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

DeepSeek, the Chinese AI lab behind the open-weight R1 and V4 models, has restarted talks for a second funding round targeting roughly 50 billion yuan (~$7 billion) at a pre-money valuation of about 500 billion yuan (~$70 billion) — up 43% from its first round two months ago. If it closes, DeepSeek will have raised over $14 billion in under five months. This piece is for readers tracking China AI capital, STAR Market IPO paths, and compute buildouts. Every second-round figure below comes from anonymous dealmakers cited by financial media — not an official DeepSeek statement.

SECTION 01 How DeepSeek funding headlines trip readers up

• "In talks" is not "closed": Round 2 is still negotiating — 50B yuan target, ~500B yuan pre-money, late-August signing — all from anonymous sources, not filings;
• Pre-money vs post-money mix-ups: Round 1 is usually quoted as post-money >350B yuan; Round 2 rumors are pre-money ~500B yuan — the ~43% jump only makes sense if you keep the basis aligned;
• Missing the no-vote structure: Most Round 1 capital entered via an LP controlled by Liang Wenfeng — no voting rights, five-year lock-up; only China's National AI Industry Investment Fund invested directly with votes and no lock-up;
• Treating P/S as a cash-flow verdict: At ~$70B pre-money vs $400–500M ARR, implied P/S sits around 140–150x vs OpenAI ~65x and Anthropic ~21x — dealmakers call it an options bet, not cash-flow pricing;
• Ignoring the compute bill: Analysts estimate ~70% of each 10B yuan raised goes to chips, data centers, bandwidth, and liquid cooling — fundraising cadence tracks compute expansion;
• Detaching capital from product: Parallel threads include the V4 Flash GA build and custom inference silicon — verify each line separately.

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.

SECTION 02 Timeline: from "no fundraising" to a $70B valuation in four months

• April 2026: Corporate filing shows registered capital increased; Liang Wenfeng's direct stake rose from 1% to 34%. Combined with an entity he controls, total control reached ~84.29%. Same month: first external round opened and V4 previewed;
• June 2026: Round 1 closed at ~50B yuan (~$7.4B), post-money >350B yuan ($52–59B across sources) — largest first-round raise in Chinese AI history. Liang personally contributed 20B yuan; Tencent 10B; CATL 5B; JD.com, NetEase, IDG, and the National AI Industry Investment Fund also participated;
• July 14–17, 2026: Outlets reported STAR Market IPO prep plus Round 2 talks at ~$71B pre-money (~480B yuan), ~37% above Round 1 post-money. ARR of $400–500M (mostly API tokens) surfaced publicly;
• July 25–26, 2026: Round 2 talks paused. Bloomberg and others cited Liang's frustration that closed-door investor remarks had circulated online;
• August 4–5, 2026: Dealmakers cited by Caijing said the round restarted — still targeting 50B yuan at ~500B yuan pre-money (~$70B), signing expected late August, both sides wanting a low profile.

SECTION 03 Inside the deal: compute bills, voting rights, and a 148x P/S

The real bill is compute, not headlines. Shortly after Round 1, DeepSeek said it would double headcount across data-center and AI-agent teams; Reuters reported hiring for in-house AI inference chips. Industry analysts estimate that for every 10 billion yuan raised, roughly 7 billion goes into compute-related spend — chips, data centers, bandwidth, liquid cooling. The cadence is a race to keep pace with compute buildout.

Most investors don't get a vote. In Round 1, most outside capital flowed through a limited partnership controlled by Liang Wenfeng: no voting rights, five-year lock-up. The exception: China's National AI Industry Investment Fund invested directly with voting rights and no lock-up. That keeps Liang near 84% control — and draws governance scrutiny from outlets like Forbes about state influence versus founder control.

A 148x P/S is either a bet on the future — or a red flag. One dealmaker's line, translated from Chinese coverage: "Pricing a foundation-model company is fundamentally an options bet, not a cash-flow valuation." Investors are pricing the chance DeepSeek becomes infrastructure-level in China's compute ecosystem and enterprise agent market.

SECTION 04 Peer table + a six-step checklist for verifying the story

Controversy snapshot: (1) a leaked closed-door transcript stalled talks in late July; (2) voting-rights asymmetry keeps governance questions alive; (3) 140–150x P/S is extreme even versus high-growth SaaS at 30–50x — whether it holds depends on scaled enterprise revenue after a STAR Market listing, which markets have not yet tested.

1. Label the source tier: separate official filings from anonymous dealmakers + financial reprints — Round 2 numbers are still the latter;
2. Align valuation basis: when comparing rounds, state pre- vs post-money and yuan vs dollar ranges (Round 1 already spans $52–59B across sources);
3. Recompute P/S yourself: use the $400–500M ARR band to sanity-check the 140–150x implication — do not treat headlines as signed prices;
4. Read the cap table: confirm "most LPs: no vote + 5-year lock; state fund: direct votes, no lock" — that structure explains more controversy than the headline amount;
5. Track the STAR Market path: on June 17, 2026, the Shanghai Stock Exchange expanded the fifth listing standard to AI (no profit / large-revenue requirement if technology is strong enough); reports cite a late-2026 filing target and 2027 listing — still subject to regulators and company disclosure;
6. Map funding to compute and API risk: if production depends on DeepSeek APIs or agent pipelines, watch data-center expansion, custom chips, and product GA cadence — not only valuation headlines.

SECTION 05 Why it matters: STAR Market rules, global AI re-pricing, citeable numbers

• STAR Market rewrite for unprofitable AI: June 17, 2026 Lujiazui Forum announcement expanding the fifth listing standard is the regulatory backdrop for a late-2026 IPO filing / 2027 listing target;
• End of a five-year "no fundraising, no IPO, no commercialization" policy: High-Flyer funded the lab until Round 1 closed in June 2026; peers Zhipu and MiniMax already listed in Hong Kong; Moonshot keeps raising;
• Global coordinates: OpenAI was reportedly valued at $300B in 2025; Anthropic's valuation reportedly surpassed OpenAI by June 2026 — steep multiples are not China-only;
• Compute self-reliance subtext: custom inference chips and owned data centers explain why modest revenue still demands fast fundraising;
• Hard numbers memo: Round 2 target 50B yuan / ~500B yuan pre-money; if closed, >100B yuan combined in under 5 months; ARR ~$400–500M; implied P/S ~140–150x.

Primary third-party sources (re-check after publication; most figures remain anonymously sourced):

Forbes: DeepSeek Just Raised $7.4 Billion. Here's The Catch.

CIW: DeepSeek gave Beijing the only outside vote

DeepSeek official API docs (product-side cross-check)

A private-market premium does not mean your production stack already has enough stable compute. Virtualized cloud Macs often lose on Xcode builds, Metal acceleration, and long-running agents. For zero-loss native Apple silicon, stable iOS CI/CD, and 24/7 AI Agent automation, MACNOX dedicated physical Mac nodes are usually the better fit: 100% Apple hardware, full root, no hypervisor tax, day/week/month flexibility. For product-side context, see the DeepSeek custom-chip brief and V4 Flash official build review.

SECTION 06 FAQ