China's AI June Surge: DeepSeek's $7B First Round, Zhipu's STAR Market Bid, and the Agentic AI Stack

DeepSeek closes its first external round at $59B. Zhipu files for a $2B STAR Market IPO. Huawei unveils an Agentic AI stack. And on June 6, China's AI industry converges on Hangzhou for a summit that signals the shift from "model hype" to "industrial deployment."

On June 3, 2026, DeepSeek — the Hangzhou-based AI lab that shocked Silicon Valley in January 2025 with its $5.6M R1 training run — officially began its first external funding round. The target: $7 billion (approximately ¥50 billion). The implied valuation: up to $59 billion.

Less than 48 hours later, on June 1, Zhipu AI (智谱) — already the world's first publicly traded pure-play LLM company after its January 2026 Hong Kong IPO — announced it would seek a second listing on Shanghai's STAR Market (科创板), raising ¥15 billion (approximately $2.1 billion). Its Hong Kong stock had already surged 1,100% from its IPO price.

And on June 4–5, while capital markets digested these numbers, Huawei Cloud held its INSPIRE 2026 conference in Shanghai, releasing a full Agentic AI stack — from enterprise agent development platforms (AgentArts) to embodied intelligence toolchains (CloudRobo) — backed by partnerships with Chery Auto, Changan Auto, and AI chip startup Infinigence-AI.

By June 6, the pieces converged: the inaugural China AI High-Quality Development Summit opened in Hangzhou, with government officials, investors, and founders sharing the same stage. The message was unmistakable: China's AI industry is no longer in the "model release" phase. It is in the "capital + infrastructure + deployment" phase.

This is the story of how three separate vectors — venture capital, public markets, and industrial policy — collided in a single week to reshape the global AI order.

Executive Summary: The Week in Numbers

Sources: Caixin, Reuters, Sina Finance, TMT Post, Huawei Cloud official announcements, CSRC filings, CSET (Center for Security and Emerging Technology)

1. DeepSeek: The $59B Question

1.1 From "Zero External Funding" to the Biggest First Round in Chinese History

For nearly three years, DeepSeek operated on a principle that seemed almost quaint in the AI gold rush: no outside capital. Founded in July 2023 by Liang Wenfeng (梁文锋) and incubated within his quantitative hedge fund High-Flyer Quant (幻方量化), DeepSeek funded itself through High-Flyer's trading profits, which peaked at over ¥70 billion AUM and delivered 56.55% average returns in 2025.

That self-reliance made DeepSeek's January 2026 announcement — that it was exploring external funding — a market-moving event. The numbers have escalated rapidly:

Source: Caixin, Reuters, Sina Finance, TMT Post

The $59 billion valuation cap represents a 140% increase from the $450 billion figure discussed just one month prior. If the round closes at the upper end, DeepSeek will be valued at roughly half of NVIDIA's market cap at the time of its original R1 breakthrough — a staggering multiple for a company that had never taken a dollar of venture capital.

1.2 Who Is Buying, and Why?

The investor list reveals a coalition that spans China's entire tech-industrial complex:

Source: Reuters, Tencent News, Cailianshe (财联社)

What makes this coalition unusual is its diversity. Tencent and NetEase are rivals in gaming. JD.com and Alibaba are e-commerce enemies. CATL is a battery manufacturer with no historical AI presence. And the National AI Fund — established in January 2025 with approximately $8.8 billion in capital — represents the state's direct entry into a startup cap table.

This is not a typical venture round. It is a national strategic placement disguised as a private financing event. Every major investor has a reason beyond financial returns: they need DeepSeek to exist, to succeed, and to remain Chinese.

1.3 The Valuation Debate: Bubble or Bargain?

At $59 billion, DeepSeek would be valued at approximately 40x its estimated 2026 revenue (based on API and token economics). By comparison, OpenAI's last private valuation was around $300 billion on roughly $20 billion in revenue — a 15x multiple. Anthropic, at $61.5 billion post-money after its May 2026 $6.5B round, trades at roughly 13x its estimated revenue.

Source: Company announcements, TMT Post, Reuters, LatePost, industry analyst estimates

The multiple looks extreme. But DeepSeek's investors are not buying a revenue stream. They are buying:

1. A national asset in an era of U.S. chip export controls

2. The world's most efficient training methodology (MLA, auxiliary-loss-free MoE)

3. An open-weight ecosystem that has become the default for Chinese developers and a significant share of Southeast Asian, Middle Eastern, and African AI deployments

4. A hedge against NVIDIA dependency — DeepSeek's models run on fewer, less advanced GPUs than their Western counterparts

Whether the valuation holds depends on DeepSeek's next move: the rumored R3 model, expected in June 2026, which could justify the price or burst the bubble.

2. Zhipu AI: Building the "A+H" Dual-Listing Empire

2.1 From "Global LLM First Stock" to "A+H" Pioneer

On January 8, 2026, Zhipu AI became the first pure-play LLM company to list on any global exchange, debuting on the Hong Kong Stock Exchange at HK$116.20 per share. The IPO raised approximately HK$4.17 billion ($530 million) — a modest sum by Chinese tech standards, but historically significant as the "first" of a category.

Five months later, Zhipu's stock traded at HK$1,462 — a 1,161% gain from its IPO price. Its market capitalization exceeded HK$6,500 billion (approximately $88 billion), surpassing Baidu's market cap despite generating roughly 0.56% of Baidu's annual revenue.

Source: Zhipu AI prospectus, HKEX filings, Baidu earnings reports, TMT Post

2.2 The STAR Market Filing: ¥15B for What?

On June 1, 2026, Zhipu's board approved a plan to issue between 9.1 million and 38.8 million new A-shares on the STAR Market, representing 2% to 8% of post-issuance total shares. The target: ¥15 billion (approximately $2.1 billion).

The use of funds is revealing:

Source: Zhipu AI HKEX announcement, June 1, 2026

The ¥12 billion allocation to foundation model R&D is the largest single AI R&D commitment by any publicly listed Chinese company. It signals that Zhipu — despite its soaring stock price — is doubling down on technical depth rather than acquisition or marketing.

2.3 Why List on STAR Market When HKEX Cash Is Unused?

Zhipu's Hong Kong IPO, including over-allotment, raised approximately HK$4.9 billion ($630 million). As of May 2026, the company had used only HK$2.06 billion of those funds — leaving roughly HK$2.84 billion ($360 million) untouched on its balance sheet.

So why raise another ¥15 billion so soon?

The answer is capital base architecture. Zhipu's chairman, Liu Debing (刘德兵), has publicly stated that "listing is a means to obtain long-term survival capability." Foundation model companies carry enormous costs — compute, training, inference services, and talent — that require a capital structure capable of absorbing multi-year losses without liquidity pressure.

Source: Zhipu AI filings, Caixin, TMT Post

The STAR Market listing also serves a political purpose. The STAR Market (科创板) is China's designated board for "hard technology" companies — semiconductors, biotech, advanced manufacturing, and AI. A successful Zhipu listing would validate the STAR Market as the primary venue for China's next-generation technology champions, creating a template for other AI companies to follow.

2.4 MiniMax Follows Suit

On May 29, 2026, MiniMax — Zhipu's co-entrant in the "LLM first stock" race (it listed on HKEX January 9, 2026, one day after Zhipu) — signed a tutoring agreement with CITIC Securities (中信证券) and submitted a STAR Market IPO辅导备案 to the Shanghai CSRC bureau.

Source: HKEX filings, Sina Finance, 21st Century Business Herald

The divergence in stock performance — Zhipu up 1,115% versus MiniMax up 305% — reflects investor preferences: Zhipu's enterprise-focused, open-source, domestic-market story commands a premium over MiniMax's consumer-companion, overseas-dependent model in the current geopolitical environment.

3. The Capital Tsunami: May 2026 in Context

3.1 A Record Month for Chinese AI Funding

May 2026 set a new benchmark for AI investment in China. According to industry data compiled by CSET and domestic research houses, disclosed funding in China's AI sector reached approximately ¥49.4 billion ($6.9 billion) in May alone — an 800% year-over-year increase.

Source: CSET China AI Investment Monitor; 36Kr; IT桔子; industry estimates

The ¥49.4 billion figure includes only disclosed rounds. The actual total, including undisclosed state-backed investments and strategic corporate placements, is likely significantly higher. The month also saw:

- Moonshot AI (Kimi) close its $20 billion round at a $200+ billion valuation

- Stepfun (阶跃星辰) complete a ~$2.5 billion round, removing its VIE structure for a Hong Kong IPO

- 01.AI (零一万物) raise a $300 million extension round

- SparsCore AI (墨芯人工智能) close a nearly ¥1 billion ($140M) Series C for sparse-compute AI chips

3.2 The "4+6" Competitive Landscape

Chinese industry analysts now describe the market as a "4+6" structure:

Source: CSDN analysis, 51CTO, TMT Post

The consensus among Chinese tech analysts: of the approximately 1,509 large models officially counted by Chinese regulators as of mid-2025, fewer than 10 will survive as independent entities by the end of 2026. The rest will be acquired, pivot to narrow verticals, or simply close.

4. Huawei Cloud's Agentic AI Offensive

4.1 INSPIRE 2026: The Product Matrix

While capital markets obsessed over DeepSeek and Zhipu, Huawei Cloud held its INSPIRE 2026 conference in Shanghai on June 4–5, releasing a comprehensive Agentic AI stack that may prove more consequential for enterprise AI deployment than any individual model release.

Source: Huawei Cloud INSPIRE 2026 official agenda; IT之家

4.2 The "Agentic Core" Network Strategy

Huawei's Agentic AI push is not limited to cloud software. At MWC 2026 in March, Huawei unveiled Agentic Core — a network architecture designed to enable agent-to-agent (A2A) communication across 5G and future 6G networks.

Source: Huawei MWC 2026 announcements; IT之家

The significance: Huawei is building infrastructure for a world where AI agents, not humans, are the primary network endpoints. While Western discussions of "agentic AI" focus on individual model capabilities (Claude Code, AutoGPT, etc.), Huawei is building the network layer that allows agents to coordinate across devices, clouds, and geographies.

4.3 Partnership Ecosystem

Huawei's INSPIRE 2026 featured partnerships that span China's AI industrial base:

Source: Huawei Cloud INSPIRE 2026 agenda; company announcements

5. The Hangzhou Summit: Policy Meets Capital

5.1 June 6, 2026: The Inaugural China AI High-Quality Development Summit

On June 6, 2026 — the day this article is published — the inaugural China AI High-Quality Development Summit (首届人工智能高质量发展大会) convened in Hangzhou, Zhejiang Province. The summit brought together government officials from the Ministry of Science and Technology, Zhejiang provincial leadership, and representatives from DeepSeek, Zhipu, Alibaba, Huawei, and dozens of other AI companies.

The summit's timing is not coincidental. It occurs exactly one week after:

- DeepSeek's funding announcement (June 3)

- Zhipu's STAR Market filing (June 1)

- Huawei's INSPIRE conference (June 4–5)

Source: Summit agenda; Zhejiang provincial government announcements; industry analysis

5.2 Hangzhou's AI Ambition

Hangzhou, already home to Alibaba and DeepSeek, is positioning itself as China's "AI First City." The Yuhang District (余杭区) — where DeepSeek is headquartered — has announced plans to build an "AI Innovation Development Core Zone" with targets including:

Source: Yuhang District Government; 2026 GAITC (Global AI Technology Conference) announcements

6. What the World Should Watch

6.1 The Export Control Paradox

U.S. export controls on advanced GPUs were designed to slow Chinese AI development. The data suggests they may have accelerated it — by forcing efficiency breakthroughs:

Source: Reuters, Caixin, academic analysis

6.2 Global Investment Implications

Source: Reuters, Sina Finance, investment analysis

Social Comments: What Chinese Netizens Are Saying

"DeepSeek现在就是国家队的产品了，基本是国家在AI行业的标志。"

— CSDN top-voted comment on DeepSeek analysis

*"DeepSeek is now essentially a national team product — basically the symbol of China's AI industry."*

"智谱回A不是融资，是在建长期生存的资本底座。港股的28亿还没用完呢。"

— TMT Post comment section

*"Zhipu's return to A-shares isn't fundraising — it's building a capital base for long-term survival. They haven't even used the HK$2.8B from the Hong Kong IPO yet."*

"华为云的AgentArts要是真能做到零代码部署，那企业AI落地才算真正开始。"

— Zhihu technical discussion

*"If Huawei Cloud's AgentArts can truly achieve zero-code deployment, that's when enterprise AI adoption truly begins."*

"5月AI融资494亿，这泡沫比2021年元宇宙还大。"

— Weibo finance blogger

*"¥49.4 billion in AI funding in May — this bubble is bigger than the 2021 metaverse hype."*

"字节在跳动，阿里在躁动，AI六小龙被震动。"

— Xiaoyuzhou FM podcast title (December 2024, still relevant)

*"ByteDance is jumping, Alibaba is restless, and the 'AI Six Little Dragons' are shaken."*

"中国AI的打法已经明显区别于美国同行。美国靠最先进的GPU和巨额投资，中国靠效率和轻量化。"

— Xinhua News Agency comment (January 2026)

*"China's AI approach is clearly different from American peers. The US relies on the most advanced GPUs and massive investment; China relies on efficiency and lightweight models."*

Conclusion: The Convergence Phase

Three narratives converged in the first week of June 2026:

1. Capital — DeepSeek's $7 billion first round and Zhipu's ¥15 billion STAR Market filing prove that China's AI industry has graduated from angel investing to institutional-grade capital formation. The state, private tech giants, and international investors are now competing for the same cap table slots.

2. Infrastructure — Huawei's Agentic AI stack and the Hangzhou Summit's policy framework signal that China's AI strategy is shifting from "model releases" to "industrial deployment." The winners will not be determined by who has the best benchmark score, but by who can deploy agents into factories, hospitals, vehicles, and networks at scale.

3. Global Positioning — With open-weight models downloaded 700 million times (Qwen), companion apps reaching 200 million overseas users (MiniMax Talkie), and training costs compressed by 50x (DeepSeek), China's AI industry is no longer "catching up." It is running a different race — one where efficiency, distribution, and ecosystem lock-in matter more than frontier model parameters.

The question for the next 90 days: Can DeepSeek's R3 model justify a $59 billion valuation? Can Zhipu's STAR Market IPO clear regulatory review? Can Huawei's AgentArts platform achieve the enterprise adoption that has eluded every AI agent product to date?

The answers will determine whether June 2026 is remembered as the peak of a bubble — or the beginning of China's AI industrial age.

Related Articles:

- China's AI Unicorn Stampede: Kimi's $20B Round, DeepSeek's Valuation Surge, and the Great Model Shakeout

- MiniMax: The 212 Million User AI Companion Empire

- DeepSeek-V4: The Million-Token Context Model That Redefined China AI Sovereignty

- ByteDance Doubao: The 200 Million User AI Assistant

Data Sources:

- Caixin, Reuters, Sina Finance, TMT Post, Cailianshe (财联社)

- Huawei Cloud INSPIRE 2026 official agenda and announcements

- Zhipu AI HKEX filings (June 1, 2026)

- MiniMax HKEX filings and CSRC tutoring records

- CSET China AI Investment Monitor

- Yuhang District Government AI development plan

- 2026 GAITC (Global AI Technology Conference) proceedings

- Industry analyst estimates (IDC, LatePost, 36Kr)

*Last updated: June 6, 2026*

*Word count: ~3,400 words*

*Reading time: 17 minutes*