Zhipu AI stock slides as compute shortages stall global expansion

Chinese AI company Zhipu AI saw its shares drop nearly 23% on February 23, erasing over HK$70 billion (US$9 billion) in market value amid concerns over its computing resources.

The drop followed the company’s public appeal for global partnerships with inference-compute providers, as user complaints about response delays and rate limits persisted despite recent investor interest.

Zhipu, known as Z.ai internationally, issued its first public call for support from domestic and international GPU providers to help power its GLM models.

Since its IPO in early January, Zhipu has struggled to secure additional computing resources needed for international growth, especially after being blacklisted by Washington earlier this year, which limited access to US chips.

🔗 Source: South China Morning Post

🧠 Food for thought

Implications, context, and why it matters.

Zhipu’s stock plunge follows a breakout moment in the U.S.

Zhipu’s shortage of computing power appears linked to rising use of its AI coding tools, including from U.S. users 1.
That U.S. uptake led Zhipu to cap access for new sign-ups to its GLM coding plan ahead of its public request for more resources 1.
Shares fell nearly 23% after an earlier spike, when the stock rose as much as 34% following the launch of GLM-5 2.
After releasing GLM-5, Zhipu raised prices for its GLM Coding Plan by 30% 2.

Zhipu’s strain offers a read on China’s AI hardware supply chain

Zhipu has been a visible example of China’s domestic hardware push. It said its multimodal GLM-Image was trained on Huawei’s Ascend stack. The stack includes Ascend AI processors (chips designed to train and run AI models) plus Huawei’s MindSpore framework (software used to build and train AI models) 3.
That claim suggests domestic chips can handle training for advanced models. Zhipu’s inference-service problems match user complaints about slow replies and rate limits, which raises questions about enough compute at scale.
Sources say Huawei’s advanced Ascend chips have shipped in small volumes because a yield-constrained manufacturing process means too many chips miss quality targets 4.
Zhipu’s public search for compute partners puts China’s domestic supply chain under pressure to support a fast-growing AI company, which ties to its push for tech self-reliance 4.