Tokens: The New Focal Point

Alibaba has unveiled the Alibaba Token Hub (ATH) Business Group today, a pioneering entity dedicated to the core missions of 'token creation, token distribution, and token application'.

As reported by Huxiu, the acronym ATH also evokes the financial term 'All Time High', symbolizing an anticipated peak in stock price history.

The ATH Business Group encompasses Tongyi Labs for foundational model research, the MaaS business unit, the consumer-facing AI assistant Qianwen, the business-oriented AI work platform Wukong, and a dedicated team exploring innovative applications.

Jack Ma remarked at Yungu School earlier this year: 'The influence of AI surpasses our imagination, and we are all underprepared.'

This sentiment rings equally true for Alibaba.

Over the past two years, Alibaba Cloud has been weaving AI narratives into its financial reports, with AI-related product revenues soaring by triple digits for nine consecutive quarters.

However, beneath these narratives, the organizational structure around AI remained fragmented. Models resided in Tongyi Labs, platforms in Bailian MaaS, consumer applications in Qianwen, and business solutions in DingTalk Wukong, each functioning as independent business units with their own KPIs.

Now, with Wu Yongming overseeing the entire token ecosystem, two critical issues—resource allocation and coordination—have been addressed.

Lin Junyang's departure was not without its controversies, with dissatisfaction at Tongyi Labs primarily stemming from limited computing resources.

This was partly due to the previous lack of focus from group leadership on this crucial aspect. Additionally, computing power is a scarce commodity, even for a tech giant like Alibaba.

The distribution of computing power between R&D and applications required negotiation, previously managed by Lin Junyang and Zhou Jingren, who balanced and coordinated across different business lines.

With Wu Yongming now directly responsible for the new business group, the previously fragmented resource coordination has been elevated to a unified, group-level scheduling issue.

Lin's departure indirectly highlighted issues such as inadequate computing resources. However, his personal disagreements with the group's strategic direction, particularly the need for closer collaboration between foundational models and products, left little room for compromise.

Alibaba's push for stronger collaboration between product development and research is undoubtedly sound. Yet, as long as the Qianwen team and Tongyi Labs remain separate entities, such collaboration, even if mandated from the top, will face significant hurdles.

While Alibaba's internal jargon may sometimes invite ridicule, misalignment indeed causes substantial organizational friction.

This issue is not confined to the clash between Qianwen and Tongyi's foundational model team. Qianwen, having recently garnered the most group resources and being a key growth project, has seen its conflicts surface and escalate earlier.

Business-end issues have not yet been fully exposed, but it is likely only a matter of time.

Compared to consumer applications, enterprise scenarios involve longer chains and more intricate steps, demanding higher coordination among models, platforms, and products.

The current friction between Qianwen and Tongyi may well recur with DingTalk Wukong if the organizational structure remains unchanged.

Therefore, integrating Wukong into ATH is a preemptive move. Business-end AI products ultimately need to seamlessly integrate into enterprise workflows, requiring not just model capabilities or product entry points but also platforms, toolchains, and delivery capabilities.

Early integration of these elements can at least mitigate predictable internal friction and disputes.

ATH: Drawing the Boundaries

Organizational adjustments are never solely about efficiency.

Efficiency is one aspect; power realignment is another. Any significant organizational change in a large company inevitably involves resource regrouping, responsibility redefinition, and influence redistribution.

The same holds true for Alibaba this time.

With ATH singled out and directly overseen by Wu Yongming, it is set to become Alibaba's AI strategic hub.

Who is included, who is excluded, and who reports to whom—these changes will ultimately crystallize into a new power landscape.

Let's briefly discuss three directly impacted business lines.

My initial reaction upon seeing the Qianwen Business Unit being incorporated into the new business group was: what becomes of the newly established Qianwen C-end Business Group? Will it need a rebrand?

On December 9 last year, Alibaba had already formed the Qianwen C-end Business Group, led by Wu Jia, encompassing businesses such as Qianwen APP, Kuake, AI hardware, UC, and Shuqi.

Now, with Qianwen separately integrated into ATH, Wu Jia's management scope and strategic weight may significantly diminish.

Qianwen is no longer just an ordinary product but one of Alibaba's most critical strategic AI businesses outside of e-commerce.

Whoever controls Qianwen holds Alibaba's most crucial consumer-end AI entry point.

Initially, Alibaba had hoped Kuake would serve as this entry point, but Kuake, being a tool product with a mix of functions like cloud storage, scanning, and search, failed to establish a clear user mindset. The baton was then passed to Qianwen.

Previously, even if Kuake's AI performance was lacking, the impact on Wu Jia was limited because Qianwen, the new trump card, was still in his hands. Now, with Qianwen taken away, the situation has fundamentally changed. Wu Jia will continue to manage Kuake, UC, Shuqi, and other businesses, but their strategic priority pales in comparison to Qianwen.

Alibaba is essentially transferring consumer-end AI leadership from its original information business system to ATH, directly overseen by Wu Yongming.

With Tongyi Labs and MaaS being transferred out, Alibaba Cloud's scope has also narrowed.

This adjustment is not hard to fathom. While computing power and tokens are both foundational resources in the AI era, they are ultimately distinct businesses.

Computing power still adheres to the traditional cloud services logic, except that today's scarcest resources have shifted from CPUs, memory, and hard drives to GPUs, with clients primarily being business-end.

Tokens, on the other hand, are more standardized, suitable for both enterprise sales and consumer product integration, offering different imaginative spaces and business models.

After OpenClaw's meteoric rise, in February 2026, the usage of Chinese model tokens on OpenRouter surpassed that of American models for the first time.

Therefore, separating model R&D and MaaS from Alibaba Cloud and merging them into ATH aligns with business logic.

After this separation, Alibaba Cloud's role has also evolved. Previously, it served as both a provider of underlying infrastructure and, to some extent, a participant in AI front-end businesses.

Now, with Tongyi and MaaS taken away, Alibaba Cloud appears more like a pure foundation, responsible for providing computing power, storage, and cloud infrastructure support.

This means Alibaba Cloud's position within the group's AI landscape will become more utility-like. In the future, ATH will define needs, decide how models are built, how tokens are sold, and how applications are implemented; Alibaba Cloud will supply resources and respond to the former's computing power demands.

Of course, this is temporarily just a scope adjustment for business lines and does not involve personnel changes. Wu Yongming currently serves as both Chairman and CEO of Alibaba Cloud, with both ATH and Alibaba Cloud under his purview.

These boundary changes are, at this stage, more about functional reorganization than a redistribution of interest patterns.

As mentioned earlier, when Alibaba sought to compete for AI entry points, both Kuake and DingTalk were heavily favored, one for the consumer-end and the other for the business-end. Later, Kuake faltered, and Qianwen took over.

However, DingTalk has maintained its position. The competition for consumer-end entry points is inherently a battle for native AI apps, making it difficult for legacy tool products to adapt.

The business-end is different. In enterprise collaboration, DingTalk remains Alibaba's most readily available and foundational entry point.

Therefore, incorporating Wukong into ATH makes logical sense. If the consumer-end has been redone with the Qianwen APP, the business-end will naturally follow suit.

Especially considering the recent resource tilt Qianwen has received, Wukong is likely to receive similar support to undertake the crucial responsibility of a 'business-end AI-native work platform'.

DingTalk and Wu Zhao have secured a core position in Alibaba's AI main battlefield.

Wu Zhao is truly fortunate.

When someone ascends to the top, many attribute it to ability, experience, or reputation.

But anyone who has worked in a large company knows that timing, structure, and organizational trust play significant roles.

When Wu Zhao initially left to start a business, he received support from Jack Ma and current CEO Wu Yongming. Later, when his venture didn't succeed, Alibaba stepped in to acquire it, allowing him to return and lead DingTalk.

Now, with the opportunity of an AI-native work platform, he has once again caught a favorable moment.