Alibaba shakeup signals shift toward token-based AI business model

Summary

• Alibaba restructures to separate core AI development from its cloud infrastructure division.
• The surge in autonomous AI agents is driving a massive increase in token consumption and new revenue streams.
• Launch of “Wukong” platform highlights the move toward multi-agent coordination for enterprise workflows.

HONG KONG, March 17, 2026 — Alibaba Group is overhauling its artificial intelligence operations, pivoting toward a token-driven business model to better monetize the rapidly expanding AI agent economy.

The $288 billion technology giant announced a structural separation of its specialized AI units from its cloud computing arm. This strategic pivot moves away from traditional full-stack enterprise bundling, which has faced headwinds as Chinese corporations tighten IT budgets. By decoupling these units, Alibaba aims to provide more granular, usage-based pricing that reflects how modern AI is consumed.

Alibaba’s cloud division is currently on track to generate 152.8 billion yuan ($21.2 billion) in revenue for the fiscal year ending March 2026. While the division remains a market leader, its growth is increasingly dependent on high-volume model inference rather than standard storage and compute.

Focus shifts to AI agents and token usage

The center of this restructuring is a new entity, the Alibaba Token Hub Business Group. Under the direct leadership of CEO Eddie Wu, the group will consolidate the company’s disparate AI initiatives into a single pipeline focused on “tokenomics”—the pricing and distribution of raw model outputs.

The shift is a direct response to the rise of AI agents. Unlike the chatbots of 2024, today’s agents operate autonomously to handle complex logistics, coding, and scheduling. This continuous operation consumes tokens at a rate roughly 40 to 60 times higher than manual user queries, fundamentally changing the unit economics of the industry.

This demand has been bolstered by the “OpenClaw” phenomenon. As an open-source framework for agentic workflows, OpenClaw has forced major players like Tencent and Alibaba to accelerate their own proprietary alternatives to capture the high-frequency usage associated with autonomous software.

Competitive landscape in China

The move also highlights a divergence in how AI is monetized in China versus the West. While U.S. firms have largely leaned on high-margin SaaS subscriptions, Chinese users have shown a preference for usage-based “pay-as-you-go” models.

Startups like MiniMax are already validating this path, with estimated annual recurring revenue hitting $150 million in February. Alibaba’s new Wukong platform, launched today, follows this trend. Wukong allows businesses to deploy multiple “agent swarms” that coordinate document editing and research in real-time, billed entirely on the volume of tokens processed.

Alibaba’s Hong Kong-listed shares reacted positively to the news, rising 2.7% to HK$137.80 in midday trading as investors cheered the clearer path to AI profitability.

Why this matters

• Monetization Pivot: Marks the transition from fixed software fees to dynamic, usage-based “token” billing.
• Agent Economy: Recognizes that autonomous agents, not human chatters, are now the primary consumers of LLM power.
• Cost Leadership: Positions Alibaba to leverage China’s domestic chip advancements and lower energy costs to undercut global token prices.
• Operational Agility: Separating AI from Cloud allows the AI units to move at the speed of the startup ecosystem rather than the slower pace of infrastructure cycles.

FAQs

Q1: Why is Alibaba separating its AI and cloud businesses? 

The move allows the AI division to focus on high-frequency token sales and agent development, independent of the capital-intensive infrastructure requirements of the cloud business.

Q2: What exactly is the “Wukong” platform? 

Launched on March 17, Wukong is an enterprise interface that coordinates different specialized AI agents to work together on a single project, such as a full marketing campaign or a financial audit.

Q3: How does “token-based” pricing work? 

Instead of a monthly flat fee, businesses pay only for the volume of text, code, or data the AI generates or processes, similar to a utility bill for electricity.

Q4: Is Alibaba’s stock recovering because of this? 

Market sentiment suggests that investors are responding to the clear “AI-first” organizational structure, which clarifies how the company will compete with OpenAI and domestic rivals like Tencent.