Zhipu's Market Cap Surge and Rivalry | Generated by AI

Home 2026.06

Question: What’s the current state of Zhipu (GLM models) — is its market cap really larger than Xiaomi + Kuaishou (short video)?

Answer:

This story is wild and very fresh — Zhipu had a 48% single-day surge just three days ago.

Zhipu / Z.ai — Current State (June 2026)

Market Cap Right Now

As of June 16, 2026, Zhipu’s stock price is $188.14 HKD, giving it a current market cap of ~$60.4B USD (446M shares).

But that number is misleading because it’s post-crash from peak. Zhipu hit a historical peak of HK$1,993 per share on May 29, with market cap briefly exceeding HK$880 billion — nearly a 40x increase from its IPO price — before crashing by HK$300–400B in just a few trading days.

Then it bounced hard again: the stock surged as much as 48% to HK$1,620 on June 16 after releasing GLM-5.2 and a million-token context window, closing up 32.8% at HK$1,457.

vs. Xiaomi + Kuaishou

As of May 27–28, 2026, Xiaomi Group’s total market cap was ~HK$740 billion. Kuaishou (short video) trades around HK$400–500B range. So at peak (HK$880B), Zhipu was approaching Xiaomi alone but probably not Xiaomi + Kuaishou combined. The claim likely refers to the peak or a specific trading day.

Why the Insane Rally

Two catalysts stacked:

  1. On June 8, Zhipu was officially included in the Hang Seng Tech Index and Hong Kong Stock Connect — forcing passive index funds to buy.

  2. On June 12, Anthropic cut off non-US users’ access to Claude Fable 5 and Mythos 5 due to US government export control orders, paralyzing hundreds of millions of overseas users. Zhipu reacted immediately — on June 13 it opened GLM-5.2 fully, emphasizing its 1M token context, to absorb displaced Claude users.

Fundamentals vs. Narrative

The valuation is almost entirely narrative-driven: a P/S ratio of 576x means that even at 100% revenue growth for 5 straight years, the current price is still extremely expensive — at least 80% is the scarcity premium from the “Chinese OpenAI” narrative plus a liquidity bubble from small tradable float.

Actual business: 2025 revenue was 724M yuan (~$100M), up 132% YoY, with 73.7% from government and financial clients. Adjusted net loss was 31.82B yuan, growing 29% YoY.

Tech Reality

On SWE-bench Pro, GLM-5.1 achieved a pass rate of 58.4% — topping GPT-5.4 (57.7%) and Claude Opus 4.6 (57.3%). Real model quality is genuinely competitive, especially on coding/agent tasks.

GLM-5.2, released June 16, supports a 1M-token context window (up from GLM-5.1’s 200K). In tests, it showed performance comparable to Opus 4.8 on long-context tasks like 740K-line logs and thousands of lines of code.

Key Context for You

The geopolitical angle is the critical signal here — US export controls on frontier models are directly redirecting developer traffic to Chinese alternatives. GLM-5.2 under MIT license + 1M context is a credible substitute for many use cases. Zhipu’s pivot to MaaS and coding agents (AutoGLM, GLM Coding Plan) mirrors the exact direction you’re working in. Worth watching as a potential API provider especially since you’re already using MiMo-V2.5 from Xiaomi — the Chinese model ecosystem is consolidating fast.

References:

Back Donate