Deepseek’s Massive Fundraise Gains CATL and Tencent Backing as AI Startup Targets 2027 IPO

Headline: Deepseek Nears $12 Billion Raise as V4-Flash and 160,000 Chips Shape IPO

Meta description: Deepseek nears a $12 billion funding round, with talks at $15 billion, as V4-Flash results and 160,000 Huawei chips shape early 2027 IPO plans.

Tags: Deepseek, AI funding, V4-Flash, IPO, Huawei chips

Featured image: abstract technology style, glowing server racks and chip patterns, no real people, no company logos, no text

Deepseek is close to raising at least $12 billion in a new funding round, Bloomberg reported, and the total could approach $15 billion. Investor interest is tied to V4-Flash, which set new benchmarks for cost and performance against Anthropic and OpenAI, while the company plans to restructure for an IPO in early 2027 and is building a data center with at least 160,000 AI chips from Huawei.

Background: A pause after chip dependence became public

The current round follows a pause after comments by founder Liang Wenfeng about Deepseek's reliance on Nvidia chips went viral. That episode put the company's supply constraints into view. For an AI lab, training and serving models depends on access to powerful processors, and the market's reaction to such comments can affect fundraising.

The original target was smaller: about $7.5 billion at a valuation of roughly $75 billion. The latest reported size is larger, and CATL and Tencent are contributing the largest shares. The scale of the new round therefore signals more than a routine bridge financing; it points to a company trying to convert model attention into a long-term infrastructure position.

Development: From April model previews to a June round

Deepseek released V4-Pro and V4-Flash preview versions in April. The source highlights V4-Flash because it set new benchmarks for cost and performance against Anthropic and OpenAI. For readers, that is the most important non-financial detail: the funding story is anchored to a claim about model economics. In AI, lower training and inference cost can make a model more attractive to developers and enterprises, even if headline capability is not the only factor.

In June, the company closed its first external funding round, about $7.4 billion at a valuation above $50 billion. That round established that Deepseek could raise large sums before the current effort. The new round then moved from a completed financing to a much larger target, with reported participation from CATL and Tencent. CATL declined to comment, while Tencent and Deepseek did not respond.

Latest progress: IPO preparation, Huawei chips, and an own inference chip

After the new round closes, Deepseek plans to restructure for an IPO in early 2027. The capital plan sits alongside a concrete infrastructure build-out: the company is building a data center with at least 160,000 AI chips from Huawei. It is also working on its own inference chip to reduce dependence on Nvidia and Huawei. Inference chips are specialized processors used to run trained models rather than train them.

Liang Wenfeng wants to keep developing open models with freely available weights. In this context, weights are the trained parameters that can be downloaded and used in products or research. A strategy of open model releases requires repeated model production and serving capacity, which helps explain why a funding round of this size is tied to a data center and custom chip work.

What to watch next

The next checks are whether the round closes at the reported minimum of $12 billion, whether it reaches the $15 billion range, whether the IPO restructuring happens on schedule for early 2027, and whether V4-Flash's reported benchmark results against Anthropic and OpenAI translate into durable cost-performance advantages. Deepseek now appears to be moving from a startup financing stage into a pre-IPO infrastructure stage, with large reported backers, a major chip build-out, and a stated goal of keeping model weights open.