From Zero to $1.2T: Zhipu AI's Open-Source Wins Big as Anthropic Stumbles
Key Takeaways
- Chinese AI startup Zhipu AI saw its Hong Kong-listed entity soar to a $1.2 trillion valuation after open-sourcing its GLM-5.2 model, capitalizing on the regulatory missteps that crippled Anthropic.
- The episode illustrates how startups can weaponize open-source and jurisdictional agility.
Mentioned
Key Intelligence
Key Facts
- 1On June 12, 2026, the US Commerce Department ordered Anthropic to block foreign access to its Fable 5 and Mythos 5 models, forcing a global shutdown and downgrading subscribers to an older model.
- 2Anthropic’s own technical review found the ‘jailbreak’ vulnerability was minor and already existed in other publicly available models, undercutting the official justification.
- 3On June 15, Zhipu AI announced the open-source release of its GLM-5.2 foundation model, triggering a 48% intraday surge in shares of its Hong Kong-listed entity, Knowledge Atlas Technology.
- 4By June 22, Knowledge Atlas shares hit an intraday high of HK$2,980 (US$380), pushing market capitalisation briefly past HK$1.2 trillion — a 25-fold increase from its January 2026 listing debut.
- 5The episode highlights the failure of treating software model controls like physical export restrictions, as code moves instantly and denial accelerates foreign development of alternatives.
Knowledge Atlas Technology (Zhipu AI)
Company- Founded
- 2019
- Employees
- ~500
Hong Kong-listed entity for Chinese AI startup Zhipu AI. The company released the open-source GLM‑5.2 foundation model in June 2026, sparking a historic stock rally.
Market cap peaked above HK$1.2T
Analysis
For startup founders and VCs, the Zhipu–Anthropic saga is a masterclass in strategic positioning. While Anthropic, a well-funded US AI lab, was paralyzed by its government's export controls, Zhipu pivoted decisively to an open-source model and a Hong Kong listing. The result: a 48% one-day stock pop and a 25x overall return inside six months. It's a stark reminder that in the AI race, regulatory resilience and market access matter as much as the technology itself.
The sudden and chaotic US regulatory action against Anthropic on June 12, 2026, and the explosive market response to Zhipu AI's open-source release just three days later, expose the deep contradictions of America's techno-nationalist AI containment strategy. The US Commerce Department ordered Anthropic to block all foreign access to its cutting-edge Fable 5 and Mythos 5 models, citing a jailbreak vulnerability. Unable to verify user nationalities in real time, Anthropic was forced to shut down the models globally, lock out its own non-US staff, and downgrade paying subscribers to legacy systems. The twist: Anthropic's own technical review found the flaws were minor and already present in other public models, and the company had actively lobbied for the very export controls that ensnared it.
By June 22, the stock touched HK$2,980 (US$380), briefly giving the company a market capitalization above HK$1.2 trillion — a 25-fold increase from its January 2026 debut.
The market's verdict was swift and merciless. On June 15, Chinese AI pioneer Zhipu AI announced the open-source release of its GLM-5.2 foundation model. Shares of its Hong Kong listing vehicle, Knowledge Atlas Technology, vaulted 48% intraday. By June 22, the stock touched HK$2,980 (US$380), briefly giving the company a market capitalization above HK$1.2 trillion — a 25-fold increase from its January 2026 debut. This staggering capital migration underscores a structural shift: investors are pricing in a future where restrictive US controls make American AI firms less reliable, while more open jurisdictions become hubs for innovation and valuation.
At the heart of the fiasco is a category error. Washington has treated software models like physical nuclear components, applying hardware-era export controls to code that can replicate across borders in milliseconds. The result is not containment but acceleration of foreign alternatives. Since October 2022, the US has steadily tightened restrictions on advanced semiconductors, expanded entity lists, and leaned on allies to follow suit. Each move was intended to slow China's AI progress. Instead, it has spurred domestic investment, encouraged open-source strategies that erode proprietary moats, and driven capital toward platforms like Hong Kong that offer both proximity to mainland Chinese talent and a reliable common-law regulatory framework.
What to Watch
Hong Kong now stands at a historic crossroads. The Anthropic shutdown — which punished the company's own global workforce and paying customers — demonstrated that even heavily regulated US firms cannot guarantee service continuity when political winds shift. In contrast, Zhipu's open-source release proved that Chinese firms can leverage transparency as a competitive weapon. For Hong Kong, this creates a unique opportunity: if it can craft predictable, innovation-friendly AI regulations that respect both national security and global interoperability, it could attract a wave of companies seeking a safe harbor between the US and China's increasingly fragmented tech ecosystems.
The implications extend far beyond a single stock rally. The episode suggests that further US export controls may trigger similar blowback, accelerating the diffusion of AI capabilities while reducing Washington's ability to set global norms. For companies, the lesson is clear — over-compliance with unilateral controls carries severe business risk. For regulators, the challenge is to design rules that protect legitimate security interests without inadvertently destroying the competitiveness of one's own industries. As the article argues, "software cannot be embargoed like hardware." This reality, married to the explosive growth in Hong Kong's AI-linked valuations, points to a future where regulatory overreach becomes a primary driver of both market dislocation and techno-political realignment.
How we covered this story
Every story in our startup coverage is assembled from multiple primary sources, cross-referenced for factual consistency, and scored along three independent dimensions: sentiment, operational impact, and source-cluster confidence. Single-source rumors and unverifiable claims do not pass our editorial gate. When a story shows "Verified by N sources" with N≥2, the development is independently corroborated; when N=1, we mark it explicitly so readers can weigh the signal accordingly.
Impact scoring uses a 1-10 scale weighted toward regulatory, financial, and operational consequence rather than coverage volume. A topic that runs in every outlet but moves no real decisions ranks lower than a niche regulatory filing that reshapes how operators in the startup space have to behave. Read our full methodology for the scoring rubric, our glossary for term definitions, and our trends index for the longitudinal view across the beat.
| Signal on this page | What it tells you |
|---|---|
| Verified by N sources | Independent corroboration count. N≥2 is our confidence floor; N=1 is marked explicitly. |
| Impact score (1-10) | Regulatory + financial + operational weight. 8+ signals an experienced-operator action item. |
| Sentiment | Five-tier classification trained on labeled startup-specific corpora. |
| Timeline | Where applicable, the related-events sequence that contextualizes today's development. |