Zhipu AI Hits $112B Record Valuation, Defying Concerns

Shares of Hong Kong-listed Chinese AI developer Zhipu AI (SEHK: 2513) surged to an all-time high on May 29, briefly reaching HK$1,993 intraday and pushing the company’s market capitalization above HK$880 billion ($112 billion), according to Caixin Global. The milestone cements Zhipu’s position as one of the world’s most valuable AI companies, with the stock climbing nearly 1,600% since its January 2026 IPO at HK$116.20.

A Meteoric Rise Rooted in Fundamentals

Zhipu, founded in 2019 by researchers from Tsinghua University, is one of China’s six “AI tigers” — startups building large language models to rival OpenAI and Anthropic. It was the first of this cohort to go public via an IPO, making its Hong Kong Stock Exchange debut on January 8, 2026, a landmark moment for China’s AI sector.

The company’s 2025 financial results, released on March 31, revealed revenue of 724 million yuan ($99 million), up 132% year-on-year, as detailed by 1950.ai. Yet the headline growth masks a deeper story: Zhipu’s adjusted net loss widened to 3.18 billion yuan as research and development spending surged 45% to 3.2 billion yuan. Gross margins contracted from 56% to 41%, reflecting a shift toward lower-margin cloud deployments.

CEO Zhang Peng defended the spending, stating that “engineering optimizations and platform scaling are essential to maintaining leadership in a market defined by rapid model iteration and computational demand.”

The “Volume and Price Simultaneously Increase” Phenomenon

What has captivated investors is Zhipu’s unusual pricing power. In the first quarter of 2026, the company raised API call pricing by a cumulative 83% — yet call volume still increased by 400%, according to 36Kr. This “simultaneous increase in volume and price” is nearly unprecedented in an AI industry otherwise defined by aggressive price wars.

As of March 2026, Zhipu’s annualized recurring revenue (ARR) from its open platform and API offerings reached 1.7 billion yuan ($250 million), a 60-fold increase from the previous year. Chairman Liu Debing explained that the company employs a tiered token model: low-complexity tasks leverage low-cost or ad-supported tokens, while high-complexity tasks maintain sustained pricing power.

GLM-5: A Technological Milestone on Domestic Chips

Zhipu’s February 2026 release of GLM-5, its flagship large language model, marked a significant technological achievement. The model was developed using domestically manufactured chips, including Huawei’s Ascend processors, underscoring China’s progress toward semiconductor self-sufficiency despite U.S. export restrictions. As The Diplomat noted, GLM-5 can directly compete with Anthropic’s Claude Opus 4.5 in coding benchmarks and surpasses Google’s Gemini 3 Pro in certain evaluations.

The open-source MIT License strategy mirrors Meta’s approach to building ecosystem adoption, and Zhipu’s pricing — around $0.80 per million input tokens — undercuts Western rivals by a factor of six.

The Valuation Debate: Narrative or Fundamentals?

Zhipu’s market cap of over HK$700 billion implies a price-to-sales ratio of approximately 890x. By comparison, rival MiniMax trades at roughly 425x PS, while OpenAI commands about 66x PS. The 36Kr analysis notes that the HK$400 billion valuation gap between Zhipu and MiniMax cannot be explained by fundamentals alone — both companies reported similar revenues (724 million yuan vs. 569 million yuan) and both remain deeply unprofitable.

Instead, analysts point to “narrative economics”: Zhipu’s “Agentic AI” story aligns perfectly with global trading themes, creating a self-reinforcing cycle where rising stock prices attract more funds. UBS analysts cited by Caixin note that lofty valuations also reflect a scarcity premium and limited tradable shares due to lock-up periods.

Geopolitical Risks and the Road Ahead

Zhipu’s inclusion on the U.S. Commerce Department’s Entity List in January 2025 remains a critical risk factor. The restriction on access to advanced U.S. semiconductor technology has forced the company to accelerate adoption of domestic chips — a constraint that may limit performance compared to models trained on NVIDIA’s most advanced GPUs.

Professor Ming Zhao of Tsinghua University observed that “Zhipu exemplifies how independent AI model developers in China can achieve both technological parity and market scalability with global counterparts, while simultaneously navigating policy and supply chain constraints.”

What to Watch

Key questions remain unanswered. Can Zhipu maintain its pricing power as competitors like DeepSeek and Xiaomi slash prices aggressively? When will the company reach profitability, given that R&D spending is growing faster than revenue? And what happens when insider lock-up periods expire, potentially flooding the market with additional shares?

For now, Zhipu’s record valuation represents a triumph of market narrative as much as business fundamentals — a story that speaks to China’s AI ambitions, investor psychology, and the extraordinary premium the market places on perceived winners in the global AI race.