Amid $7.4 Billion AI Funding Push

Context CATL supplies lithium‑ion batteries for electric vehicles and is expanding into power‑supply equipment for AI data centers. DeepSeek gained attention after releasing its open‑source large language model R1 in January 2025, which was trained at lower compute cost than comparable U.S. models. The startup is now looking to build its own computing infrastructure to support faster model releases and reduce reliance on third‑party cloud providers. DeepSeek says its latest V4 model is optimized for inference on Huawei’s Ascend 950PR chips, showing a tight coupling of software and hardware. The company plans to release V4.1 in June to shorten its update cycle toward industry norms.

Key Facts DeepSeek’s funding round targets about 50 billion yuan, equivalent to roughly $7.36 billion, and may close by June. If the round succeeds, the startup’s valuation could exceed 350 billion yuan, or about $48 billion. Separate job postings show DeepSeek plans to construct a large‑scale data center in Ulanqab, Inner Mongolia, a region chosen for abundant electricity and low temperatures that cut cooling costs.

What It Means A CATL investment would give the battery maker a foothold in the AI hardware market, aligning its power‑supply business with the growing energy needs of large language models. For DeepSeek, securing the funding would accelerate its data‑center build‑out, potentially improving model training speed and lowering operating expenses. The move also reflects a broader trend of Chinese tech firms integrating chip, hardware, and software development to compete globally. Analysts note that owning a data center could reduce latency and improve control over model iteration, though capital expenditures remain high. The AI infrastructure race in China is attracting interest from battery makers, telecom firms, and semiconductor funds seeking diversified growth.

What to watch next Investors will monitor whether CATL finalizes its stake, the exact timing of DeepSeek’s funding close, and progress on the Ulanqab data center ahead of the planned V4.1 model release in June.