Chinese AI stocks pay for the model price war

On September 10, shares in two Chinese model companies on the Hong Kong exchange fell sharply after DeepSeek cut its prices. Zhipu (02513.HK) closed the day at 819 Hong Kong dollars, down 10.34 percent. MiniMax (00100.HK) dropped 8.98 percent to 292 dollars. Over the month, losses reached 31.46 percent for Zhipu and 16.3 percent for MiniMax.

The direct cause is simple: a cheaper rival model takes customers. DeepSeek cut prices on its Flash series that same day. Morgan Stanley analysts noted that other competitive models, including Alibaba's Qwen3.8-Flash, also added pressure. Their conclusion is blunt. When service quality and availability are similar, customers choose the model with the lower total cost. This hits retention, pricing, and margins for the remaining suppliers.

The financial statements show the scale of the issue. In the first half, Zhipu generated 954 million yuan in revenue, up 399.7 percent year-on-year and more than in the entire previous year. Cloud deployments accounted for 86.5 percent of revenue. The margin on the open platform and API improved from minus 0.4 percent to 24.6 percent. The problem is that the cloud business model carries far lower margins than on-premise deployments. The company's overall margin fell from 50 percent to 26.4 percent. Research and development spending hit 2.13 billion yuan, while the adjusted net loss stood at 1.96 billion yuan.

MiniMax also relies increasingly on serving model calls. First-half revenue came to about 117 million dollars. The open platform and enterprise services made up 73.9 million dollars of that, with their share rising from 30.3 percent to 63.4 percent. This is the segment where a competitor's price cut hurts most.

Both companies are trying to offset this with efficiency. Zhipu reported that since the start of the year, the unit cost of inference per token fell by 80 percent. The number of tokens handled rose more than 40-fold, and the average API price increased by about 101 percent. MiniMax attributes its margin improvement to infrastructure upgrades. Jefferies, however, lowered its valuation of Zhipu's cloud business, cutting the revenue multiple from 50 to 30 times.

A broader question hangs over the entire Chinese market. Can independent model labs hold their margins when competing not just with DeepSeek, but with conglomerates that subsidize the effort with their own cloud, ecosystem, and sales channels?

The half-year results show where value is created. At Zhipu, most revenue comes from cloud deployments, and the open platform has only just moved into positive gross margin after running below the line. MiniMax has shifted its focus from consumer apps to enterprise services. Both face a similar problem. Revenue is growing, but so are research and infrastructure costs. A competitor backed by a conglomerate does not need to make money on the model itself. In this setup, margin is the last thing to stabilize.