DeepSeek V4 Flash Fuels $12B Raise, 160K Huawei Ascend Cluster

For AI engineers and researchers, the money matters because of where it goes: at least 160,000 Huawei accelerators in a single Inner Mongolia data center, plus publicly released software to program Huawei AI chips. That combination gives DeepSeek a domestic hardware pipeline that could decouple its cost-performance gains from Nvidia supply constraints — a critical test of the Chinese AI stack.

DeepSeek, the Hangzhou-based artificial intelligence startup that has become synonymous with cost-efficient frontier models, is close to securing at least 80 billion yuan ($12 billion) in fresh funding, according to people familiar with the matter cited by Bloomberg. The round, which is expected to close soon, already far exceeds DeepSeek's initial target of about 50 billion yuan, and term sheets signed with investors suggest the final tally could approach 100 billion yuan. Contemporary Amperex Technology Co. Ltd. (CATL), the battery pioneer, and Tencent Holdings Ltd., operator of WeChat, have committed among the largest amounts, giving DeepSeek a mix of industrial and internet-platform backers ahead of what is expected to be a landmark initial public offering in early 2027.

“

The company has not disclosed a figure, but Moonshot AI, a domestic rival, recently closed funding at a $50 billion valuation and is also targeting an early 2027 IPO.

The fundraising surge follows the release of DeepSeek's V4 Flash model, which reset expectations for cost versus performance against frontier leaders Anthropic and OpenAI. That release capped a strong run for founder Liang Wenfeng, whose startup gained global attention in 2025 with a model that matched the capabilities of leading Silicon Valley offerings at a fraction of the cost. In this context, the oversubscription is not merely a vote of confidence in DeepSeek's research talent but also a bet that its efficiency-first approach can be sustained at much larger scale.

Capital is the binding constraint for AI development, and DeepSeek has made clear it intends to spend aggressively on infrastructure. The company plans to deploy at least 160,000 of Huawei Technologies Co.'s top accelerators at a massive data center it is building in Inner Mongolia, which would create one of the largest known clusters of Huawei AI chips. DeepSeek has also publicly released software created with Huawei to program AI chips, signaling a deeper partnership with the hardware leader and a potential path to reduce reliance on Nvidia's graphics processors. That alignment matters strategically: it links DeepSeek's model performance to a domestic hardware supply chain at a time when U.S. export controls have constrained Chinese access to cutting-edge semiconductors.

The size of the round, if it lands near 100 billion yuan, would rank among the largest private AI financings in China and would give DeepSeek a substantial pre-IPO cash cushion. It also raises questions about valuation. The company has not disclosed a figure, but Moonshot AI, a domestic rival, recently closed funding at a $50 billion valuation and is also targeting an early 2027 IPO. That parallel creates a crowded calendar of Chinese AI listings that will test global investor appetite. Tencent's participation is notable given its own cloud and enterprise AI ambitions, and CATL's involvement shows how industrial giants are seeking exposure to AI infrastructure beyond their core businesses.