Summary: China’s first pure-play large language model developer now trades on the public market. Spun out of Tsinghua University in 2019, Zhipu ranked among the earliest Chinese teams to bet on general-purpose large language models. On Thursday, January 8, 2026, Zhipu AI — registered as Knowledge Atlas Technology under ticker 2513.HK — began trading on the Hong Kong Stock Exchange.
China’s first pure-play large language model developer now trades on the public market. On Thursday, January 8, 2026, Zhipu AI — registered as Knowledge Atlas Technology under ticker 2513.HK — began trading on the Hong Kong Stock Exchange. The shares priced at HK$116.20, opened 3.3% above the line, and finished the session up 13.2% at HK$131.50.
The raise brought in HK$4.35 billion (about US$558 million) at a valuation approaching HK$51 billion (roughly US$6.5 billion). With OpenAI and Anthropic still private, “the first frontier LLM company to list” stands as a milestone on its own — and rival MiniMax (0100.HK) followed one day later, with image-sensor maker OmniVision waiting in the queue.
Debut Day: Up 3.3% at the Open, 13.2% at the Close
Zhipu sold roughly 37.4 million shares and shared its debut day with two other tech listings. Per Reuters, the trio — Zhipu, GPU designer Shanghai Iluvatar CoreX (9903.HK; raised HK$3.48 billion, closed up 8.4%), and surgical-robotics firm Shenzhen Edge Medical (2675.HK; up 30.9%) — pulled in a combined HK$9.3 billion (US$1.19 billion).
The backdrop is a Hong Kong market on the mend: 115 listings raised US$37.2 billion through 2025, the strongest year since 2021 on LSEG data as of January 5. Zhipu’s first-day tape is being read as the thermometer for whether the 2026 AI listing wave holds its heat.
Who Zhipu Is: Tsinghua Roots and the Entity List
Spun out of Tsinghua University in 2019, Zhipu ranked among the earliest Chinese teams to bet on general-purpose large language models. Its GLM family, branded Z.ai, ships open weights alongside paid APIs and enterprise deployments. Cornerstone investors included the JSC International Investment Fund and JinYi Capital’s multi-strategy fund. Per CNBC, the company booked 312.4 million yuan (about US$43 million) of revenue in 2024 and plans to steer 70% of IPO proceeds into general-purpose model R&D.
Two facts stretch this listing beyond the first-day pop. First, the US Commerce Department added Zhipu to its Entity List in January 2025 over alleged military ties, restricting access to advanced chips and American expertise; the company operates out of the UK, Singapore, Malaysia, and the Middle East, and executives describe the overseas push as “normal business.” Second, OpenAI publicly tagged Zhipu in June 2025 as a fast-rising Chinese rival. A company severed from US technology has just raised US$558 million in public markets — the sharpest pricing signal yet for a Chinese AI stack engineered around US export controls.
China Is Fast-Tracking AI and Chip Listings
Reuters reports that Chinese regulators are accelerating AI and semiconductor listings to fortify domestic alternatives to US technology. In the pipeline: xFusion, the Huawei AI-server spin-off, has hired Citic Securities for a mainland IPO, while memory maker ChangXin Memory and Baidu’s AI-chip unit Kunlunxin are charting their own paths to market. Zhipu and MiniMax trading on consecutive days form the visible crest of that policy wave.
Market Read: Modest Gains Do Not Mean Fading Enthusiasm
“Modest first-day gains should not be interpreted as fading AI enthusiasm,” Marco Sun, chief financial markets strategist at MUFG China, told Reuters. “Early-stage investments have yet to be cashed out for profit… China’s AI story is only beginning to unfold.” Janice Hu, UBS’s China head, laid out the scale gap: China’s listed AI and high-tech companies carry a combined market value near US$5 trillion against roughly US$30 trillion in the US, with no trillion-dollar player yet — “only a matter of time,” in her phrasing.
What It Means for Developers and Product Teams
Three takeaways. First, transparency: a frontier LLM lab is now a reporting company — revenue mix, R&D spend, and use of proceeds arrive on a fixed schedule, handing observers the only public financial sample of a Chinese model business. Second, the open-source route faces its capital-markets exam: GLM open weights plus paid API is written into the prospectus as the business model, not marketing gloss. Third, geographic hedging: an Entity List company completing a US$558 million raise proves Hong Kong will price supply chains built around US export controls. If your product leans on a single model vendor, treating open-weight options like GLM as backup infrastructure has been promoted from model selection to supply-chain management.