DeepSeek IPO Push: $74B Valuation as Funding Round Tops $12B

A round that outgrew its target

DeepSeek is a lab known for building frontier AI models cheaply. It is now heading toward a public listing with one of the largest private funding rounds a Chinese technology company has raised. Reuters, citing a person with knowledge of the deal, reported on October 6 that the Hangzhou company is set to raise more than 80 billion yuan, about $11.93 billion. DeepSeek began the fundraising in July with a 50 billion yuan goal and a target valuation of 500 billion yuan, or about $74 billion.1519 Bloomberg broke the story earlier that day and said the final total could reach 100 billion yuan. Tencent Holdings and battery maker CATL are reported to have committed some of the largest amounts.1417 The round is expected to close this month.14

CNBC added another layer. It reported that DeepSeek is considering expanding the round to as much as 100 billion yuan, about $14.9 billion, which would double the original target. State-backed funds and corporate investors are competing for allocations.1 Most coverage treats this as a growth story. The more useful way to read it is as a pre-IPO cap table being shaped on purpose. DeepSeek is choosing which investors get in, and it is pricing the shares as if it already had a listing date.

Where the reporting agrees, and where it doesn't

Almost every outlet agrees on the basic facts. The round is oversubscribed, the valuation sits around 500 billion yuan, CATL and Tencent are major backers, and CITIC Securities has been hired to prepare a listing on Shanghai's STAR Market.618 The disagreements are in the details, and they matter for anyone thinking about IPO pricing.

Size. Estimates of the round range from roughly $7.4 billion to $15 billion. One analysis described that as a twofold spread on a single transaction, with every version traced back to the same Bloomberg reporting.11 Reuters' figure of "more than 80 billion yuan" is the most conservative firm number. Bloomberg's 100 billion yuan is the upper limit that signed term sheets could allow.35

Valuation. Most reports use $74 billion. CNBC and Channel Insider converted the same 500 billion yuan to about $75 billion.13 More important is whether that number is before or after the new money. Coverage from August described 500 billion yuan as a pre-money figure.816 If that holds, a raise of 80 to 100 billion yuan would put the post-money valuation well above $74 billion. American Bazaar wrote that the valuation "could now be higher" than the figure investors expected before the Bloomberg report.12 Our reading is that $74 billion is a floor, not the final price.

Status. Whalesbook said DeepSeek had "finalized" the round and called the company Beijing-based.4 Other outlets put DeepSeek in Hangzhou, and Reuters and CNBC both say the deal is still being negotiated.131 CNBC's sources said the terms are still fluid.1 Until the round closes, "committed" is the most that can be said.

A valuation that rose very quickly

The speed of the repricing is the main reason this round matters for the IPO. DeepSeek took outside money for the first time in a round that closed in June. It raised about 50 billion yuan, and a related investor filing implied a valuation of roughly 350 billion yuan.1 A securities filing disclosed in July pointed to a post-money valuation above $52 billion.16 Founder Liang Wenfeng put about $3 billion of his own money into that round.7 Forbes and Reuters also noted an unusual structure: most investors came in through a limited partnership controlled by Liang.16

In mid-July, the Financial Times reported that DeepSeek was already talking to new investors at a pre-money valuation of about $71 billion. At the same time, Bloomberg said the company was preparing IPO paperwork and aiming to complete its financial statements by the end of December.79 The round did not go smoothly. In late July, DeepSeek paused it after posts attributed to Liang about China lagging the US in AI spread widely online. Talks resumed in early August.16 CNBC also reported that the round, first meant to close by the end of August, ran late because some investors balked at the price.1

The pause and the price resistance are a counterweight to the headlines about heavy demand. Overall demand is clearly strong. But the valuation was negotiated, and some investors pushed back on it.

Who gets shares, and why that matters

CNBC reported one detail that will shape how the listing works. DeepSeek is turning away private funds raised from individual investors and is limiting the round mostly to government and corporate money.1 Investors with commitments reportedly include CATL, Geely Auto, Monolith Management and Loyal Valley Capital.12 Reporting from August also named Shixiang Capital, CPE, Legend Capital, semiconductor-focused Stony Creek Capital, funds backed by GigaDevice, and Hefei state investment vehicles as participants or investors in talks.8

This list is mostly strategic and policy-aligned money, not a typical venture syndicate. Tencent put 10 billion yuan into the earlier round. Coverage links its stake to plans to build AI into its platforms, while CATL may be interested in supplying components to data centers.26 For a STAR Market listing, a shareholder base like this has two effects. Investors are less likely to sell in a hurry after the listing, and the company sits closely in line with Beijing's goal of a self-sufficient technology sector. Excluding retail-funded vehicles now also leaves the first chance for ordinary investors to own DeepSeek shares to the public offering.

What the money is for

Computing power is the most consistent theme across the coverage. DeepSeek plans to install at least 160,000 Huawei AI accelerators at a data center being built in Inner Mongolia. It is also working with Huawei on software for the Ascend chip line.23 Whalesbook said the chips would be Ascend 950DT units.4 Earlier reports described plans for a gigawatt-scale data center campus and an in-house inference chip.97

This is where the round's significance goes beyond DeepSeek itself. One analysis argued that capital spent on domestic accelerators at this scale is demand that Nvidia never sees. In that view, US export controls are redirecting Chinese AI spending rather than reducing it.11 That argument is persuasive, but it depends on the Huawei build-out actually working. Whalesbook notes that moving workloads off Nvidia without losing performance will decide how attractive DeepSeek looks when it lists.4

The IPO path, and the gap in the numbers

DeepSeek's listing preparations look concrete. CITIC Securities is working on a STAR Market offering, though Reuters said the timing, valuation and offering size are still undecided.14 Bloomberg reported that the company could restructure after this round to prepare for a listing in early 2027.3 DeepSeek also plans to appoint its first chief financial officer, Yan Wentao, a partner at GL Ventures. That kind of hire usually comes ahead of a filing.13

The weak point is revenue. DeepSeek does not publish revenue figures, and it has said it is focused on research rather than near-term commercialization.13 One private-market tracker estimates its annualized revenue at about $500 million. Against a $74 billion price, that works out to roughly 100 times revenue, though the revenue number is an estimate and not a disclosure.16 Liang has told investors he cares more about progress toward artificial general intelligence than about maximizing profit.1 DeepSeek also releases its models openly and prices its API aggressively, which limits how much it can earn from each user.1113 Channel Insider identifies the central tension: a valuation near $75 billion creates expectations of returns that a lab built on cheap, open models may struggle to meet.3

Peers set the comparison

DeepSeek is not the only AI lab moving toward public markets. Its domestic rival Moonshot AI is reportedly targeting an early-2027 listing after raising money at a $50 billion valuation.1 In the US, Anthropic and OpenAI filed confidential IPO paperwork in June.7 By one comparison, DeepSeek's $74 billion is about 5% of the $1.4 trillion pre-money valuation OpenAI is reported to be seeking.11

That gap supports two different readings. One is that DeepSeek is cheap compared with its technical standing. The other is that investors are pricing in real uncertainty about how a lab that leads with open, low-cost models will make money. We lean toward the second reading for now. The round is oversubscribed because strategic and state-linked investors want exposure to China's leading AI lab before it lists. That is a different thing from investors endorsing its business model.

What to watch

The next milestones are clear. The first is the final size of the round and whether its close this month confirms the post-money valuation. The second is the restructuring Bloomberg expects afterward. The third is a STAR Market prospectus, which would be the first time DeepSeek has to disclose audited revenue.14 That filing will show whether $74 billion was a fair price or a price set by a small group of investors with strategic reasons to pay it.