Zhipu AI: a large-model company's Hong Kong IPO

Zhipu AI is a foundation-model provider whose revenue grew at a 130% compound annual rate from 2022 to 2024, from RMB 57.4 million to RMB 312 million. It serves about 12,000 institutional customers, more than 45 million developers and more than 80 million devices, and has released agent products including AutoGLM.

At a glance

Item	Detail
Company	Zhipu AI
Industry	Technology
Use cases	Business intelligence, recommendations
Functions involved	R&D, business intelligence
Reported results	Valued at US$4.615B (about RMB 30B) in Feb 2025; more than RMB 10B raised in total; about RMB 4.4B cumulative R&D; Hong Kong IPO offering opened Dec 30, 2025, with 37,419,500 H shares on offer
Moments rating	A (recommended)
AI scenario score	82

Case details

Background

Zhipu, formally Beijing Zhipu Huazhang Technology Co., was founded in June 2019. In April 2025 it filed for pre-listing tutoring with the Beijing office of the securities regulator, with CICC as tutor, and was then tipped to become the first large-model stock on China’s A-share market. The same report put its February 2025 valuation at US$4.615 billion (about RMB 30 billion) after more than ten funding rounds totaling over RMB 10 billion. The company later switched to Hong Kong, passing the HKEX listing hearing and publishing its prospectus on December 19, 2025.

What they did

Zhipu has released agent products including AutoGLM and AutoGLM Rumination (compare Doubao vs AutoGLM), and partnered with Samsung on agent features for the Galaxy S25 series. Its GLM base models are updated roughly every three to six months. The business is mainly MaaS (model as a service), split between on-premises and cloud deployment. In 2024, dedicated private models brought in RMB 264 million, over 80% of revenue and largely one-off projects; token-billed cloud services were under 20%. R&D spending rose from RMB 84.4 million in 2022 to RMB 2.195 billion in 2024, about RMB 4.4 billion in total.

Results and data

• Revenue: RMB 57.4M in 2022, RMB 125M in 2023, RMB 312M in 2024 (130% CAGR); RMB 191M in H1 2025, up 325%.
• Losses: about RMB 144M, 788M and 2.958B in 2022–2024; a net loss of RMB 2.358B in H1 2025. Compute service fees were RMB 1.552B in 2024.
• Scale: as of September 30, 2025, about 12,000 institutional customers, more than 45 million developers and more than 80 million devices.
• Listing: the offering opened December 30, 2025, with 37,419,500 H shares offered globally and a planned January 8, 2026 listing on the HKEX main board under code “2513”. 70% of net proceeds go to general large-model R&D. The reports reviewed do not state total proceeds.

What to take from it

With private deployment above 80% of revenue and cloud API calls under 20%, Chinese enterprises still buy large models mainly as on-premises deployments. When choosing a vendor, companies can weigh its R&D spending and model update cadence, while watching how the size of its losses could affect long-term service.

Limits

This is a capital-markets event for a model vendor, not an enterprise AI deployment, so it cannot be used directly to estimate your own costs and returns. The financial figures come from media summaries of the prospectus; the HKEX filings are authoritative.

Sources

Original source: Zhipu AI case material (Securities Times) Additional source: The Paper · Zhipu passes HKEX listing hearing Additional source: Wallstreetcn · Prospectus analysis Additional source: Securities Daily · Zhipu opens Hong Kong IPO

This entry is industry research: a publicly documented, third-party AI case. It is not a Moments client case. For work Moments delivered itself, see client cases.