Zhipu AI Stock Surges 70% After Musk Endorses Chinese AI Models
Zhipu 2513.HK stock surges 71% after Musk says Chinese AI could match Anthropic Fable by Q1 2027. GLM 5.2 open source.
When Elon Musk replied “Probably Q1” to a question about when Chinese AI models could match Anthropic’s Fable, he probably didn’t expect to move a stock 70% in a week. But that is exactly what happened to Zhipu AI (2513.HK), the Beijing-based foundation model company whose shares surged from HK$1,134 to HK$1,940 in five trading days after Musk’s comment went viral on Chinese social media.
The Musk Effect on Chinese AI
The exchange, which took place on X (formerly Twitter) in mid-June 2026, was brief but explosive. A user asked Musk when Chinese large language models would reach the capability level of Anthropic’s Fable — the model that the U.S. government had just added to its export restriction list. Musk’s one-line answer — “Probably Q1” (referring to Q1 2027) — was interpreted by Chinese markets as validation that the country’s AI industry is closing the gap with American frontier models faster than consensus expected.
Grok, Musk’s own AI assistant, confirmed the exchange: “Yes, Elon recently replied in a discussion noting Chinese AI models (like Zhipu’s GLM) could reach Anthropic’s Fable level ‘Possibly by Q1.’” The confirmation amplified the signal, turning a casual X reply into a market-moving event.
| Metric | Before Musk Comment | After (5 Days) | Change |
|---|---|---|---|
| Zhipu (2513.HK) stock price | HK$1,134 | HK$1,940 | +71% |
| Zhipu market cap | ~$18B | ~$31B | +$13B |
| Daily trading volume | ~HK$200M | HK$2.1B | 10x increase |
| Social media mentions (Weibo) | ~5,000/day | 180,000/day | 36x increase |
The stock move is remarkable not just for its magnitude but for its speed. Zhipu went from a relatively obscure Hong Kong-listed AI company to the most talked-about stock in China’s tech sector in under a week. The company’s market capitalization surged by approximately $13 billion — making founder Tang Jie, who holds a significant stake, a billionaire on paper.
GLM 5.2: The Open-Source Catalyst
Musk’s comment landed in a specific context that amplified its impact. Days before the X exchange, Zhipu had released GLM 5.2 as a fully open-source model — a strategic decision that positioned the company as China’s answer to Meta’s Llama and Mistral’s open-weight models. The announcement, made by Zhipu co-founder and chief scientist Tang Jie, carried an explicit ideological message: “Frontier intelligence belongs to everyone.”
The timing was deliberate. The U.S. Commerce Department had just added Anthropic’s Fable 5 to its export restriction list, prohibiting the sale of the model’s weights and API access to entities in China, Russia, and other restricted countries. Zhipu’s response — open-sourcing GLM 5.2 for free — was a direct challenge to the premise of export controls: that restricting access to frontier AI models would slow Chinese AI development.
“The U.S. government restricts Anthropic’s best model from China, and China’s response is to give away a competitive model for free,” said a Beijing-based AI researcher who asked not to be named. “It’s a propaganda victory as much as a technical one.”
How Good Is GLM 5.2?
The honest assessment is: very good, but not frontier. GLM 5.2 performs competitively on standard benchmarks — MMLU, HumanEval, GSM8K — placing it roughly in the same tier as GPT-4 and Claude 3.5 Sonnet. On Chinese-language tasks, it outperforms all Western models. On English-language reasoning and coding, it trails Anthropic’s Fable and OpenAI’s GPT-5 by a meaningful but narrowing margin.
The gap matters less than the trajectory. Two years ago, Chinese models were a full generation behind. Today, they are six to nine months behind, according to most independent evaluations. Musk’s “Probably Q1” estimate implies he believes the gap will close to near-zero within nine months — a timeline that most Chinese AI researchers consider plausible but Western researchers consider optimistic.
The Export Control Paradox
UPSTREAM: The U.S. export controls on AI chips (NVIDIA H100/A100) and AI models (Anthropic Fable) were designed to maintain American AI supremacy. Instead, they have accelerated Chinese self-sufficiency. Zhipu, DeepSeek, and other Chinese labs have developed training techniques that extract more performance from less hardware, and the open-source movement means that restricted models’ capabilities are being replicated through alternative architectures.
DOWNSTREAM: The stock market is pricing in a world where Chinese AI is competitive. Zhipu’s $31 billion market cap — for a company with estimated annual revenue of $200-300 million — reflects investor belief that the company will capture a significant share of China’s enterprise AI market, which is projected to reach $50 billion by 2028.
BOTTLENECKS: The real constraint is compute. Even with software optimization, training frontier models requires massive GPU clusters that are increasingly difficult for Chinese companies to acquire. Huawei’s Ascend chips are an alternative, but they lag NVIDIA’s latest offerings by 20-30% in raw performance. Zhipu’s ability to close the gap with Western models depends as much on hardware access as on algorithmic innovation.
What the Market Is Telling Us
BULL CASE: Zhipu’s open-source strategy creates a developer ecosystem around GLM that drives enterprise adoption. The Musk endorsement attracts international attention and partnership inquiries. By 2027, Zhipu is the leading AI platform in China, with revenue exceeding $1 billion and a market cap above $50 billion.
BEAR CASE: The stock is a bubble driven by retail FOMO and Musk hype. Zhipu’s revenue is too small to justify the valuation, and the open-source model commoditizes AI, destroying pricing power. The stock crashes 50%+ when Q2 earnings disappoint.
BASE CASE: Zhipu maintains its position as a top-3 Chinese AI lab, growing revenue to $500-800 million by 2027 through enterprise contracts and cloud services. The stock corrects 20-30% from its peak but maintains a premium valuation due to scarcity value (few pure-play AI companies are publicly listed in Hong Kong).
WHAT TO WATCH: Zhipu’s Q2 2026 earnings (revenue growth rate is the key metric); GLM 5.2 adoption metrics (downloads, API calls, enterprise deployments); any U.S. government response to the open-source challenge; and whether Musk makes further comments about Chinese AI capabilities.
CII Analysis
Zhipu’s 70% stock surge is a symptom of a larger shift: the market is pricing in a world where Chinese AI is no longer playing catch-up but competing for leadership. Musk’s casual “Probably Q1” endorsement mattered not because he is an AI expert (he isn’t, despite xAI), but because he is the most visible technology leader in the world, and his comment forced a global audience to take Chinese AI seriously. The deeper story is the export control paradox: every restriction the U.S. imposes on Chinese AI access accelerates Chinese self-sufficiency. GLM 5.2 being open-sourced days after Fable was restricted is not a coincidence — it is a strategic countermove in a technological cold war. For investors, the lesson is that Chinese AI companies are investable, but the valuations are extreme. Zhipu at $31 billion on $200-300 million revenue is priced for perfection. The risk-reward is unfavorable at current levels, but the sector as a whole — DeepSeek, Moonshot, MiniMax, Baichuan — deserves serious attention.
Related reading:
Sources
- CNBC — Zhipu surges as Wall Street raises bets on China AI
- Pandaily — Musk Says Chinese LLMs Could Reach Fable Level by Q1 2027
- Fortune — Anthropic disables Fable models after U.S. export controls
- Reddit — GLM 5.2 open source announcement
- Yahoo Finance — Zhipu (2513.HK) stock data
