DeepSeek Reportedly Nears $12 Billion in Tencent-Backed Funding Round

Liang Wenfeng’s DeepSeek was close to raising at least 80 billion yuan, about $12 billion, as funding talks neared an end. People familiar with the matter told Bloomberg on Tuesday, October 6, 2026, that the conversations were near enough for a close to be in reach, open enough that nothing had locked, and still subject to change on the final terms. Signed term sheets — preliminary agreements that set out the main terms of an investment — already pointed past that floor. Based on those sheets, the final tally could approach 100 billion yuan, roughly $15 billion. The round reportedly valued the company at about 500 billion yuan pre-money, the price investors were using before their new capital was added, or roughly $74 billion to $75 billion.

Lead investors included Tencent Holdings, the company behind WeChat, and Contemporary Amperex Technology Co., the battery pioneer known as CATL. Both had committed among the largest amounts. Tencent already knew the ground. In a prior round it had written the biggest check of any investor: 10 billion yuan. Now the WeChat operator and CATL were back among the heaviest participants. When the Bloomberg account appeared, a CATL representative declined to comment and Tencent had no immediate comment. DeepSeek did not respond to an emailed request during a Chinese public holiday.

Behind the talks lay what the company had already raised under Liang. The prior round closed with approximately $7 billion to $7.4 billion and a post-money valuation of more than $50 billion, about 350 billion yuan. Liang contributed approximately $3 billion of his own assets. The check came after High-Flyer, the hedge fund he had co-founded, had carried the company as its principal backer through the early years. Outside investors had only just been invited in.

By mid-July 2026 the company’s IPO preparations became known. Talks with banks and auditors were already under way, and a public path was taking shape even while private capital remained necessary to keep pace with the scale of model work still ahead. When the next raise opened, its initial target sat at about 50 billion yuan, roughly $7.4 billion to $7.5 billion, a figure almost identical to what the prior round had brought in. Demand was about to rewrite it.

In September 2026 DeepSeek released DeepSeek-V4.1-Flash, its latest large language model. A large language model is an artificial intelligence system trained on enormous collections of text so that it can generate language, answer questions, and carry out complex written tasks. The new model intensified competition with OpenAI and Anthropic, and the September release became the catalyst for the surge in investor demand that followed. An initial target of about 50 billion yuan gave way to term sheets that could near 100 billion.

The capital still had to find its hardware, and the software that would run on it. DeepSeek partnered with Huawei Technologies to build programming tools for Huawei’s Ascend AI chips. An Ascend chip is a domestic accelerator built to handle the dense mathematical work of training and running large artificial intelligence models. Programming tools sit between a researcher’s code and the silicon, translating high-level instructions into the precise operations the chip can execute quickly and in volume. The partnership joined software to hardware on home ground. DeepSeek would help shape tools that let models run on Ascend processors rather than on accelerators designed abroad. As competition with U.S. laboratories hardened, that link became the condition that made new capital useful. Without software matched to the chips, money could not become working capacity. Code fitted to domestic silicon turned the raise into infrastructure.

DeepSeek hired CITIC Securities to prepare for a listing on Shanghai’s STAR Market. CITIC Securities is a Chinese securities firm brought on to guide the company through the steps of a public offering. An initial public offering, or IPO, is the first sale of a private company’s shares to public investors on an exchange. The STAR Market is the board of the Shanghai Stock Exchange built for science and technology companies. The planned debut was set for early 2027. DeepSeek needed the IPO primarily to finance further growth. Developing new AI models required increasingly large investments in computing power, data centers and talent. Once the round closed, the next step was expected to be a restructuring of the company to prepare for the offering. The timing, valuation and size of the eventual IPO had not been decided.

Until then the money had to become the hardware the models required. DeepSeek planned to deploy at least 160,000 Huawei AI accelerators at a new data center in Inner Mongolia. An accelerator is a specialized chip designed for the dense mathematical operations large language models need during training and when generating answers. A data center is the facility built to house and power large numbers of such chips so they can operate together. Gathering 160,000 accelerators inside one facility creates a computing cluster, processors working in parallel on shared workloads at a scale few installations reach. The new funding was intended to expand that capacity. The Inner Mongolia cluster could become one of the largest known assemblies of Huawei AI chips.

At least 160,000 Huawei AI accelerators would stand at the site in Inner Mongolia.