Alibaba.com says its Accio e-commerce agent has passed US$60 million in annual recurring revenue after five months on the market. The business-to-business platform also says Accio has more than 60,000 paid users whose average spending exceeds US$1,000. The figures were presented by Alibaba.com president Zhang Kuo at the company’s CoCreate conference in Los Angeles, according to Nikkei Asia.
The disclosure is a useful data point in a market where Chinese AI firms are under pressure to show that agents generate revenue rather than only attention. Accio is designed for e-commerce tasks such as sourcing and shipping, putting it in a narrower category than a general-purpose coding or chat agent. Alibaba’s pitch is that specialization can lower costs and create enough practical value for small and medium-sized businesses to pay.
The company’s numbers are self-reported and not independently audited. They nonetheless show how Alibaba wants to measure agent commercialization: paid users, annual recurring revenue, and cost per completed task. That differs from the model race, where parameter counts and benchmark rankings often dominate.
Accio’s Revenue Claim Comes Five Months After Launch
Alibaba.com launched Accio in April 2026. At the September conference, Zhang said the platform had surpassed 60,000 paying users and US$60 million in annual recurring revenue. The company’s stated average spending above US$1,000 is consistent with those figures at a broad level, although Alibaba did not provide audited financial statements or a detailed definition of which customer contracts count toward recurring revenue.
Accio targets e-commerce workflows including product sourcing and shipping. Those tasks are attractive for AI agents because they involve repeated searches, supplier comparisons, documentation, and coordination steps. A specialized system may be able to use structured commerce data and domain-specific tools more efficiently than a broad agent designed for many unrelated tasks.
The revenue figure is particularly important because it gives investors a rare company-reported measure of willingness to pay for a Chinese AI agent. Many agent products are still promoted through demos, free tiers, or enterprise pilots. A meaningful base of paid users, if sustained, would suggest that businesses see enough value in automation to budget for it.
Alibaba has already put enterprise agents at the center of its strategy. Its recent QoderWake update focuses on software workers inside Chinese workplace applications. Accio addresses a different market: international merchants and small businesses using Alibaba.com for commerce. Together, the products show Alibaba trying to turn agents into task-specific services rather than a single all-purpose assistant.
Commerce Agent Bench Anchors Alibaba’s Cost Argument
Alibaba also released a cost comparison based on its Commerce Agent Bench. The company says the open-source framework contains 107 end-to-end tasks across seven categories and four levels of autonomy. Alibaba says it was derived from 1.6 million merchant conversations and 200,000 execution traces.
In Alibaba’s evaluation, Accio completed the task set at an estimated token cost of US$3.69. The company put OpenAI’s Codex at US$9.27 and Anthropic’s Claude Code at US$9.51. That is the basis for Alibaba’s assertion that Accio can be more than 50 percent cheaper for commerce-related tasks.
The comparison should be treated with care. Alibaba developed the benchmark using its own merchant data and reported its own results. The systems compared are not necessarily optimized for the same workflow, model choice, tool configuration, or cost structure. Token cost also captures only one part of commercial value. Accuracy, completion rate, operational safety, and the cost of human review can change the economics of an agent deployment.
Still, Alibaba’s decision to publish a commerce-specific benchmark signals a serious competitive move. General-purpose models are flexible, but specialized agents can be tuned around repeated business processes and structured datasets. If the benchmark gains outside use, it may become a way for merchants and developers to compare agent performance on tasks that resemble actual commerce operations.
Agent Economics Move From Demos to Retention
The key question for Accio is whether early paid usage turns into durable retention. An annual recurring revenue figure is meaningful only if customers continue to pay, use the tool regularly, and expand it into more workflows. Alibaba did not disclose churn, renewal rates, customer geography, or how much revenue comes from the largest users.
Those missing details are normal for an early disclosure, but they will matter. A small business may find an agent useful for a limited sourcing project without adopting it as a core operating system. Conversely, a specialized agent that becomes part of supplier discovery, negotiation, compliance, and shipping could develop high switching costs.
Alibaba’s international orientation also differentiates Accio from agents aimed only at China’s domestic office market. It also broadens Alibaba’s commercial AI story beyond the AI compute price competition among Chinese cloud providers. Alibaba.com serves cross-border merchants, so the product can be evaluated through the practical work of global trade: finding suppliers, comparing offers, and managing logistics. That creates a broad potential market, but it also requires the agent to cope with multiple languages, rules, and commercial norms.
EastFrontier has previously covered Alibaba’s effort to build AI services into its cloud and business ecosystem. Accio’s reported US$60 million revenue claim adds a concrete, though unverified, commercial milestone to that strategy. The next evidence will be external adoption data, customer case studies, and whether competitors can match its claimed cost advantage on the same tasks.
For now, the significant development is not that Alibaba has declared an agent cheaper. It is that the company is asking the market to evaluate an AI product through paid users and recurring revenue. That is a more demanding standard than a launch event, and Accio’s future will depend on whether the figures continue to rise after the first five months.