DeepSeek, the Chinese AI lab that offers its models through web and mobile services and a developer API, is close to securing at least $12 billion in its latest financing, Bloomberg reported Tuesday in a story republished by Moneycontrol. Tencent and battery maker CATL have made two of the largest commitments. The deal hasn’t closed, and the report said terms could still change.
People can use DeepSeek directly, while software teams connect its models to their own products through an API billed by token usage. The hosted service handles the model responses; developers provide prompts and can build around visual input and tool calls. That creates a recurring, usage-based business alongside DeepSeek’s consumer-facing services.
DeepSeek’s newest V4.1 Flash model, released September 10, is available through the API with visual input and lower prices. The company says its architecture sharply reduces the memory needed for cached context, which matters when agents reuse long histories. Lower serving costs help attract developers, but they don’t eliminate DeepSeek’s own need for chips and data centers.
That capital demand is becoming more concrete. Bloomberg reported that DeepSeek plans to deploy at least 160,000 of Huawei’s top accelerators in an Inner Mongolia data center and has released software developed with Huawei to program AI chips. The planned cluster puts the financing in context: efficiency at the model layer still sits on top of costly physical infrastructure.
Bloomberg said signed term sheets support the reported $12 billion minimum, although DeepSeek did not respond during a Chinese public holiday and Tencent had no immediate comment. The round would follow DeepSeek’s first external financing this summer, after High-Flyer funded its early development. People cited by Bloomberg said a restructuring for an early-2027 IPO could come after the financing closes.
Published versions remain available when an article is updated or corrected.
- Revision 1 · Initial publication
Initial publication