From a $56 Billion Arms Race to a Calculated Token Economy Strategy.

/

By He Lanyi Edited by Yang Xiaoruo

Produced by Business Show

Image from Alibaba's official website

Alibaba's AI strategy is undergoing redefinition.

Within just 23 days, the e-commerce giant has completed two high-intensity organizational adjustments, pointing to a core proposition: Alibaba's AI business is transitioning from costly new technology experimentation to systematic, refined operations.

This is not merely a land grab for technological positioning but rather a commercial cost-benefit analysis regarding investment and returns.

The capital market's response reflects investors' complex sentiment towards Alibaba's high-stakes gamble. As of noon on April 14, 2026, Alibaba's Hong Kong-listed stock price stood at HK$124.60, with a price-to-earnings ratio of approximately 22.8 times.

Recently buoyed by its AI strategy, Alibaba's stock price surged by 6.75% in a single day, showcasing capital's enthusiasm for the new narrative of "AI + e-commerce." However, over the past two months, its stock price has still experienced a correction of around 30% to 40%. This volatility indeed reflects the market's oscillation between AI narratives and profitability concerns.

On the one hand, Wu Yongming's goal of "surpassing $100 billion in cloud and AI commercialization revenue over the next five years" presents a highly imaginative vision. On the other hand, the planned $56 billion investment in cloud and AI infrastructure from 2025 to 2027, coupled with persistent actual losses, raises investor concerns about whether this could become an insatiable black hole.

Recently, Alibaba's video generation model, Happy Horse, topped the global blind testing rankings on Artificial Analysis. Additionally, on March 16, Alibaba announced the establishment of the ATH Business Group, with CEO Wu Yongming directly leading the charge. It is evident that Alibaba is shifting from a technological idealist arms race to a commercially pragmatic token economy track.

It's time for Alibaba to conduct a cost-benefit analysis of its AI endeavors.

01 Token Strategy

Alibaba's strategic pivot responds to shifts in the industry's underlying logic.

Historically, competition among AI large models focused on parameter scale and benchmark rankings, with token consumption regarded as pure IT costs. However, with the explosive application of AI agents, this logic appears to be undergoing a reversal.

As large models begin to demonstrate the capability to handle complex enterprise-level workflows, token invocation is no longer merely a simple API request but may become a production resource driving automated tasks.

This trend is also supported by data. According to Fortune China, during the Q3 FY2026 earnings call, Alibaba CEO Wu Yongming noted that from October to December 2025 (Q3 FY2026), token consumption in the public model service market on the Bailian MaaS platform increased sixfold. Tokens, the basic unit of AI information processing, are becoming a core metric for measuring AI value.

Alibaba has recognized this shift and incorporated it into its top-level strategy. On March 16, 2026, the company established the Alibaba Token Hub (ATH) Business Group, directly led by CEO Wu Yongming. This group consolidates core assets such as the Tongyi Lab, MaaS (Model as a Service), and the Qianwen Business Unit, forming a clear "AI productivity supply chain":

• Token Creation: Led by the Tongyi Lab, the goal is to develop the Qwen series of large models and enhance the "intelligence density" of each token.
• Token Delivery: Handled by the MaaS business line, model capabilities are opened up to enterprises and developers through the Bailian platform, with token consumption as the core performance indicator.
• Token Application: Managed by application-end units like Qianwen and Wukong, the aim is to convert token consumption into tangible commercial revenue.

These moves are also based on Alibaba's assessment of the second half of the AI race. The key competition has shifted from "whose model is more capable" to "who can more efficiently produce and consume tokens."

This is not merely a simple business integration but a calculated strategy centered around the "token economy," aiming to transform substantial early-stage technological investments into quantifiable commercial outcomes through a full-stack layout of "chips + cloud + models + applications."

02 Organizational Restructuring

To support this commercial transformation, Alibaba has undergone a significant organizational overhaul.

On April 8, Wu Yongming announced the establishment of the Group Technology Committee and made key personnel appointments. This was not a simple rotation of positions but a "separation of powers" based on rigorous causal logic. The backgrounds and divisions of responsibility among the three technical leaders clearly reveal the underlying logic of Alibaba's AI strategy.

Zhou Jingren is a veteran of Alibaba's AI strategy. He previously led iDST, the precursor to the DAMO Academy, and later took charge of the Tongyi Lab. As a key operator in Alibaba's journey from a follower to a leader in large models, he stepped down as Alibaba Cloud's CTO to become the Chief AI Architect of the Group Technology Committee, fully responsible for the upgraded Tongyi Large Model Business Unit.

The rationale behind this adjustment is to elevate model R&D from the complexities of cloud business operations to the highest strategic level within the group. After incidents such as the departure of Lin Junyang, Alibaba aims to clarify that models are core assets in the AI era, and Zhou Jingren, as the chief designer of Alibaba's AI "brain," must focus on enhancing model value.

Li Feifei is Alibaba's Group Vice President and a top expert in the database field. Public reports indicate he holds academic backgrounds from Tsinghua University and Boston University, representing a typical academic technical leader. He led the development of core database products such as PolarDB.

His appointment as Alibaba Cloud's CTO, responsible for AI cloud infrastructure construction, signifies that Alibaba's future AI competition will hinge not only on model parameters but also on computational efficiency and cost control.

Wu Zeming is a core member of Alibaba's e-commerce system technical architecture, having risen through the ranks with extensive practical business experience. He will focus on the Group CTO role, responsible for the business technology platform and AI inference platform. His mission is clear: to integrate AI capabilities into all business scenarios, including Taobao, Tmall, and DingTalk.

If Zhou Jingren forges the "weapons," Li Feifei maintains the "arsenal," then Wu Zeming is the one responsible for "training the troops." He must ensure these weapons can be effectively utilized in real commercial battlefields, achieving transformation from technology to revenue.

This structural adjustment also breaks down the "silos" that previously existed between large models, cloud infrastructure, and business applications.

In the past, model teams focused on academic rankings, cloud teams on revenue scale, and business teams on daily active users. Now, through the coordination of the ATH Business Group, all teams are aligned under the unified metric of "tokens." Models must be smarter, clouds more efficient, and applications more profitable.

03 Growing Pains and Concerns

Historically, every significant organizational transformation at Alibaba has been accompanied by growing pains.

On the eve of ATH's establishment, the departure of Lin Junyang, a core figure behind the Qwen model, sent shockwaves through the industry. Although Alibaba quickly quelled doubts through internal communications and executive appearances, this incident undoubtedly sounded an alarm for Alibaba's AI talent strategy.

In the AI field, which heavily relies on top-tier talent, balancing the personal aspirations of leading scientists with the company's commercial objectives under a collectivist operational model remains a major challenge for Alibaba.

While a unified technical framework and performance metrics (tokens) enhance efficiency, they may also stifle disruptive innovation. If all resources are tilt towards "profitable" projects, will early-stage, high-risk foundational research be marginalized?

A more pressing and severe risk lies in the fact that AI's grand vision must be built on a stable and reliable infrastructure.

Reviewing Alibaba Cloud's recent performance in security and stability, a series of frequent incidents constitute its "historical debt" that must be addressed.

In May 2025, the data breach incident involving Dior China's customers, reported by the Ministry of Public Security's Cybersecurity Bureau, was linked to improper configuration of the Alibaba Cloud platform used by Dior, exposing the complexity of cloud data security.

At the physical infrastructure level, hidden danger are equally concerning. On December 9, 2024, a fire broke out at the under-construction Alibaba Cloud campus in Heyuan, Guangdong (non-production data center). Although there were no casualties and cloud services were unaffected, the incident tested cloud providers' physical safety management.

Looking further back, according to media reports such as The Paper, the "Hong Kong outage" on December 18, 2022, marked a dark moment in Alibaba Cloud's history. A cooling failure triggered fire sprinklers, causing a service disruption lasting over 15 hours. This incident directly led to the resignation of then-president Zhang Jianfeng, with Daniel Zhang stepping in to manage the crisis.

Coupled with the earlier six-month suspension as a cooperation partner by the Ministry of Industry and Information Technology for failing to promptly report the Log4j2 vulnerability, this series of incidents—spanning from the application layer to the physical layer, from compliance to operations—clearly demonstrates that when AI business becomes the core growth engine of the group, any failure in underlying cloud services could disrupt MaaS services and paralyze AI applications, with far-reaching chain reactions and commercial losses exceeding past experiences.

This adjustment also means that the new Alibaba Cloud CTO, Li Feifei, must not only address computational efficiency challenges but also shoulder the pressure of repairing technical debt and ensuring the uninterrupted supply of AI "utilities."

Drawing from the experience of the "Big Middle Platform" era, overly centralized organizational structures often face the risk of rigidity. How to keep technical leaders like Zhou Jingren and Li Feifei agile in their respective domains and avoid getting bogged down in corporate bureaucracy is a question Wu Yongming must consider.

Q3 FY2026 financial results show that Alibaba's Non-GAAP net profit was RMB 16.71 billion, a year-on-year decline of 67%.

After its rapid expansion, Alibaba is slowing down to conduct a cost-benefit analysis. Following the analysis, adjustments must be made, painfully divesting inefficient businesses and reallocating every saved penny towards AI and global expansion. Wu Yongming has set a goal of surpassing $100 billion in cloud and AI commercialization revenue over the next five years, which is undoubtedly Alibaba's bet on the future.

However, the harsh reality is that in this token-based marathon, whether Alibaba can maintain its technological moat while achieving a closed-loop commercialization remains to be seen over time.「End」