"Lobster" frenzy, "Token" explosion—can they hype up the Chinese BAT giants?
```
Alibaba Cloud's maximum price increase of 34% is just the tip of the iceberg. A revolution in Token consumption triggered by the open source AI agent framework OpenClaw is reshaping China's AI industry at a speed far beyond market expectations.
At noon on March 18, Alibaba Cloud officially announced on its website that due to the global explosion in AI demand and supply chain price hikes, its AI computing power and storage products would see price increases of up to 34%. Among them, computing power card products such as Pingtouge Zhenwu 810E increased by 5%-34%, and file storage product CPFS (Intelligent Computing Edition) increased by 30%.
The official explanation is "global AI demand surge and supply chain price increase," but according to insiders, the deeper driving force is — a surge in Token calls. Alibaba Cloud MaaS business Bailian achieved its highest growth rate in history from January to March 2026, with computing resources being prioritized for Token-related businesses.
The starting point of this Token storm is an open source AI agent framework affectionately called "Little Lobster" by Chinese netizens — OpenClaw. As of March 15, 2026, it has garnered over 315,000 stars on GitHub and ignited an unprecedented collective "ChinaClaw" battle on the Chinese internet.
UBS and Barclays both released special research reports on the 17th, pointing to the same core conclusion: The shortage of computing power is intensifying, cloud service demand is surging, and the business model for enterprise AI SaaS is emerging. Core Chinese tech companies such as Tencent, Alibaba, and Baidu will be the most direct beneficiaries.
"Little Lobster" Sweeps China: From Developer Toy to National AI Infrastructure
The disruptive aspect of OpenClaw lies in its architectural logic—it is not a chatbot answering questions in the cloud, but an "AI executor" stationed in instant messaging software, able to directly control computers to perform real-world tasks.
Its ecosystem is anchored by the ClawHub skill platform, dubbed the "App Store for AI Agents," which has accumulated more than 20,000 skills, covering scenarios such as file management, browser operations, email handling, code generation, and more.
What's more crucial, a single OpenClaw can generate hundreds or even thousands of model calls per day, with Token consumption at multiples of traditional AI chat applications.
This feature has directly fueled the rapid expansion of the global AI model API market: According to data from the API aggregation platform OpenRouter, the weekly Token usage recently reached about 16 trillion, roughly three times the level of January 2026 (before OpenClaw was launched).
Compared to the US and European markets, the spread of "Little Lobster" in China has taken on a completely different form. While Western markets are largely confined to small circles of technical developers (with complicated setups requiring Docker and API key management), Chinese tech giants rapidly transformed it into a "one-click deployment" cloud service, rolling it out en masse to both enterprise and consumer markets.
ChinaClaw Blooms: Giants Bet Collectively, SaaS Monetization Breaks Through
From late February to March 2026, major Chinese internet and AI model companies intensively launched their own "Claw" products.
Before its launch, Tencent's WorkBuddy had been used internally by more than 2,000 employees, mainly covering data analysis and office automation scenarios for HR, administration, and operations. All products support invocation via popular enterprise messaging platforms like WeChat Work, DingTalk, and Feishu, directly integrating into enterprise workflows.
Token Explosion: Data Reveals Global AI Demand Inflection Point
The popularization of OpenClaw is triggering an unprecedented revolution in Token consumption.
The UBS report notes that global weekly Token usage for AI models, monitored via the API aggregation platform OpenRouter, recently reached about 16 trillion, nearly three times the level in January 2026 (before OpenClaw went online).
The logic behind this explosive growth is clear: compared to ordinary AI chat tools, Token consumption by Claw agents is an order of magnitude higher—a single OpenClaw can generate hundreds to thousands of model calls per day, far exceeding standard chat scenarios.
Chinese models are the biggest beneficiaries of this Token feast. According to OpenRouter data for the past month as of March 9, 2026:
MiniMax M2.5: Token usage reached 7.5 trillion, No. 1 globally;
Kimi K2.5: 4.2 trillion, No. 2;
DeepSeek V3.2: 3.3 trillion;
GLM-5: 2.2 trillion;
According to vendor data, the growth is also eye-catching: MiniMax disclosed that in February 2026, the daily Token consumption of its M2 series text models soared more than six-fold compared to December 2025, with coding plan Token consumption surging over ten-fold. The MiniMax M2.5 was released on February 11, 2026, and rose to the top of the OpenRouter popularity rankings within 12 hours; in the first week after launch, weekly Token usage shot up to 3 trillion, surpassing the combined numbers of Kimi K2.5, GLM-5, and DeepSeek V3.2.
The data from Moonshot (Kimi) is even more stunning: Kimi K2.5, released at the end of January 2026, had revenue in its first 20 days exceeding the entire year of 2025, and overseas revenue surpassed domestic sales. On this basis, OpenClaw also designated Kimi K2.5 as its first official free core model.
China Models' Cost "Nuclear Weapon": Pricing Less Than 10% of US Counterparts
With explosive Token consumption and China's AI models priced far below US equivalents, the logic guiding global developers’ choice of models is being fundamentally changed.
According to Barclays research (as of March 9, 2026), mainstream Chinese AI models are generally priced below 20% of OpenAI's GPT-5.4, and most are below 10%:
Taking MiniMax M2.5 as an example, running at an output rate of 100 Tokens per second for an hour costs only about $1, which is 1/10 to 1/20 the equivalent cost of Claude Sonnet. This price advantage means that, even if users initially built their Claw agent with Claude or GPT, when US cutting-edge model costs become untenable, they can easily switch to Chinese large models to control costs.
It is noteworthy that, despite steep price reductions, the gross margins for inference at Chinese AI companies continue to improve.
MiniMax disclosed that its per-Token inference cost dropped by more than 50% from December 2025 to February 2026, with pricing unchanged during the period. UBS research shows that Zhichu and MiniMax’s API and enterprise AI deployment gross margins are about 60%-70%, mainly from inference business.
From "Free" to "Subscription": China’s SaaS Awakening Moment
What excites the capital market most about ChinaClaw may not be just the surge in Token consumption, but the fact that it is opening the historic door for SaaS monetization in China.
For a long time, Chinese consumer AI chat products have been virtually all free. The launch of ChinaClaw introduced subscription pricing to ordinary users and the enterprise market—monthly fees range from 40 to 500 RMB, with additional pay-as-you-go points packages.
Barclays calls this the "China SaaS moment"—China has never had a true SaaS industry, but thanks to AI and the ChinaClaw wave, it may leap directly into a new era of productivity subscription services driven by AI.
Computing Power Shortage Worsens: Who Are the Most Certain Beneficiaries?
The contradiction between supply and demand for computing power is deteriorating rapidly under the impact of the ChinaClaw wave.
Barclays channel research shows that Tencent and Baidu are both considering raising Token prices, while Alibaba Cloud's price hike has already taken effect. The worsening shortage of computing power means larger capital expenditures (Capex) are inevitable.
Synthesizing UBS and Barclays' analysis, the "Little Lobster" craze affects China’s AI value chain at three levels:
Infrastructure Layer: Most Certain Beneficiaries
Alibaba: China’s largest cloud provider, with full-stack AI capabilities. As computing power shortages suddenly intensify, Barclays expects Alibaba to further increase Capex during earnings releases.
Kingsoft Cloud: AI workloads account for over 30% of total revenue, and the company has established deep partnerships with several top AI labs; UBS rates it a Buy.
Large Model Layer: Direct Token Monetization
Chinese AI labs (MiniMax, Moonshot, Zhichu, DeepSeek, etc.) are direct beneficiaries of the Token surge. Their cost advantages are attracting global developers to switch to Chinese models, especially as ChinaClaw spreads, with single-task Token consumption growing exponentially; data has validated the logic of high-speed growth in model API revenue.
Application Platform Layer: Most Anticipated SaaS Monetization Path
Tencent: The combination of Tencent QClaw (personal version) + WorkBuddy (enterprise version) covers China’s most widely used instant messaging entry points—WeChat, QQ, and WeChat Work, with AI agent development and execution capabilities continually improving.
Baidu: Baidu Search ranks No. 1 on the ClawHub skill download list, with over 36,000 downloads. Strong native search skill demand is driving active renewal of cloud service contracts.
Barclays concludes that the shortage of AI computing power has been sharply aggravated by the rise of Claw—not only because Claw is a brand new AI usage scenario, but also because Token consumption in this scenario is an order of magnitude higher than ordinary chat. Once ChinaClaw secures a foothold in the Chinese enterprise market, it will open a new, sizeable TAM. More importantly, this monetization opportunity is not necessarily limited to China—the significant cost advantage of Chinese AI models is expected to help Chinese tech companies compete effectively overseas where China AI models are welcomed.
The above exciting content is from Zhifeng Trading Desk.
For more detailed interpretation, including real-time analysis and frontline research, please join the 【Zhifeng Trading Desk▪Annual Membership】
Risk Disclaimer and Exemption ClauseThe market carries risks, and investment should be made with caution. This article does not constitute personal investment advice and does not take into account individual users' specific investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article fit their specific circumstances. Investments made based on this content are at your own risk. ```