Z.ai Stock Jumps After Launching New AI Model That Runs Exclusively on Chinese-Made Chips
Table of Contents
You might want to know
1. Can an advanced AI model truly operate entirely on domestically produced chips without external components?
2. What are the broader implications for China’s tech self-reliance and the global AI hardware market?
Main Topic
BEIJING — Z.ai, a Chinese artificial intelligence developer listed in Hong Kong, announced the release of a new model on Wednesday it says runs entirely on semiconductors manufactured in China. The model, launched under the name GLM-5.3-Flash and previously circulated under the code name "Ox Alpha," is described by Z.ai as a lower-cost variant of its flagship offering. The company said the model handled online inference requests using a farm of 100,000 domestically produced chips, a claim that helped lift Z.ai’s Hong Kong shares by more than 8% in Thursday trading.
According to Z.ai, GLM-5.3-Flash performed strongly on usage metrics soon after its August 20 release, ranking first by usage on the OpenRouter platform in the most recent week and placing tenth on the Artificial Analysis Intelligence Index. The company positioned the model as a cost-efficient alternative within its product line, citing both usage and accessibility advantages of the Flash variant.
CNBC and other outlets noted they were unable to independently confirm Z.ai’s assertion that the system relied solely on Chinese-made chips. Z.ai declined to provide details identifying the chip makers involved, leaving observers to infer potential suppliers. Industry analysts say that running an AI model in production for inference uses significantly less computing power than the resource-intensive training phase, which makes claims about deployment using domestic chips more plausible than a claim about training on those same chips.
Geopolitical dynamics are central to this development. Restrictions on sales of advanced Nvidia chips to China, driven by U.S. export controls, have complicated the supply of leading-edge accelerators into the country. In response, Chinese technology firms including Huawei have accelerated efforts to develop alternative processors and end-to-end computing stacks. Counterpoint Research senior analyst Ivan Lam suggested that Z.ai is likely employing Huawei’s Ascend chips among others, and that closer collaboration between Chinese chip makers, server builders, and AI model developers has become increasingly common. He emphasized that domestic players are allocating more resources toward building AI servers and computing infrastructure around homegrown semiconductors.
The move by Z.ai is therefore both a technical announcement and a signal in the broader push for technological self-reliance. Over recent years, China has invested heavily in both semiconductor manufacturing and AI capability development as a strategic response to export controls and restricted access to some foreign technologies. This context helps explain why a company’s claim to operate a production AI model entirely on domestic chips attracts significant market attention.
Market reaction extended beyond Z.ai. MiniMax, another AI company listed in Hong Kong, saw its shares rise by roughly 3% following the release of its earnings, which reported a 283% year-over-year increase in revenue for the first half of the year. Despite the revenue jump, MiniMax disclosed that its adjusted net loss more than doubled to $293 million during the period. MiniMax’s flagship M3 model ranked 18th on the same Artificial Analysis Intelligence Index referenced for Z.ai.
Z.ai and MiniMax both completed Hong Kong listings in January. Since their initial public offerings, Z.ai’s shares have soared by over 800%, while MiniMax’s stock has risen by more than 80%. Z.ai is scheduled to disclose its financial results for the first half of the year on Monday, an event investors will likely monitor for further clarity on adoption, margins, and capital expenditures, including any investments tied to domestic hardware adoption.
Analysts and industry watchers caution that while deployment on domestic chips is a meaningful milestone, transparency matters for validating such claims. Independent verification, detailed disclosures about hardware partners and performance benchmarks, and third-party audits would provide stronger support for Z.ai’s assertions. Nonetheless, the announcement reflects broader market and policy trends in which Chinese firms aim to reduce dependencies and demonstrate capabilities on both hardware and software fronts.
In summary, Z.ai’s release of GLM-5.3-Flash — and the company’s claim it runs on only China-made chips — is notable for its market impact, its signaling about China’s semiconductor progress, and the questions it raises about verification and the practical limits of domestically sourced AI infrastructure. The combination of increased usage rankings, a favorable movement in the company’s share price, and the broader strategic push toward self-sufficiency in semiconductors frames this as a development with both commercial and geopolitical dimensions.
Key Insights Table
| Aspect | Description |
|---|---|
| New Model | GLM-5.3-Flash — a lower-cost variant of Z.ai’s flagship model. |
| Domestic Chips Claim | Company asserts the model runs on 100,000 China-made chips; not independently verified. |
| Market Reaction | Z.ai shares rose over 8% in Hong Kong trading. |
| Usage and Rankings | Ranked first by recent usage on OpenRouter and 10th on the Artificial Analysis Intelligence Index. |
| Industry Context | Reflects China’s push for semiconductor self-reliance amid export restrictions on foreign chips. |
| Competitor Update | MiniMax reported strong revenue growth but larger losses; its M3 model ranks 18th on the index. |
Afterwards...
Looking ahead, Z.ai’s announcement will be monitored for corroborating evidence about the hardware used and for performance metrics that compare domestically deployed stacks to international alternatives. If verifiable, the deployment could accelerate collaboration among Chinese chipmakers, server integrators, and AI developers, reinforcing the country’s strategy to reduce foreign dependencies. Investors and industry participants will watch upcoming financial disclosures, third-party benchmarks, and potential partner disclosures to gauge how robust and scalable the domestic-chip-based deployments truly are. For global markets, continued advances in China’s AI hardware ecosystem could reshape competitive dynamics, prompting responses from international suppliers and potential changes in trade and technology policies.
