Daily Report on Hong Kong-Listed LLM Concept Stocks (July 21, 2026)

NewTimeSpace · 2026/07/21 17:41 · By Zeyu Wu

On July 21, Hong Kong equities consolidated, while the LLM and semiconductor supply chains staged a powerful rebound. ZHIPU AI skyrocketed nearly 37% after launching a 1GW all-domestic-chip data center and acquiring Zhongke Jiahe to solidify its infrastructure capabilities; MiniMax and XUNCE also posted significant gains. The semiconductor sector erupted across the board, driven by robust AI hardware demand. On the industrial front, Tencent Hunyuan launched Hyra-1.0, an agent capable of recursive self-improvement; ZHIPU AI operationalized its 10k-chip domestic computing clusters; and Alibaba released Qwen-Image-3.0, supporting ultra-long contexts for complex visual generation. This highlights the accelerating integration of foundational compute infrastructure and frontier agentic engineering.

I. Industry Observations

On July 21, 2026, Hong Kong stocks consolidated within a narrow range. The Hang Seng Index (HSI) closed down 0.04%, the Hang Seng TECH Index rose 1.32%, and the Hang Seng China Enterprises Index (HSCEI) edged down 0.25%. Total daily turnover reached HKD 289.943 billion, with Southbound trading recording a net inflow of HKD 7.167 billion.

The AI and LLM supply chains staged a robust rebound. ZHIPU AI skyrocketed 36.9% as its 1GW all-domestic-chip computing data center and top-tier infrastructure deployed simultaneously, complemented by the completion of its acquisition of Zhongke Jiahe, effectively rounding out its dual critical capabilities in compute supply and deployment. MINIMAX-W surged 15.2%.

The semiconductor sector rebounded across the board: SMIC rose 8.2%, Hua Hong Semiconductor jumped 17.9%, Montage Technology surged 17.1%, GigaDevice advanced 15.1%, and Iluvatar CoreX gained 12.1%. Lenovo Group climbed 8.5% on news that its AI server orders on hand have exceeded USD 21 billion.

II. Market Performance of LLM Concept Stocks

Stock Code	Stock Short Name	Closing Price (HKD)	Change
02513.HK	KNOWLEDGE ATLAS	1,219.00	36.89%
03317.HK	XUNCE	126	16.99%
00100.HK	MINIMAX-W	222.4	15.17%
00020.HK	SENSETIME-W	1.4	2.94%
02718.HK	MININGLAMP-W	115.4	2.76%
09988.HK	BABA-W	117	0.17%
01024.HK	KUAISHOU-W	45.1	-0.35%
00700.HK	TENCENT	474	-0.80%
03690.HK	MEITUAN-W	85.2	-1.27%
09888.HK	BIDU-SW	107.2	-1.83%

Data Source: HKEX. For reference only, not investment advice. Data as of market close onJuly 21, 2026.

III. Industry and Corporate Dynamics

1. Tencent Hunyuan: Launches Hyra-1.0 Scientific Discovery Agent, Supporting Recursive Self-Improvement

On July 21, Tencent Hunyuan officially announced the release of Hyra (Hunyuan Research Agent) version 1.0. According to Tencent, this is a performance-oriented research and engineering agent capable of recursive self-improvement. Beyond model R&D and scientific discovery, Hyra can be deployed in open-ended scenarios such as gaming, design, and content creation, continuously iterating its strategic outputs based on self-play, self-evaluation, and user feedback. In the "AI for AI" domain, Hyra surpassed previously public records in three benchmarks: NanoChat Autoresearch, NanoGPT Speedrun, and SOL-ExecBench. In the "AI for Science" domain, it established new historical best results on 29 out of 55 open mathematical problems.

2. ZHIPU AI: Completes 1GW All-Domestic-Chip Data Center, Entering Partial Operational Phase

On July 21, ZHIPU AI announced the completion of a 1GW-scale data center powered entirely by domestically developed AI chips. The peak power supply scale of this single campus can concurrently meet the daily electricity demands of 750,000 standard households. The facility has now entered a partial production and operational phase. ZHIPU AI has currently built or operates multiple computing clusters, each equipped with over 10,000 chips.

3. Alibaba: Releases Qwen-Image-3.0 Foundation Model, Supporting 4.5k-Token Ultra-Long Contexts

On July 21, Alibaba released its latest foundation model for image generation, Qwen-Image-3.0, marking the third generation in the Qwen image generation and editing series. The new model supports an ultra-long input of up to 4.5k tokens, enabling the one-shot generation of multi-element integrated outputs—such as knowledge diagrams encompassing formula symbols, geometric shapes, and logical derivations, as well as complex UI interfaces. It natively supports rendering in 12 languages and over 20 font styles. The model is currently online for beta testing by invitation in mainland China and international regions.

IV. Introduction to LLM Concept Stocks

Stock Name	Stock Code	Market Position	LLM Products & Business Profile
KNOWLEDGE ATLAS	02513.HK	A leading player in China's independent LLM sector	Knowledge Atlas Technology Joint Stock Company Limited (02513.HK; commercially known as Zhipu AI) is a leading player in China's independent large language model (LLM) sector (recognized for its capabilities and market position by Frost & Sullivan). The company has released its next-generation flagship model, GLM-5.2, achieving open-source State-of-the-Art (SOTA) performance in coding and agentic capabilities. It has also open-sourced the multimodal image generation model GLM-Image in collaboration with Huawei. Focusing on novel model architecture design, generalized reinforcement learning paradigms, and autonomous model evolution, its business layout closely aligns with the trend of enterprise-level AI productivity transformation.
MINIMAX-W	00100.HK	A top-tier player dedicated to omni-modal Artificial General Intelligence (AGI)	The company continuously iterates its video generation model matrix and has partnered with mainstream domestic and overseas chip manufacturers and inference platforms to complete underlying adaptations. The total monthly active users (MAU) of its B2C and B2B LLM applications, along with multimodal API call volumes, maintain rapid growth, further broadening its computing power ecological layout.
MININGLAMP-W	02718.HK	LLM concept stock, dubbed the "First Agentic AI Stock in Hong Kong"	Mininglamp Technology is dubbed the "First Agentic AI Stock in Hong Kong." It has released DeepMiner, a multi-agent collaborative data analysis platform for commercial data analytics, and the open-source edge GUI (Graphical User Interface) model series, Mano. Deeply engaged in the enterprise service and data intelligence tracks, the company adopts a multi-agent architecture focusing on commercial data analysis. By utilizing structured factual data to mitigate AI hallucination risks, it provides decision-making agent pipelines for the marketing and mass consumption sectors. Additionally, it has launched AI Native hardware to expand the carriers for its agents.
SENSETIME-W	00020.HK	A leading domestic provider of computer vision and generative AI infrastructure	Accelerating its transition towards Model-as-a-Service (MaaS) and generative AI services. The company has released and open-sourced the SenseNova U1 series, a native unified understanding and generation model, and the lightweight multimodal model SenseNova 6.7 Flash-Lite. Leveraging the computing reserves of its AI infrastructure (AIDC), it translates foundational computing power into omni-modal generation pipelines tailored for vertical industries such as automotive and healthcare.

NewTimeSpace Disclaimer: All content herein is the original work of NewTimeSpace. Any reproduction, reprinting, or use of this content in any other manner must clearly indicate the source as "NewTimeSpace". NewTimeSpace and its authorized third-party information providers strive to ensure the accuracy and reliability of the data, but do not guarantee the absolute correctness thereof. This content is for reference only and does not constitute any investment advice. All transaction risks shall be borne by the user.

Tags：

LLM Knowledge Atlas MININGLAMP-W