Why Is DeepSeek Raising Another $7 Billion Just Months After Its First Round?

• April 2026: A corporate filing shows DeepSeek increased its registered capital and founder Liang Wenfeng personally subscribed to the new shares, raising his direct stake from 1% to 34%. Combined with an entity he controls, his total control over the operating company reached roughly 84.29% — the ownership structure that would later shape how outside capital entered. That same month, DeepSeek opened its first-ever external funding round and previewed its V4 model series.
• June 2026: The first round closed at roughly 50 billion yuan (~$7.4 billion), giving DeepSeek a post-money valuation of more than 350 billion yuan (reported in a range of $52–59 billion across sources) — the largest first-round raise in Chinese AI history. See our earlier roundup on the 2026 AI funding supercycle.
• July 14–17, 2026: Multiple outlets reported DeepSeek had begun preparing a STAR Market (Shanghai) IPO and was simultaneously in talks for a second round at a pre-money valuation of roughly $71 billion (~480 billion yuan) — about 37% above the first round's post-money value. This is also when DeepSeek's annualized revenue (ARR) — reportedly $400–500 million, mostly from API token usage — became public for the first time.
• July 25–26, 2026: The second-round talks abruptly paused. According to Bloomberg and other outlets, part of the reason was that founder Liang Wenfeng was unhappy that comments he made during closed-door investor meetings had circulated online; DeepSeek reportedly told some investors on the standby list to hold off on signing.
• August 4–5, 2026: Dealmakers cited by Chinese business magazine Caijing said the round had restarted, still targeting 50 billion yuan at a pre-money valuation of roughly 500 billion yuan (~$70 billion), with signing expected in late August. Both DeepSeek and the investors reportedly want to keep this round low-profile.

Pain points for readers trying to act on the news:

• Unconfirmed terms: Amount, valuation, and timeline come from anonymous dealmakers and financial media — not an official DeepSeek statement.
• Breakneck cadence: A second round two months after first-round close fragments the decision frame.
• Opaque governance: Most outside investors reportedly get no vote and a five-year lock-up, while a state fund holds a different structure.
• Valuation-to-revenue gap: A ~140–150x P/S multiple is extreme versus mature-sector norms and untested in the public market.

Caveat: every figure above about the second round — the amount, the valuation, the timeline — comes from anonymous dealmakers cited by Chinese financial media, not from an official DeepSeek statement. Terms could still shift before signing.

The real bill is compute, not headlines. Shortly after closing its first round, DeepSeek said it would double headcount across data-center and AI-agent teams, and Reuters reported it was hiring chip-design engineers to develop its own AI inference chips (see DeepSeek custom chip coverage). Industry analysts estimate that for every 10 billion yuan DeepSeek raises, roughly 7 billion yuan goes straight into compute-related spending — chips, data centers, bandwidth, liquid cooling. The fundraising cadence is a race to keep pace with its own compute buildout. On the model side, V4 Flash official also lifted Agent and inference demand.

Most investors don't get a vote. In the first round, most outside capital flowed in through a limited partnership controlled by Liang Wenfeng, meaning those investors received no voting rights and are locked in for five years. The one exception: China's National AI Industry Investment Fund, which invested directly and got both voting rights and no lock-up. This structure keeps Liang's control near 84% — and it is exactly the kind of detail that has drawn scrutiny from outlets like Forbes about governance and state influence versus founder control.

A 148x price-to-sales ratio is either a bet on the future — or a red flag. At a $70 billion pre-money valuation against $400–500 million in ARR, DeepSeek's implied P/S ratio sits around 140–150x, dwarfing OpenAI's ~65x and Anthropic's ~21x. One dealmaker's assessment, translated from Chinese coverage, is blunt: "Pricing a foundation-model company is fundamentally an options bet, not a cash-flow valuation." Investors aren't pricing today's revenue — they're pricing the chance that DeepSeek becomes infrastructure-level in China's compute ecosystem and enterprise agent market.

Fundraising races compute buildout; the cap table protects founder control; the multiple prices an infrastructure option. Together they explain the $70B headline.

The pattern: DeepSeek and Moonshot — the two still-private frontrunners — both carry P/S multiples around 140–150x, well above what the already-listed Zhipu and MiniMax trade at in the secondary market. Private-market investors are, for now, paying a steeper premium for the two labs that have not yet faced public-market scrutiny.

Controversy:

• A leaked closed-door transcript stalled the deal. The immediate trigger for the July pause was reportedly Liang Wenfeng's frustration that remarks he made during first-round investor meetings had spread online — a reminder that as DeepSeek's investor base grows, keeping a low profile is getting harder.
• The voting-rights structure is drawing outside scrutiny. Most external investors have no vote and a five-year lock-up, while only the state-backed National AI Industry Investment Fund gets direct voting rights with no lock-up. This remains media-reported rather than officially confirmed, and DeepSeek has not publicly addressed the governance questions.
• The valuation-to-revenue gap remains unresolved. A 140–150x P/S ratio is extreme by any industry's standard — even top-tier, high-growth SaaS companies typically trade at 30–50x. Whether this valuation holds up depends on whether DeepSeek can convert its technical lead into scaled enterprise revenue after a STAR Market listing — something the market has not yet tested.

To be clear: every detail above about deal size, valuation, and ownership structure comes from anonymous-sourced reporting (Caijing, Reuters, Bloomberg, Forbes, among others). DeepSeek has not officially confirmed the terms of its second round, so treat the specific numbers as reported-but-unconfirmed until a formal announcement.

Why it matters:

• STAR Market rule changes: On June 17, 2026, the Shanghai Stock Exchange announced at the Lujiazui Forum that it was expanding its "fifth listing standard" on the STAR Market to cover AI companies — meaning a company does not need to be profitable, or even have significant revenue, to file, as long as its technology is strong enough. That rule change is the regulatory backdrop making DeepSeek's planned late-2026 IPO filing (targeting a 2027 listing) plausible.
• End of a five-year "no fundraising, no IPO, no commercialization" policy: For years, DeepSeek was funded entirely by founder Liang Wenfeng's quant fund, High-Flyer, and refused outside capital. That policy ended with the first round in June 2026 — a symbolic moment as rivals Zhipu and MiniMax listed in Hong Kong and Moonshot kept raising.
• Global re-pricing of frontier AI labs: OpenAI was reportedly valued at $300 billion in 2025, and Anthropic's valuation reportedly surpassed OpenAI's by June 2026. Against that backdrop, investors paying a steep premium for a globally competitive Chinese lab is partly a bet that Chinese model companies can keep pace technically.
• Compute self-reliance is the subtext: Reports that DeepSeek is developing its own AI inference chips and building out its own data centers mirror a broader trend among leading Chinese AI labs — and help explain why a company with comparatively modest revenue still needs to raise capital this fast.

Six-step watchlist:

1. Label source status: Mark "500B yuan pre-money / 50B yuan target / late-August signing" as anonymously sourced until official confirmation or close.
2. Reconcile both rounds: Use Section 02 tables to keep raise size, pre- vs post-money, and cumulative totals distinct from "closed" facts.
3. Recompute the P/S band: Check ~140–150x against $400–500M ARR and a ~$70B pre-money mark; compare with dealmaker estimates for OpenAI and Anthropic.
4. Parse voting rights: Track whether Round 2 keeps the no-vote LP structure, five-year lock-up, and National AI Fund exception.
5. Watch the STAR Market clock: File-by-end-2026 / list-in-2027 remains a media-reported plan, not a confirmed prospectus date.
6. Stress-test local compute hosts: If your team accelerates Agent or eval pipelines on the back of this narrative, lock a stable, rootable Apple Silicon host before chaining APIs and local weights.

# DeepSeek R2 watchlist (unofficial)
status: in_talks_not_closed
target_raise_cny: 50_000_000_000
pre_money_cny: ~500_000_000_000
vs_round1_post_money: +43%
implied_ps: ~140-150x (ARR $0.4-0.5B)
official_confirm: NO
next_checkpoint: late_August_2026_signing

Citable figures (media reports, 2026-08-06):

• Round 2 target: ~50B yuan raise; ~500B yuan (~$70B) pre-money
• Vs Round 1: ~+43% vs post-money >350B yuan; >$14B combined if Round 2 closes
• ARR / P/S: ~$400–500M ARR; implied ~140–150x P/S
• Control: Liang-related control ~84.29%; most outside investors no vote, five-year lock
• IPO path: STAR Market prep; file by end-2026, target 2027 listing
• Compute allocation: Analysts cite ~70% of each 10B yuan raise going to compute-related spend

Has DeepSeek's second funding round actually closed?
Not yet. As of this writing, the round is still in negotiation, targeting a close by late August 2026. The final amount and terms could differ from what's currently being reported.

Why is DeepSeek raising money again so soon after its first round?
The company is funding a rapid buildout of data centers, in-house AI chips, and headcount across its agent and infrastructure teams — capital expenditure that is outpacing what its first raise covered, according to multiple reports.

Is the $70 billion valuation confirmed?
No. It comes from dealmakers cited anonymously by Chinese financial media (primarily Caijing), not from an official DeepSeek statement, and it could change before any agreement is signed.

Does this valuation mean DeepSeek's investors get more control over the company?
Not necessarily — and that is part of the controversy. In the first round, most outside investors received no voting rights and a five-year lock-up, while only China's National AI Industry Investment Fund got direct voting rights, which has raised governance and state-influence questions that remain unresolved.

When might DeepSeek go public, and can international investors buy in?
DeepSeek is reportedly preparing to file for a STAR Market listing in Shanghai by the end of 2026, targeting a 2027 debut. The STAR Market is a mainland China exchange, so retail access for international investors would likely be indirect (e.g., through connect programs) rather than direct participation in this private round, which is currently limited to institutional backers.

Sources (selected): Caijing magazine reporting, as relayed by Sina Finance and Wall Street CN; The Standard (HK), Gate News, ChainCatcher; Forbes: "DeepSeek Just Raised $7.4 Billion. Here's The Catch."; South China Morning Post, Caixin Global, Reuters, Bloomberg; CIW on the cap table structure; DeepSeek API documentation, TechCrunch, and Hugging Face write-ups on DeepSeek-V4; 36Kr and TMTPost coverage comparing Moonshot, Zhipu, and MiniMax. Most figures come from anonymous sources and media reports rather than official company disclosures. Verify the latest confirmed numbers before publishing.

As funding headlines accelerate roadmaps, common alternatives still have real flaws: shared VPS hosts for long-horizon Agents hit bandwidth jitter and oversubscription; ad-hoc inference nodes drop long-lived connections; self-built racks spike delivery time and ops cost. For more stable production environments suited to AI Agents and local eval pipelines, JEXCLOUD multi-region bare-metal Mac nodes are the better fit: dedicated Apple Silicon, 7x24 uptime, monthly elasticity, roughly 120-second provisioning. See nodes and pricing on the JEXCLOUD pricing page.