Stock Markets July 7, 2026 11:18 PM

Zhipu AI stock jumps after report it is weighing a bespoke AI chip for GLM models

Knowledge Atlas Tech (HK:2513) rallies as company holds early talks on a custom ASIC to handle surging GLM-5.2 demand amid export constraints

By Leila Farooq
Share
Twitter Reddit Facebook LinkedIn

Shares of Knowledge Atlas Tech (HK:2513), the company behind Zhipu AI, rose sharply after reports that the firm has initiated preliminary discussions with domestic chip designers about creating a bespoke ASIC to run its GLM series of large language models. The initiative is at an early stage, with no partner selected and development likely to exceed two years. The move is framed against strong early demand for Zhipu’s GLM-5.2 model and ongoing U.S. export restrictions that limit access to leading Nvidia chips.

Zhipu AI stock jumps after report it is weighing a bespoke AI chip for GLM models
Summarize with
ChatGPT Perplexity Claude Grok Gemini

Key Points

  • Knowledge Atlas Tech (HK:2513) shares rose 9.9% after reports it is exploring a custom ASIC for GLM models; broader Hang Seng gained about 2% the same day.
  • Zhipu AI has held preliminary discussions with multiple domestic chip design firms but has not selected a partner; the project is described as early-stage and development could take more than two years.
  • Demand for GLM-5.2 has surged, with daily token usage reported to have increased 27-fold in the first week after launch, increasing pressure on compute resources and prompting consideration of proprietary silicon.

Shares of Knowledge Atlas Tech Joint Stock (HK:2513) climbed on Tuesday following reporting that the company behind Zhipu AI is exploring the design of a custom application-specific integrated circuit (ASIC) intended to accelerate its GLM family of large language models.

In Hong Kong trading, Zhipu AI stock rose 9.9%, outpacing a roughly 2% advance in the Hang Seng. The gains followed accounts that the firm has held initial talks with several domestic chip design houses about building an AI processor tailored to its models.

The discussions are preliminary, and the company has not selected a partner, according to the report. Sources cited in the reporting described the project as early-stage, and estimated that development would likely take in excess of two years should Zhipu proceed.

Interest in a proprietary chip is being driven in part by very rapid uptake of Zhipu’s most recent model. The report noted that daily token usage for GLM-5.2 surged 27-fold during the first week after that model’s launch last month, a jump that has increased demand on compute capacity.

These capacity pressures come at a time when U.S. export restrictions continue to constrain Chinese access to Nvidia’s most advanced AI accelerators, according to the reporting. That dynamic has prompted a number of Chinese developers to look for alternatives that can be produced domestically or by non-U.S. suppliers.

ASICs differ from general-purpose graphics processors in that they are engineered for particular computing workloads. When deployed at scale, specialized silicon can deliver lower operating expenses and better energy efficiency for those tasks, the report said. For Zhipu, a proprietary processor could reduce dependence on third-party suppliers while lowering the cost of running inference workloads for its models.

If it moves forward, Zhipu would join other technology firms that have invested in custom AI silicon to improve performance and reduce reliance on external suppliers. The reporting listed companies that have pursued in-house processors, including Google, OpenAI, Alibaba and ByteDance, and said Chinese AI developers are increasingly seeking domestic options amid tightened export controls.

The next key development for the project would be the selection of a chip design partner, which would indicate that the plan has advanced beyond exploratory talks. Until such a partner is chosen, the initiative remains in an early, uncertain phase.


Summary: Zhipu AI's parent, Knowledge Atlas Tech (HK:2513), saw its shares jump after reports the company is exploring the development of a custom ASIC to run its GLM LLMs. The effort is preliminary, with partner selection pending and a development timeline likely exceeding two years. Rapid uptake of GLM-5.2 and U.S. export limits on advanced Nvidia chips are cited as contextual drivers.

Risks

  • Project remains at an early stage with no partner chosen, creating uncertainty about whether and when a custom chip will be produced - impacts semiconductor and AI infrastructure sectors.
  • Estimated development timeline likely exceeds two years, meaning benefits to operating costs and supply independence are not immediate - impacts cloud/datacenter and AI service economics.
  • U.S. export restrictions limiting access to Nvidia's most advanced AI chips are a cited driver; continued constraints could shape procurement and deployment strategies but also introduce supply-chain uncertainty for Chinese AI developers - impacts chipmakers and enterprise AI deployments.

More from Stock Markets

Inpex Shares Jump After Long-Term LNG Pact with ADNOC; CCS Drilling Begins in Japan Jul 7, 2026 BHP Faces Potential Disruption at Port Hedland After Unions Announce Eight-Hour Stop Work Jul 7, 2026 BHP Port Hedland workforce to stage eight-hour stop work on July 16 Jul 7, 2026 Asia markets steady as AI chip stocks rebound, but oil and geopolitics restrain gains Jul 7, 2026 Wall Street Futures Hold Steady as Iran Tensions and Fed Minutes Command Attention Jul 7, 2026