DeepSeek spent most of 2026 doing the one thing that is supposed to scare off investors: cutting prices until the margin on its own product looked uncomfortably thin. Its API rates undercut OpenAI’s discounted tier within days of that discount landing. Its models kept shipping for a fraction of what Western labs charge. By any normal startup logic, that is a company racing itself to the bottom.
Investors just responded by valuing it at roughly $74 billion.
According to multiple reports citing people familiar with the deal, DeepSeek is closing in on a funding round of about 50 billion yuan, close to $7.4 billion, at a pre-money valuation near 500 billion yuan, around $74 billion. The round is expected to close by the end of August, and it is widely read as the runway DeepSeek needs before an initial public offering on Shanghai’s STAR Market, a listing that could come as early as the second quarter of 2027.
The short version
- What: DeepSeek is closing a roughly $7.4 billion funding round at a pre-money valuation of about $74 billion
- When: expected to close by the end of August 2026
- Who’s paying: Tencent and JD.com through a limited partnership tied to founder Liang Wenfeng, plus China’s National AI Industry Investment Fund
- Why now: the round is widely seen as pre-IPO positioning ahead of a possible Shanghai STAR Market listing as early as Q2 2027
- The tension: the valuation jump is happening at the same time DeepSeek is undercutting its own pricing and everyone else’s
The round, in plain numbers
Strip away the yuan-to-dollar conversion and the shape of the deal is straightforward. DeepSeek’s last confirmed external round, closed in the late spring of 2026, valued the company at roughly $52 billion. This new round, if it closes as reported, pushes that up to about $74 billion, a jump of more than 40 percent in a few months, without a flagship model launch to point to as the trigger.
| Milestone | Approximate timing | Valuation / raise |
|---|---|---|
| R1 model release triggers global attention | January 2025 | No new funding event, but the moment that put DeepSeek on every investor’s watchlist |
| First major external funding round | Late spring 2026 | Roughly $7 billion raised at about $52 billion valuation |
| Pre-IPO round (reported) | Closing end of August 2026 | About $7.4 billion at roughly $74 billion pre-money |
| Possible IPO filing | As early as late 2026 | Target listing on Shanghai STAR Market, Q2 2027 |
Who is actually writing the check
The investor list is the part of this story that looks least like a normal venture round, and it tells you almost as much as the valuation itself.
Tencent and JD.com are putting in billions of yuan, but not directly. Their money reportedly flows through a limited partnership controlled by Liang Wenfeng, DeepSeek’s founder, and both companies are accepting five-year lockups with no voting rights attached to the stake. That is an unusual amount of control to hand back to a founder in a round this size. It only makes sense if the two companies are betting on continued access, distribution and integration rather than on a normal governance seat at the table.
The other investor, China’s National AI Industry Investment Fund, is on completely different terms. It gets direct equity, real voting rights, and the freedom to sell its stake later, making it the only backer in the round not locked into Liang’s structure. A state-linked fund walking away with the most flexible terms in the deal is its own kind of signal about who Beijing wants steering the country’s most visible AI company, and how much say private capital is being asked to give up to get in.
Why Shanghai, and why now
A Hong Kong or New York listing would once have been the default path for a Chinese tech company this size. Shanghai’s STAR Market exists specifically for cases like this one: a board created in 2019 to let loss-making, R&D-heavy technology companies list domestically, at a moment when routing a strategically sensitive AI company through a US exchange carries obvious friction given the current export-control climate around advanced chips.
Listing at home also keeps DeepSeek inside a regulatory and capital environment Beijing controls directly, at a time when the government has made it clear it wants a homegrown AI champion that does not depend on Nvidia’s newest hardware or Washington’s good graces to keep training. The National AI Industry Investment Fund’s presence in this round, on the most favorable terms of any investor, fits that picture closely.
The part that should feel contradictory
Here is the tension nobody in the reporting quite resolves: DeepSeek’s valuation is going up while its prices are going down, on purpose. Earlier this year, OpenAI cut its own API prices by 80 percent, and DeepSeek promptly undercut that too, in a race that has compressed margins across the entire industry. A company does not usually get more valuable by making its core product cheaper for everyone, including itself.
Unless the bet is not really about the API business at all. Investors writing checks at $74 billion are not pricing DeepSeek on token margins. They are pricing distribution, the ability to train frontier-competitive models on a fraction of the compute budget Western labs report, and a live seat at the table as ChatGPT’s own usage numbers keep climbing while its share of the overall AI market keeps slipping to exactly this kind of cheaper, fast-moving competitor. Cheap models plus fast iteration is a market-share strategy, not a profit strategy, and pre-IPO investors are often the ones being asked to fund the gap between the two.
What to actually watch next
| The August close | Whether the round actually finalizes at the reported terms, or slips, as pre-IPO rounds in China frequently do |
| A formal IPO filing | Expected as early as late 2026, the first hard signal of a real STAR Market timeline |
| Model releases between now and listing | Any flagship launch will be read by the market as evidence for or against the valuation, fairly or not |
| Pricing moves from OpenAI and Google | The price war shows no sign of pausing, and DeepSeek’s IPO story depends on staying ahead of it, not settling into it |
None of this makes DeepSeek unique. Startups get priced on a growth story instead of a profit and loss statement all the time, and other AI companies have posted even steeper valuation jumps this year on far less revenue. What makes this round worth watching is the specific combination on display: a company that is simultaneously racing to the bottom on price and to the top on valuation, backed by investors who gave up nearly all their leverage to get in, with a state fund holding the only exit ramp. Whether that combination survives contact with a public listing is exactly the question a Shanghai IPO in 2027 would answer.
