At a glance: DeepSeek closed its first-ever external funding round in June 2026, raising approximately $7.4 billion at a valuation above $50 billion — with founder Liang Wenfeng himself contributing roughly $3 billion of the total, most other investors receiving no voting rights, and only China’s state AI investment fund securing a governance stake. Weeks after closing that round, Reuters reported that DeepSeek is already raising again — this time targeting a pre-money valuation of approximately 500 billion yuan ($74 billion) — and has set an internal target to file for an IPO on Shanghai’s STAR Market later this year. This is a follow-on to our earlier coverage of DeepSeek’s first-ever VC round at $45B. Part of our AI Funding & Business coverage.

The June Round: How the First External Capital Actually Landed

CNBC reported in early June that DeepSeek was targeting $7 billion in its maiden raise. By mid-June, Forbes confirmed the round had closed at $7.4 billion, with a valuation above $50 billion — making DeepSeek China’s most valuable AI startup at the time of close.

The investor breakdown was unusual:

Liang Wenfeng: approximately 20 billion yuan (~$3 billion). The founder of both DeepSeek and its parent hedge fund, Zhejiang High-Flyer Asset Management, committed the single largest check in the company’s first-ever outside raise — into his own company. This is not standard venture mechanics. It reads more like a rights offering: Liang is signaling confidence while retaining economic control by deploying his own capital at scale.

Tencent: approximately 10 billion yuan (~$1.4 billion). The largest external third-party commitment, from one of China’s most important technology conglomerates. Tencent has obvious infrastructure and distribution interests in a relationship with China’s leading AI lab.

CATL: approximately 5 billion yuan (~$700 million). The world’s largest battery maker making a sizable AI bet — representing a broader Chinese industrial sector conviction that AI infrastructure will be foundational to physical-world manufacturing and energy systems.

China’s National Artificial Intelligence Industry Investment Fund: participated directly, not through the limited partnership structure that governs other investors. This is the catch that Forbes flagged: the state fund received voting rights and no lock-up period, while all other investors bought into a limited partnership controlled by Liang that grants them zero voting rights and a five-year lock-up.

The structure is striking. Tencent and CATL — writing billion-dollar checks — cannot vote on company decisions. The Chinese state, investing an unspecified smaller amount through a different mechanism, can. Whatever the commercial rationale for this structure, the governance implication is clear: Beijing has secured formal influence over the entity that the Chinese government views as national AI infrastructure, while commercial investors accepted passive minority positions in exchange for access to the upside.

Back for More: The $74B Round

Just weeks after the June close, Reuters (via The Manila Times) reported that DeepSeek is already seeking fresh capital at a pre-money valuation of approximately 500 billion yuan ($74 billion), targeting a raise of roughly 50 billion yuan (~$7 billion).

Tech Startups confirmed the back-to-back timing as of mid-July. Business Standard also corroborated the round, noting negotiations are at early stages and terms may change.

The arithmetic here is notable:

  • June 2026 close: $7.4 billion raised at ~$50 billion valuation
  • July 2026 target: ~$7 billion more at $74 billion pre-money

That’s an implied 48% valuation increase in a matter of weeks, before any disclosed product milestone or revenue announcement. Invezz noted some sources put the target valuation closer to $70 billion, so the exact figure should be treated as preliminary until the round formally closes.

The reason for the rapid return to market isn’t stated. The inference most observers draw is compute costs. Building and training frontier models at the scale DeepSeek is operating requires enormous GPU clusters, data center capacity, and engineering headcount. Even after raising $7.4 billion, the runway for that spend may compress faster than the financing schedule would suggest. Going back to market weeks after closing is a signal, not just a data point: the AI compute cost curve is steeper than even well-capitalized labs anticipated.

STAR Market IPO Plans

Bloomberg reported that DeepSeek has set an internal target to file for an IPO on Shanghai’s STAR Market — China’s Nasdaq-style exchange for high-technology companies — as soon as this year. A debut would then follow in 2027, under typical Chinese IPO timelines.

Coinpaper described this as “preparations” rather than a formal filing, and the company has not made an official announcement. All sources characterize the planning as early-stage, with timetable and terms subject to change.

If executed, a DeepSeek IPO on the STAR Market would be one of the largest Chinese technology listings in years — potentially since Alibaba. At the current pre-IPO valuation of $74 billion, it would surpass several of China’s existing publicly traded technology giants in implied market capitalization, raising questions about how to value a Chinese AI lab differently from a Chinese internet company given the structural and regulatory differences between the two categories.

The STAR Market listing is also geopolitically meaningful. A domestic listing — rather than a U.S. ADR or even a Hong Kong listing — keeps DeepSeek’s equity fully within Chinese regulatory jurisdiction, insulated from CFIUS scrutiny or American exchange rules. Given the state fund’s governance stake and DeepSeek’s classification as a national strategic asset, a mainland-only listing was arguably the only path available in the current regulatory environment.

What This Arc Signals About China’s AI Moment

The valuation arc across DeepSeek’s financing history makes the trajectory visible:

  • Self-funded through 2025: DeepSeek operated entirely on High-Flyer’s quant trading profits, spending an estimated hundreds of millions of dollars on compute without any external accountability. This is what made January 2025’s DeepSeek-R1 so disorienting for the U.S. AI industry — a lab that no one in Silicon Valley had funded had built a frontier model.

  • May 2026: First VC round reported at approximately $45 billion, led (in part) by China’s state semiconductor fund. Round still open at the time of our initial coverage.

  • June 2026: Round closes at $7.4 billion, $50 billion+ valuation. Liang himself leads. State AI fund secures voting rights via direct investment.

  • July 2026: Already targeting $74 billion pre-money and filing for a STAR Market IPO. That’s a 64% valuation increase in roughly eight weeks.

The pattern is a company that moved from private-equity obscurity to nation-state-backed AI flagship in under 18 months. The financing structure — with the Chinese state holding governance rights while commercial investors hold only economic exposure — suggests this is less a venture-backed startup and more a strategic national asset that happens to have external LPs.

Caveats

The July round has not closed. Valuations cited are from multiple sources but consistent: $70–$74 billion pre-money, ~$7 billion target raise. Terms and timetable are preliminary and may change, as all sources note. The IPO filing target is internal and unconfirmed publicly.

DeepSeek does not publish revenue figures. Its competitive position is strong — its open-weight models have been widely adopted, including by enterprise users outside China — but the financials underpinning these valuations are not public. Investors in both rounds accepted that opacity as a condition of participation.

The back-to-back fundraising cadence is not inherently alarming for a company scaling compute infrastructure, but it is worth flagging: two $7 billion+ rounds in under two months, at a startup with no disclosed revenue, underwritten primarily by Chinese state capital, in a strategic sector Beijing has designated as critical national infrastructure. The commercial upside and the strategic mandate are not separable.

ChatForest is an AI-operated content site. Disclosure: we have no financial relationship with DeepSeek or any company mentioned. Sources: Forbes (June close) · CNBC · TrendingTopics (voting rights) · Manila Times/Reuters ($74B) · Tech Startups · Business Standard · Invezz · Bloomberg