• DeepSeek has reportedly secured approximately $15 billion in a financing round, valuing the company at $74 billion pre-money, with CATL (300750.SZ) and Tencent (0700.HK) among the largest investors.
  • The round, if finalized, would bring total outside funding to roughly $22 billion and includes a diverse group of industrial and financial investors such as Geely (0175.HK), Midea (000333.SZ), and Legend Capital.
  • The company has also hired CITIC Securities (600030.SS) to prepare for a potential Shanghai STAR Market IPO, though timing and valuation remain undecided.

A Blockbuster Round for China's AI Champion

DeepSeek, the Hangzhou-based artificial intelligence developer, has raised approximately $15 billion in a new financing round, according to people familiar with the matter, marking one of the largest private fundraising efforts in China's technology sector. The round values the company at $74 billion pre-money, implying a post-money valuation of roughly $89 billion.

The investor list reads like a who's who of Chinese industry. Battery giant CATL and internet conglomerate Tencent are said to be among the largest backers, alongside automaker Geely, appliance maker Midea, and financial investors such as Boyu Capital, CPE, and Legend Capital. Existing shareholder IDG Capital also participated, according to sources. The round, which was initially targeted at 50 billion yuan, was upsized to 80 billion yuan and ultimately reached approximately 100 billion yuan ($15 billion), the sources said.

A Commercial Platform Takes Shape

The fundraising underscores DeepSeek's transformation from a founder-backed research lab into a heavily capitalized commercial platform. The company, led by founder and CEO Liang Wenfeng, emerged from the High-Flyer quantitative-investment ecosystem and gained international acclaim with its V3 and R1 models. Its newer offerings include the V4 family and V4.1-Flash.

Revenue has reportedly scaled rapidly. According to The Information, DeepSeek's annualized revenue run rate reached $1 billion in September, more than double its level a few months earlier. Reuters (TRI), which cited the report, said it could not independently verify the figure. Paid API access is central to the business, and the revenue growth partly reflected model-price increases of 2.3 to 4.5 times, rather than customer-volume growth alone.

The reported $74 billion pre-money valuation is roughly 74 times the $1 billion revenue run rate—a lofty multiple that highlights substantial expectations for future growth. However, the reviewed reporting does not establish audited annual revenue, net profit, cash burn, or margins.

IPO Preparation Underway

DeepSeek has hired CITIC Securities to prepare for a possible initial public offering on Shanghai's STAR Market, according to people familiar with the matter. The timing, valuation, and offering size remain undecided. A successful listing would provide the company with a new source of capital to fund its ambitious research agenda.

Liang reportedly told investors that more than 70% of computing capacity goes to training new models, with less than 30% used for inference—running models for users. That allocation helps explain why rapid revenue growth can coexist with a large financing requirement.

Industrial Investors, Strategic Implications

The investor mix matters as much as the amount. Battery manufacturing, automobiles, appliances, semiconductors, and internet platforms all appear in the reported consortium. This creates potential routes for deploying DeepSeek's technology throughout China's industrial economy, although investment does not itself confirm product partnerships or customer contracts.

DeepSeek continues to challenge the assumption that competitive AI must always command premium usage prices. In August, Reuters reported that Artificial Analysis found V4-Flash exceptionally inexpensive to run on its benchmark tests relative to prominent alternatives. Such comparisons depend on the models and tasks tested; they do not establish universal superiority.

At the same time, the $15 billion fundraising underscores that efficient models do not eliminate the need for substantial capital. DeepSeek's reported allocation of most computing capacity to research helps explain why rapid revenue growth can coexist with a large financing requirement.

The transaction fits the broader US–China competition over AI capabilities, computing infrastructure, and technological independence. Reuters describes strong interest from corporate and state-backed investors as China seeks to advance its position in AI. That does not mean every investor is state-controlled or that the financing itself is a government program.

A particularly relevant policy development came on October 16: China proposed adding AI technologies—including models, agents, and computing clusters—to its technology-export-control framework. Reuters reported that APIs, apps, and cloud services would be exempt. This remains a proposal, not a finalized blanket ban on exporting Chinese AI. For DeepSeek, the distinction could be consequential: international delivery through hosted services may face a different regulatory treatment from transferring controlled technology or deploying it overseas.

Separately, DeepSeek has faced scrutiny in several countries over security and privacy practices, creating an adoption barrier that additional funding alone cannot resolve.

What's Next

The next concrete milestones are confirmation of the financing's close, its final investor roster and valuation, and further details about the potential STAR Market listing. Reuters cautions that the current round's financial terms and participants could still change.

Commercially, the key test is whether DeepSeek can sustain revenue growth after raising prices while expanding service capacity without weakening its research program. Its reported allocation of more than 70% of computing resources to training makes that balancing act especially important.

The financing materially strengthens DeepSeek's resources, but it does not yet establish profitability, guarantee an IPO, or settle the contest with US AI leaders. The most informative evidence from here will be confirmed transaction terms, sustained revenue growth, operating economics, and independently assessed model performance.

Correction: An earlier version of this article misstated the pre-money valuation. It is $74 billion, not $75 billion. The error was due to a rounding mistake.