From $7.5B to $12B: A Ballooning Round

DeepSeek is close to raising at least $12 billion in a new funding round, Bloomberg reported on October 6, with Tencent and battery giant CATL contributing the largest shares. Reuters says the total will exceed $12 billion, and people familiar with the matter suggest it could approach $15 billion. The original target was about $7.5 billion at a valuation of roughly $75 billion — investor appetite is not subtle.

The reported trigger is DeepSeek's V4-Flash model, which has set new cost-and-performance benchmarks against Anthropic and OpenAI. The round had been paused earlier after founder Liang Wenfeng's comments about the company's reliance on Nvidia chips went viral; it has clearly regained momentum.

Where the Money Goes: Huawei Chips and an IPO

Key Facts at a Glance

· New round: at least $12B, possibly near $15B; Tencent and CATL contribute most

· Original plan: about $7.5B at a valuation of about $75B

· Previous round: about $7.4B first external round in June, valuation above $50B

· Use of funds: a data center with at least 160,000 Huawei AI chips; an in-house inference chip

· Next step: post-round restructuring, targeting an early-2027 IPO

According to The Decoder, citing insiders, DeepSeek plans to restructure for an IPO in early 2027 once the round closes. The company is building a data center with at least 160,000 Huawei AI chips and developing its own inference chip to reduce dependence on both Nvidia and Huawei, while founder Liang Wenfeng insists on continuing to release open models with freely available weights.

CATL declined to comment; Tencent and DeepSeek did not respond to requests. The deal is not officially announced, and terms may still change. — Bloomberg, Reuters

The Capitalization Inflection of an Open-Source Icon

The company that shook the industry with ultra-low training costs is now converting its efficiency myth into hard capital: two external rounds within a year — potentially nearly $20 billion combined — will fund a twin sprint toward compute autonomy (the Huawei cluster) and public markets. On the demand side, relentless iteration of DeepSeek-family models keeps bending the inference cost curve downward — and NineZenith's Tianxing heterogeneous model routing and Tiansuan inference-cost governance exist precisely to help government and enterprise clients ride that falling curve dynamically.

(Compiled from public reports by Bloomberg, Reuters, The Decoder, Quartz, etc.)