DeepSeek is about to raise $7.4 billion. Not one investor will get a vote.DeepSeek's Second Raise
Published: August 31, 2026 | Updated: August 31, 2026 | Category: AI | By Mahesh
DeepSeek, the Hangzhou-based AI lab whose low-cost models rattled Silicon Valley in early 2025, is close to closing a second external funding round that would raise roughly 50 billion yuan, or $7.4 billion, at a pre-money valuation of approximately 500 billion yuan, around $74 billion, according to reporting from The Wall Street Journal and the South China Morning Post, both citing people familiar with the ongoing talks.[1] The round, expected to close before the end of August, represents a roughly 43% jump from the roughly $51.9 billion valuation DeepSeek secured in its first external funding round, completed in June, according to Caijing's reporting on the deal's terms.[2]
The investor list backing this round is drawn substantially from the same syndicate that backed DeepSeek's first raise: local venture funds Monolith and Shixiang Capital, and Chinese battery giant Contemporary Amperex Technology Ltd, known as CATL, according to a source cited in reporting from IndexBox.[3] China's National AI Industry Investment Fund, Tencent, NetEase, JD.com, IDG Capital and Loyal Valley Capital were among the backers of the first round and are reportedly among those negotiating participation in this one as well.
Why This Round Almost Didn't Happen
The path to this second raise was not smooth, and the reason it stalled reveals something about how DeepSeek's founder operates that is worth understanding before looking at the numbers themselves. The round was actually paused in late July, according to Caijing's reporting, after founder Liang Wenfeng expressed displeasure over what was reportedly a leaked transcript of comments he made to investors during the first fundraising round. Multiple investors confirmed to Caijing that the second round had already begun by mid-July before being abruptly shelved at the end of that month, with some investors on the negotiation waiting list told that financing agreement signings were being temporarily halted.
The round resumed quietly weeks later, and China Money Network's reporting noted both DeepSeek and its investors are hoping this round proceeds in a low-key manner following the restart, a preference consistent with how deliberately private DeepSeek has kept its fundraising process compared to the public courtship rounds typical of major US AI labs. The list of parties negotiating for this second round includes both investors who were passed over during the first round's allocation and venture capital and industrial capital that had already invested previously, according to The Standard's reporting on the resumed talks.
The Governance Structure That Makes This Deal Genuinely Unusual
The most distinctive feature of DeepSeek's fundraising, and the detail most likely to surprise anyone used to how Silicon Valley mega-rounds typically work, is what outside investors are actually receiving in exchange for their capital. According to reporting on the first round's structure, outside investors received no voting rights and face a five-year lock-up period, with China's state-backed AI fund the only party granted direct equity and voting privileges. Capital from private investors was routed into a custom Limited Partnership vehicle managed directly by Liang Wenfeng, rather than converted into standard voting shares, a structure explicitly designed to preserve founder control over the company's direction even as its outside capital base scales into the tens of billions of dollars.
In a typical Silicon Valley mega-round, a $7.4 billion raise at this scale would come with board seats, information rights, and real governance influence for the lead investors writing the largest checks. DeepSeek's structure inverts that norm almost entirely. Investors are betting on a founder and a technology roadmap without gaining any formal mechanism to influence either one, a bet that becomes more remarkable given the scale of capital now involved. If this second round closes as reported, DeepSeek's cumulative fundraising across two rounds would exceed 100 billion yuan, roughly $14.8 billion, a figure Caijing's reporting noted would put DeepSeek far outpacing its closest domestic rival, Moonshot AI, in total capital raised.
What This Valuation Means Next to DeepSeek's American Rivals
Even at $74 billion, DeepSeek's valuation remains a fraction of what its most direct competitors in frontier AI command. Depth Grid covered Anthropic's own Q2 2026 revenue results earlier this month, when the company was reportedly valued near $965 billion in its most recent private funding, and OpenAI's own valuation has been reported in the $850 billion range in recent months. DeepSeek's $74 billion figure places it in a wholly different tier of capitalization than either US lab, even though its open-weight models have continued gaining real developer adoption globally, including, as Depth Grid noted in coverage of Nvidia's contested Hugging Face reporting earlier this week, forming a meaningful share of the open-source model traffic flowing through platforms like Hugging Face and OpenRouter.
That gap between DeepSeek's valuation and its US rivals' is not simply a reflection of smaller ambitions. DeepSeek has deliberately declined foreign institutional capital throughout its fundraising to date, according to reporting on the deal's investor composition, keeping its cap table concentrated among Chinese state, industrial and venture capital. That choice, combined with the founder-control LP structure, suggests DeepSeek's leadership is optimizing for a different set of priorities than maximum valuation or Western investor access, priorities that appear oriented instead toward maintaining operational independence and positioning for a domestic listing rather than competing dollar-for-dollar against Anthropic and OpenAI's US-centric capital bases.
The IPO Timeline This Round Is Building Toward
This funding round is explicitly framed by multiple reports as a step toward a public listing, not simply an isolated capital raise. DeepSeek could file for an IPO on Shanghai's STAR Market as early as the end of 2026, with a public debut targeted for 2027, according to people familiar with the company's plans cited by the Wall Street Journal. That timeline places DeepSeek's IPO ambitions on a similar horizon to Unitree Robotics, the Chinese humanoid robot maker whose own blockbuster STAR Market debut Depth Grid covered when Unitree shares surged as much as 629% on their trading debut earlier this month. Reading the two together suggests China's STAR Market is increasingly positioning itself as the primary public listing venue for the country's most capitalized frontier technology companies, an alternative track running in parallel to, rather than competing directly with, the US IPO pipeline currently being built out by Anthropic and OpenAI.
This latest round also marks a structural shift in DeepSeek's own capital identity. The company began as an AI research project financed primarily through Liang Wenfeng's hedge fund, with Liang personally contributing 20 billion yuan of the first round's total raise. The scale of this second round, if it closes as reported, completes DeepSeek's transition from a hedge-fund-backed research lab into a heavily capitalized, externally funded AI company preparing for public markets, even as its founder retains the kind of direct operational control that most companies of this financial scale have long since ceded to outside shareholders.
Read More on Depth Grid
- Anthropic made $787 million a year ago. Last quarter it made $11.5 billion.
- A robot that does backflips just made its founder richer than most tech IPOs ever have
- Nvidia says it hasn't bought Hugging Face. Two outlets disagree about which one is right.
- China's biggest memory chipmaker is doing something Chinese companies almost never do: suing the Pentagon
- Perplexity tripled its revenue in eight months. Nvidia wants in at $30 billion.
Article by Depth Grid News Desk | depthgrid.in