2026 DEEPSEEK
ROUND2_
$70B_PRE.
Lead: 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 follows the source brief exactly: timeline, deal tables, compute math, voting rights, peer comps, controversy, STAR Market context, a five-step watchlist, and FAQ.
30-SECOND TL;DR
| Status | In talks (not closed) · signing targeted late August 2026 |
| Round 2 | ~50B yuan (~$7B) · pre-money ~500B yuan (~$70B) |
| vs Round 1 | Post-money >350B yuan → pre-money ~500B yuan · ~+43% |
| Combined | If Round 2 closes: >100B yuan (~$14B) in under 5 months |
| Caveat | All Round-2 figures from anonymous dealmakers / financial media · not officially confirmed |
1. Why This Round Matters
- Unusual cadence. Round 1 closed in June 2026 at ~50B yuan; Round 2 talks restarted mid-July, paused late July, and restarted again Aug 4–5 — a sub-five-month path toward a combined >$14B raise.
- Valuation vs revenue gap. ~$70B pre-money against reported ARR of $400–500M implies ~140–150x P/S — vs OpenAI ~65x and Anthropic ~21x (dealmaker estimates).
- Still rumor-stage. Leaked closed-door remarks, non-voting LP structure, and state-fund voting rights are all live debates. Treat every number as reported-but-unconfirmed.
2. Timeline: From "No Fundraising" to a $70 Billion Valuation in Four Months
| Date | Event |
|---|---|
| April 2026 | Corporate filing: registered capital up; Liang Wenfeng's direct stake 1%→34%; combined control ~84.29%. First external round opens; V4 preview ships. |
| June 2026 | Round 1 closes: ~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, National AI Industry Investment Fund also participated. |
| July 14–17, 2026 | STAR Market IPO prep reported; Round 2 talks at ~$71B pre-money (~480B yuan, +37%). ARR of $400–500M (mostly API tokens) first surfaces publicly. |
| July 25–26, 2026 | Talks abruptly pause. Bloomberg and others cite Liang's frustration that closed-door investor remarks circulated online; standby investors told to hold off. |
| August 4–5, 2026 | Caijing cites dealmakers: round restarted at ~50B yuan / ~500B yuan pre-money (~$70B), signing expected late August; both sides want a low-profile process. |
Caveat: every Round-2 figure — amount, valuation, timeline — comes from anonymous dealmakers cited by Chinese financial media, not from an official DeepSeek statement. Terms could still shift before signing.
3. The Numbers at a Glance
| Round 1 (closed) | Round 2 (in talks) | |
|---|---|---|
| Talks opened | April 2026 | Restarted mid-July, paused, restarted Aug 4–5 |
| Close | June 2026 | Late August 2026 (planned) |
| Amount | ~50B yuan (~$7.4B) | Target ~50B yuan (~$7B) |
| Valuation | Post-money >350B yuan | Pre-money ~500B yuan (~$70B) |
| Increase | — | ~+43% vs Round 1 |
| Key backers | National AI Industry Investment Fund, Tencent (10B yuan), CATL (5B yuan), JD.com, NetEase, IDG, Loyal Valley, Shixiang | Round-1 runner-ups + some existing backers increasing stakes |
| Combined if Round 2 closes | — | Over 100B yuan (~$14B) in under 5 months |
| Metric | Value | Note |
|---|---|---|
| ARR | ~$400–500M | Mostly API tokens; media-sourced, not official |
| Gross margin | Reportedly >50% | Not independently audited |
| Implied P/S | ~140–150x | Vs OpenAI ~65x / Anthropic ~21x |
| MAU | 100M+ (externally reported) | Methodology undisclosed |
4. Inside the Deal: Why the Math — and the Voting Rights — Don't Add Up Cleanly
4.1 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 chip-design engineers for in-house AI inference chips. Analysts estimate that for every 10 billion yuan raised, roughly 7 billion yuan goes into compute — chips, data centers, bandwidth, liquid cooling. The fundraising cadence is a race to keep pace with compute buildout.
4.2 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 and a 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's control near 84% — and is exactly the detail Forbes and others flagged when discussing governance and state influence for a company with a global user base.
4.3 A 148x P/S is either a bet on the future — or a red flag
At $70B pre-money against $400–500M ARR, implied P/S sits around 140–150x. One dealmaker's assessment, translated from Chinese coverage: "Pricing a foundation-model company is fundamentally an options bet, not a cash-flow valuation." Investors are pricing infrastructure-level optionality in China's compute ecosystem and enterprise agent market — not today's revenue.
5. How DeepSeek Stacks Up Against Moonshot, Zhipu, MiniMax
| Company | Listing | Latest valuation / mkt cap | Reported ARR | Recent funding pace |
|---|---|---|---|---|
| DeepSeek | Private, preparing STAR Market IPO | ~500B yuan pre-money (~$70B, in talks) | ~$400–500M | 2 rounds in 4 months, targeting >$14B combined |
| Moonshot AI (Kimi) | Private | ~$20B (May 2026); reportedly seeking $30B later | ~$200M | 4 rounds in 6 months, ~$3.9B total |
| Zhipu AI (Z.ai) | Listed (Hong Kong) | ~350B yuan mkt cap (May 2026) | Undisclosed | ~8.3B yuan pre-IPO |
| MiniMax | Listed (Hong Kong) | ~210B yuan mkt cap (May 2026) | Undisclosed | ~11B yuan pre-IPO |
Pattern: the two still-private frontrunners (DeepSeek, Moonshot) both carry ~140–150x P/S multiples — steeper than already-listed Zhipu and MiniMax in the secondary market.
6. The Controversy: A Leaked Transcript, an Unhappy Founder, and Bubble Warnings
- A leaked closed-door transcript stalled the deal. The July pause was reportedly triggered by Liang's frustration that first-round investor-meeting remarks spread online — a reminder that a growing investor base makes a low profile harder.
- 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. Unresolved governance questions remain.
- The valuation-to-revenue gap remains unresolved. A 140–150x P/S is extreme even for high-growth SaaS (typically 30–50x). Whether it holds depends on converting a technical lead into scaled enterprise revenue after a STAR Market listing — untested so far.
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 Round-2 terms.
7. Five-Step Watchlist for Readers
- Separate "in talks" from "closed". Target ~50B yuan; planned late-August signing — final terms TBD.
- Anchor to Round 1. Post-money >350B yuan vs Round-2 pre-money ~500B yuan (+43%); combined >100B yuan if it closes.
- Follow the compute share. Analysts say ~70 of every 100 yuan raised goes to chips, data centers, bandwidth, liquid cooling — plus headcount doubles and in-house inference chips.
- Read the governance terms. Most external LPs: no vote + five-year lock; National AI Industry Investment Fund is the exception.
- Place the multiple in peer context. Compare Kimi / Zhipu / MiniMax and OpenAI / Anthropic P/S; remember STAR Market's fifth listing standard now covers AI — no profit requirement to file.
8. FAQ
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. Final amount and terms could differ.
Why is DeepSeek raising money again so soon after its first round?
Reports point to a rapid buildout of data centers, in-house AI chips, and headcount across agent and infrastructure teams — CapEx outpacing what Round 1 covered.
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.
Does this valuation mean investors get more control?
Not necessarily. In Round 1, most outside investors received no voting rights and a five-year lock-up; only the National AI Industry Investment Fund got direct voting rights — a live governance debate.
When might DeepSeek go public, and can international investors buy in?
Reports say DeepSeek aims to file for a STAR Market listing in Shanghai by end-2026, targeting a 2027 debut. Retail access for international investors would likely be indirect; this private round is limited to institutional backers.
9. Why It Matters: STAR Market Rule Changes, China's Compute Push, and the Global AI Funding Race
China rewrote listing rules for unprofitable AI companies. On June 17, 2026, the Shanghai Stock Exchange announced at the Lujiazui Forum that it was expanding the STAR Market's "fifth listing standard" to cover AI — meaning a company need not be profitable, or even have significant revenue, to file if its technology is strong enough. That backdrop makes DeepSeek's planned late-2026 IPO filing (targeting 2027) plausible.
DeepSeek abandoned a five-year "no fundraising, no IPO, no commercialization" policy. For years it was funded entirely by Liang's quant fund, High-Flyer. That ended with Round 1 in June 2026 — a symbolic shift as Zhipu and MiniMax listed in Hong Kong and Moonshot kept raising.
This is part of a global re-pricing of frontier AI labs. OpenAI was reportedly valued at $300B in 2025; Anthropic's valuation reportedly surpassed OpenAI's by June 2026. Paying a steep premium for a globally competitive Chinese lab is, in part, a bet on technical parity plus market scale.
Compute self-reliance is the subtext. Reports of in-house AI inference chips and self-built data centers mirror a broader domestic-compute push — and explain why modest revenue still forces a fast fundraising cadence.
10. Close: Funding Headlines Are About Clusters; Developers Still Need Landable Compute
The core tension is clear: 140–150x P/S "options pricing" on one side, ~70% of capital into chips, data centers, bandwidth, and cooling on the other. Retail investors cannot buy this round — but they will feel the aftershocks in API price wars, open weights, and agent-framework competition.
Cloud GPU / Linux clusters fit large-scale training and throughput inference. If you need the same environment for Cursor + OpenClaw, MLX quantization checks against DeepSeek weights, or a stable Metal graphics stack for agent pipelines, bare Windows / Linux nodes often struggle on unified memory and sustained background stability. A pragmatic split: keep daily coding on your laptop; park side-by-side acceptance tests and 24/7 agent soak runs on a MACGPU remote Mac mini / Mac Studio node — on-demand, SSH-accessible, metered where the compute bill actually lands.