Computing power is tight, and cloud providers are collectively raising prices.

When it comes to stock trading, rely on Golden Kylin Analyst Reports—authoritative, professional, timely, comprehensive—helping you uncover potential thematic opportunities!

Source: Beijing Business Today

The token frenzy sparked by “Lobster” (OpenClaw) is pushing cloud providers into an awkward situation: the more users use, the more they lose. On March 18, Alibaba Cloud and Baidu Smart Cloud announced they would no longer silently absorb costs, and on the same day, they announced price hikes for AI computing power, storage products, and services, with increases up to 34%. In fact, Tencent Cloud was the first to cut losses, having already raised prices for two of its self-developed models a week earlier, with increases exceeding 460%. Whether it’s raising the “rent” for computing and storage resources or directly cutting into model API endpoints, combined with the upcoming storage chip price increase cycle in late 2025, cloud providers are bearing significant cost pressures, with chain reactions continuing to unfold. AI Agents turning API calls into computing power consumption—since computing power is fundamental and tokens are commodities—the cloud bills triggered by “Lobster” are now coming due.

Two Major Cloud Providers Raise Prices

On March 18, Alibaba Cloud and Baidu Smart Cloud announced price increases on the same day. According to official notices, Alibaba Cloud will adjust prices for services related to the Pengniu Zhenwu 810E computing card and CPFS (Intelligent Computing Edition), with increases up to 34%. Baidu Smart Cloud explicitly stated that AI computing power-related products and services will see price hikes of about 5%–30%, and parallel file storage prices will increase by approximately 30%. Both companies attribute the price hikes to “the explosion of global AI demand.”

In this wave of price increases, Tencent Cloud acted the fastest. A week earlier, Tencent Cloud announced a price increase for two of its self-developed models, Tencent HY2.0 Instruct and Tencent HY2.0 Think. For example, before the adjustment, the input price for Tencent HY2.0 Instruct was 0.0008 yuan per thousand tokens; after the increase, it rose to 0.004505 yuan per thousand tokens, a 463% jump. At the same time, Tencent Cloud ended free public testing for three models—GLM 5, MiniMax 2.5, and Kimi 2.5—and transitioned to official commercial services.

“This price increase doesn’t affect us,” a technical manager at a small website told Beijing Business Today. “We use Baidu Smart Cloud, but we haven’t purchased AI computing power or parallel file storage.”

According to the official website, Baidu Smart Cloud’s parallel file storage service PFS is a fully managed, simple, scalable parallel file storage system that offers sub-millisecond access and high IOPS (Input/Output Operations Per Second) for high-performance computing scenarios. Use cases include AI training and inference, autonomous driving, high-performance computing, and video rendering.

For long-term clients, the price increases from Baidu and Alibaba will not affect them immediately. The new prices will take effect from 00:00 on April 18, 2026 (Beijing time). Customers who have already purchased related products or services before this date will not see changes during their current billing cycle; the new prices will apply in subsequent renewal periods.

“Cost Pass-Through”

On the same day Alibaba Cloud and Baidu Smart Cloud announced price hikes, Tencent released its Q4 and full-year 2025 financial results. Tencent Chairman and CEO Ma Huateng announced that Tencent Cloud would achieve scaled profitability by 2025.

During the earnings call, Tencent executives responded to rising storage chip prices, stating that the surge in AI demand has not only boosted DRAM and high-bandwidth memory (HBM) demand but also led to a full rebound in CPU, SSD, HDD, and other component demands. Orders now need to be booked months, quarters, or even years in advance. Suppliers prioritize their largest and most stable customers, such as Tencent Cloud. Smaller cloud providers now find it difficult to secure stable supply chains.

In this context, Tencent executives believe that the industry has no choice but to pass increased costs onto prices.

In an interview with Beijing Business Today, Luo Guozhao, Director of CHIP China Laboratory, also straightforwardly said, “Price hikes are inevitable and spreading.”

“Memory modules (DRAM), SSD NAND/HDD, and mechanical hard drives have all doubled or more in price. These are the main expenses for cloud service equipment. Many have the misconception that AI accelerators or GPUs are expensive; in fact, these high-priced products are used in limited quantities. For example, a CPU with 6–8 memory channels, even at the smallest configuration, is more expensive than a CPU alone, even without price increases,” Luo explained in detail about the cloud computing supply chain.

Regarding AI computing power, he further explained, “The widespread deployment of AI servers and computing centers has intensified the chip shortages, especially for memory and SSDs. Over the past two years, AI-related products have gained significant attention from ordinary consumers, which psychologically amplifies demand and drives prices up.”

From Tokens to Computing Power

Supply chain price hikes are only one aspect of the cost pressures faced by cloud providers. The visible trigger is the explosion in Token consumption caused by AI Agents, with OpenClaw (user nickname “Lobster”) being the catalyst.

Unlike traditional AI interactions that consume only a small number of Tokens per session, “Lobster” autonomously executes tasks and continuously calls tools, with each step involving large Token consumption.

“Prices will definitely become cheaper over time,” said Fu Sheng, Chairman and CEO of Cheetah Mobile, when discussing future costs of “raising Lobsters.” He explained, “Currently, the main bottlenecks are electricity and chips, which take time to resolve. Software architecture can also be optimized. Why was DeepSeek so popular last year (2025)? Because it fundamentally improved architectural efficiency. As everyone consumes Tokens, they’ll realize some processes don’t need such complex structures, which can lower costs. Overall, costs can be reduced in energy, chips, and algorithms.”

This conflicts with the current price hikes by cloud providers. Luo Guozhao told Beijing Business Today that this contradiction is essentially a shift in business models and market behavior. “Before ‘Lobster’ appeared, individual users paid very little for Tokens. Usage was low and stable, and free use was acceptable for cloud providers. But as ‘Lobster’ consumes massive amounts of Tokens continuously, it can no longer be free. Cloud providers can leverage user demand to adjust pricing strategies.”

In fact, this price increase was already set in motion two months ago. On January 22, AWS announced a 15% price increase for EC2 instances used for large model training. On January 27, Google Cloud announced price hikes for data transfer services, AI, and computing infrastructure, with increases up to 100%.

Beijing Business Today reporter Wei Wei