Chinese AI company DeepSeek is nearing the close of a financing round of about 100 billion yuan (roughly US$14 billion), according to reports published on October 13, 2026 by IT Home and other outlets. The pre-investment valuation remains around 500 billion yuan, and the funds are expected to support the planned IPO in early 2027 and the build-out of computing infrastructure using Huawei Ascend chips.
What happened?
According to IT Home, citing Caijing, multiple transaction participants confirmed that DeepSeek’s second external round is almost complete. The amount rose from the initial plan of 50 billion to about 100 billion yuan in less than half a month. New and existing investors each contribute roughly half. New ones include Shixi Capital, SMIC Juyuan, CPE Yuanfeng and others; existing ones include CATL (planning to add about 10 billion), Tencent, IDG Capital and others.
The first round, completed in June 2026, raised about 51 billion yuan with a post-money valuation near 400 billion. Founder Liang Wenfeng personally contributed about 20 billion at that time. Earlier reports from Bloomberg and 21st Century Business Herald already pointed to 80-100 billion, led by Tencent and CATL, focused on preparing the IPO.
Part of the funds is expected to go to a large data center in Inner Mongolia, with plans to deploy at least 160,000 Huawei Ascend 950DT accelerators, forming one of the largest known Huawei AI chip clusters. This marks a bet on domestic Chinese computing for frontier model pre-training.
Why does this matter?
DeepSeek has stood out for high-performance models at relatively low costs, challenging Western leaders. This massive round, without sharply raising the valuation relative to the previous one, signals industrial investors’ confidence in a landscape of Nvidia chip restrictions. Participation by CATL (batteries and energy) and Tencent reinforces synergies between AI, energy and the Chinese digital ecosystem.
For the global AI market, it indicates China’s ability to mobilize large-scale capital for sovereign infrastructure, potentially reducing dependence on foreign hardware. The path to an IPO in 2027 on the Chinese market (possibly the Sci-Tech Innovation Board) would be a milestone for foundation model companies.
What changes in practice?
With more capital, DeepSeek can accelerate development of next-generation models and expand training and inference capacity. Users and developers may see more open-weight models or accessible APIs. However, concentration on Huawei chips brings software and efficiency challenges compared to Nvidia’s CUDA ecosystem.
There is still no official confirmation from the company on the round’s closing. Follow DeepSeek announcements and reports from IT Home, Caijing and Bloomberg for updates. Sources: IT Home (13/10/2026), Caijing, Bloomberg, 21st Century Business Herald.
By GeekikiBot
