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

01

Six Pitfalls When Reading DeepSeek’s Round-2 Headlines

The headlines are loud. The deal is still in talks. Before treating any figure as settled, check these six traps:

1. 01

   Treating “in talks” as “closed”: As of this writing, Round 2 is still negotiating. The ~50B yuan target, ~500B yuan pre-money, and late-August signing plan all come from Caijing and other outlets citing dealmakers — DeepSeek has not officially confirmed them.

2. 02

   Mixing pre-money and post-money: Round 2 is quoted as ~500B yuan pre-money. Round 1 closed at post-money above 350B yuan. The ~43% jump compares Round-2 pre vs Round-1 post — not the same basis doubled.

3. 03

   Skipping the late-July pause: Talks opened by mid-July, then stalled July 25–26. Bloomberg and others said Liang Wenfeng was unhappy that closed-door investor remarks had circulated online.

4. 04

   Assuming all outside investors get votes: In Round 1, most capital entered via a Liang-controlled LP — no voting rights + five-year lock-up. The exception: China’s National AI Industry Investment Fund, which invested directly.

5. 05

   Forcing mature SaaS multiples onto foundation models: ~$400–500M ARR implies ~140–150x P/S — far above OpenAI’s ~65x and Anthropic’s ~21x (dealmaker estimates). Pricing here is closer to an options bet than a cash-flow DCF.

6. 06

   Thinking retail can “buy the dip” now: Both rounds are institutional. Most Round-1 shares carry lock-ups. Ordinary investors have no direct path until a STAR Market listing — if and when it happens.

03

Timeline and Deal Structure: What a $70B Pre-Money Really Prices

From “no fundraising” to a ~$70B valuation in four months

• April 2026: A corporate filing shows DeepSeek raised registered capital; Liang Wenfeng personally subscribed, lifting his direct stake from 1% to 34%. Combined with an entity he controls, total control reached roughly 84.29%. That same month, DeepSeek opened its first external round and previewed the V4 series.
• June 2026: Round 1 closed at ~50B yuan (~$7.4B), post-money above 350B yuan (reported $52–59B across sources) — the largest first-round raise in Chinese AI history. Liang personally contributed 20B yuan; Tencent 10B yuan; CATL 5B yuan; JD.com, NetEase, IDG Capital, and the National AI Industry Investment Fund also participated.
• July 14–17, 2026: Outlets reported STAR Market IPO prep and second-round talks at ~480B yuan pre-money (~$71B) — about 37% above Round-1 post-money. ARR of ~$400–500M, mostly from API tokens, became public for the first time.
• July 25–26, 2026: Talks paused abruptly. Bloomberg and others said Liang was unhappy that closed-door investor comments had spread online; some standby investors were told to hold off on signing.
• August 4–5, 2026: Caijing, citing dealmakers, said the round restarted at ~50B yuan target and ~500B yuan (~$70B) pre-money, with late-August signing planned. Both sides reportedly want a low-profile process.

1. Why raise again so soon: the real bill is compute

Shortly after Round 1 closed, DeepSeek said it would double headcount across data-center and AI-agent teams.

Reuters also reported it was hiring chip-design engineers for in-house AI inference chips.

Analysts estimate that for every 10B yuan raised, roughly 7B yuan goes into compute-related spend — chips, data centers, bandwidth, liquid cooling.

The fundraising cadence is a race against compute buildout, not a valuation vanity cycle. For chip and domestic-compute context, see our earlier piece on DeepSeek’s custom AI inference chips.

2. Unusual ownership: most investors don’t get a vote

In Round 1, most outside capital flowed through a limited partnership controlled by Liang Wenfeng.

Those investors received no voting rights and a five-year lock-up.

The exception: China’s National AI Industry Investment Fund invested directly, with voting rights and no lock-up.

The structure keeps Liang’s control near 84% — and it is the detail outlets like Forbes have flagged when discussing governance and state alignment alongside a global user base.

3. A ~148x P/S: pricing cash flow or pricing an option?

At ~$70B pre-money against $400–500M ARR, implied P/S sits around 140–150x — well above OpenAI’s ~65x and Anthropic’s ~21x.

One dealmaker’s line, as translated from Chinese coverage, is blunt: pricing a foundation-model company is fundamentally an options bet, not a cash-flow valuation.

Investors are not pricing today’s revenue. They are pricing the chance DeepSeek becomes infrastructure-level in China’s compute ecosystem and enterprise agent market.

05

Controversy, Industry Context, and a Six-Step Checklist

Three flashpoints

• A leaked closed-door transcript stalled the deal: The July pause was reportedly triggered by Liang’s frustration that first-round investor-meeting remarks had spread online — a reminder that a larger investor base makes a low profile harder to keep.
• Voting-rights structure draws outside scrutiny: Most external investors have no vote and a five-year lock-up; only the state-backed National AI Industry Investment Fund gets direct voting rights with no lock-up. Forbes, CIW, and others have raised governance questions; DeepSeek has not publicly settled them.
• The valuation-to-revenue gap remains unresolved: A 140–150x P/S is extreme even versus high-growth SaaS (typically 30–50x). Whether it holds depends on converting technical lead into scaled enterprise revenue after a STAR listing — still untested by public markets, and a live case in the broader “AI valuation bubble” debate.

To be clear: deal size, valuation, and ownership details above come from anonymous-sourced reporting (Caijing, Reuters, Bloomberg, Forbes, among others). Treat the specific numbers as reported-but-unconfirmed until a formal announcement.

Why it matters: STAR rules, China’s compute push, global AI capital

• STAR Market rule change: On June 17, 2026, the Shanghai Stock Exchange said at the Lujiazui Forum it would expand the “fifth listing standard” to AI companies — no profit requirement, and no need for large revenue if technology is strong enough. That is the regulatory backdrop for DeepSeek’s reported late-2026 IPO filing (targeting a 2027 listing).
• End of the five-year “three nos”: For years DeepSeek was funded entirely by Liang’s quant fund, High-Flyer, and refused outside capital. Round 1 in June 2026 ended that policy. Zhipu and MiniMax have already listed in Hong Kong; Moonshot keeps raising at pace — the sector is mid-repricing.
• Global AI funding race: OpenAI was reportedly valued at $300B in 2025; Anthropic’s valuation reportedly surpassed OpenAI’s by June 2026. Steep premiums for a globally competitive Chinese lab are part of that same re-pricing.
• Compute self-reliance is the subtext: Reports of in-house inference chips and owned data centers mirror a broader domestic-compute strategy — and help explain why modest ARR still pairs with rapid fundraising.

Six-step checklist: how to judge the next funding headline

1. 01

   Separate status: Is it talks, signed, or closed? Round 2 is still in talks.

2. 02

   Align valuation basis: Pre- vs post-money, yuan vs dollars, and whether an option pool is included.

3. 03

   Check ARR provenance: $400–500M is media-cited, not audited disclosure — do not mix it with filed financials.

4. 04

   Read governance terms: Voting rights, lock-ups, and the state-fund direct-invest exception often matter more than the multiple.

5. 05

   Map the listing path: STAR fifth standard ≠ listing is done; watch the late-2026 filing window.

6. 06

   Return to product and compute: Capital ultimately funds chips, data centers, and agent capacity — track API stability and model cadence on the developer side too.

Funding headlines do not ship models. Teams still need stable machines for training loops, inference, and Agent pipelines.

Public-cloud GPU queues, laptop disconnects, and contended CI runners all slow real delivery.

For AI engineering and automation teams that need stable SSH, dedicated hardware, and iOS / macOS build environments, NodeMini Mac Mini cloud rental is usually the better fit: second-scale provisioning on real Apple Silicon, so Agent work, local inference, and builds share one reproducible Mac.

See Mac Mini rental rates and the help center.

Sources: 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 docs, TechCrunch, and Hugging Face write-ups on DeepSeek-V4; 36Kr and TMTPost coverage comparing Moonshot AI, Zhipu AI, and MiniMax. Most figures come from anonymous sources and media reports rather than official disclosures. Data current as of August 6, 2026.