China's leading cloud providers are raising the prices of AI-related services, joining a global wave of increases as surging demand for artificial intelligence pushes up infrastructure costs and strains supply chains.

Alibaba Cloud said on Wednesday that it would increase prices for AI computing and storage products by as much as 34 percent, citing an "explosion" in global AI demand and rising supply chain costs.

The adjustments include its T-Head Zhenwu 810E computing cards, with price hikes ranging from 5 percent to 34 percent, and its CPFS intelligent computing storage product, which will rise by 30 percent.

On the same day, Baidu Cloud announced "structural optimization" of pricing across AI computing and storage offerings, pointing to rapidly growing demand for AI applications and sharply rising costs of core hardware and infrastructure.

The near-simultaneous moves underscored a broader industry shift: cloud providers worldwide are passing on higher costs for bandwidth, networking and compute capacity to customers, as AI workloads scale rapidly in size and complexity.

Data transfer and networking services have emerged as a key pressure point in the latest round of price increases, with hikes typically ranging from 10 percent to 40 percent.

Major providers including Amazon Web Services, Google Cloud, Microsoft Azure and Tencent Cloud have all adjusted pricing for network-related services, reflecting rising costs for bandwidth and underlying infrastructure.

The increases follow similar moves earlier this year.

Amazon Web Services and Google Cloud both raised prices on select offerings, with Google Cloud significantly increasing fees for data transfer services.

In China, Tencent Cloud also raised prices last Wednesday for its large model services. The company said it had adjusted billing for parts of its AI agent development platform to ensure stable, high-quality service.

Behind the price hikes lies a less visible but rapidly intensifying driver: an explosion in "token" usage, the unit by which AI models process and generate data.

People familiar with Alibaba Cloud's operations said its model-as-a-service platform recorded its fastest growth on record between January and March, as the company shifted scarce computing resources toward token-based inference workloads.

While the cost per token has declined over time, total token consumption is soaring as newer AI models perform more complex reasoning tasks. Applications such as deep research, AI agents and code generation require significantly more tokens per task, offsetting efficiency gains.

The rise of autonomous AI agents is accelerating this trend. Products like OpenClaw have rapidly gained traction among developers, marking a shift from simple chatbots to systems capable of executing long-running, multistep tasks.

Data from OpenRouter, a major AI model API aggregation platform, shows token consumption linked to OpenClaw surged from 80.6 billion on Feb 3 to 358 billion by early March — more than tripling in a month.

Weekly token usage on the platform reached 14.8 trillion in early March, up about 160 percent in two months, with OpenClaw accounting for the bulk of the increase.

Speaking at a recent conference, Nvidia CEO Jensen Huang said AI agents often require repeated calls across multiple models and tools to complete a single task, leading to "orders-of-magnitude increases" in token usage.

Analysts at Huatai Securities said the rapid rollout of agent-style applications would further accelerate demand for inference computing power, token consumption and infrastructure investment.

The growing popularity of such applications is tightening cloud resource availability. As token usage grows exponentially, demand for underlying compute and bandwidth is expected to surge in tandem, potentially driving further price increases or bringing forward additional hikes.

Sun Zhenya, a senior research manager at IDC China, said: "Facing exponential growth in token consumption, cost and energy use will become key constraints. Enterprises need to plan ahead in terms of computing resources and model strategies."