Author | Xin Jiu

On March 18, Baidu announced an internal appointment: He Jingzhou, formerly the head of the Large Model Algorithm Department in the Basic Model Research and Development Unit (BMU), officially rotated to the Mobile Ecosystem Business Group (MEG) as the head of the Baidu APP Research and Development Center, reporting to Luo Rong, the head of MEG.

Two days earlier, on March 16, Alibaba Group CEO Wu Yongming released an internal letter announcing the formal establishment of the Alibaba Token Hub (ATH) business group, which he will directly lead. This new group integrates core AI businesses, including the Tongyi Lab, MaaS business line, and Qianwen Business Unit.

At almost the same time, across the ocean, Meta was planning the largest-scale personnel adjustment in its history.

According to a Reuters report on March 14, Meta plans to lay off 20% of its non-core business units. Based on its workforce of 79,000 employees by the end of 2025, approximately 16,000 employees will be affected.

At the same time as the layoff news emerged, Meta disclosed its AI investment plan: By 2028, the company's total investment in AI infrastructure will reach $600 billion, with capital expenditure guidance for 2026 as high as $115 billion to $135 billion, nearly double the actual spending in 2025.

Within just one week, Chinese and U.S. tech giants have successively announced major organizational adjustments and strategic moves.

However, these seemingly scattered personnel changes, structural reorganizations, and resource allocations all point to the same core: A global organizational restructuring movement centered around AI is now fully underway in the tech industry.

The core of this movement, unlike the previous competition over large model parameters and computational power, is a silent race in organizational design for the AI era. Giants are using real money and organizational adjustments to explore production relations in the AI era, attempting to find their place in the new wave of technological transformation.

Breaking Down Barriers Between Technology and Business: Baidu's Organizational Logic for AI Implementation

Baidu's personnel adjustment is no ordinary rotation.

It can even be seen as a crucial step in deeply integrating large model technology with core businesses.

He Jingzhou is a composite talent in technology and management cultivated by Baidu, with deep expertise in large model algorithms and as one of the core members of the Wenxin large model's development. MEG is Baidu's cash cow business, with search and recommendation being its foundation, contributing the vast majority of the company's revenue and profits.

Previously, there have been certain barriers between Baidu's large model R&D and business implementation.

The Basic Model Research and Development Unit operates as a technology platform at the group level, responsible for the R&D and iteration of the Wenxin large model, while business groups like MEG are responsible for applying large model technology to specific business scenarios.

This "technology platform-business front" architecture helped Baidu achieve rapid technology reuse and business iteration during the mobile internet era. However, in the AI era, its limitations have gradually emerged.

The reconstruction of businesses by large models involves not just technological empowerment but also a comprehensive reshaping of product logic, business processes, and commercialization models.

Take search as an example. Traditional search engines rely on keyword matching, crawling information across the web, building indexes, and returning relevant web links based on user-input keywords.

In contrast, search in the large model era focuses on understanding user intent, using large models to deeply comprehend user queries and directly provide precise answers or even complete complex tasks.

Such a transformation cannot be achieved by simply adding a large model portal to the existing search product. It requires a comprehensive restructuring of the entire search and recommendation algorithm architecture, product interaction logic, and commercialization model.

In this context, the separation between the technology platform and the business front leads to information loss, reduced decision-making efficiency, and a disconnect between technology and business.

The business front struggles to accurately grasp the capabilities of large models and fully explore the space for product innovation brought by the technology. Meanwhile, the technology platform, detached from frontline business scenarios, finds it difficult to rapidly convert its technological capabilities into commercial value.

He Jingzhou's rotation is likely aimed at breaking down these barriers. By placing the core leader who best understands large model technology directly into a key position on the core business line, he can lead the reconstruction of search and recommendation businesses by large models, overseeing the entire process from technology R&D to product implementation.

This way, the wall between technology and business is broken down, enabling large model capabilities to be directly and rapidly embedded into every aspect of the business, accelerating product iteration and improving technology implementation efficiency.

In fact, this personnel adjustment is not Baidu's first organizational restructuring centered around AI.

As early as the end of 2025, MEG completed a large-scale organizational restructuring, integrating all search-related businesses across PC and mobile to achieve unified management of search operations. At the same time, it established a "search and recommendation integration" team to prepare for the reconstruction of search and recommendation businesses by large models.

Earlier this year, Baidu split Baidu Wenku and Baidu Netdisk from MEG to form a new Personal Super Intelligence Business Group (PSIG), focusing on personal AI application scenarios and exploring subscription-based commercialization models.

Behind these series of organizational adjustments is Baidu's clear judgment of the AI era: AI has moved from the technology R&D stage to the commercial realization stage.

In his 2026 OKRs, Robin Li increased the proportion of aggressive goals to eight, with the core being to drive the implementation of AI technologies across various business scenarios and realize commercial value. Organizational restructuring is the foundation for achieving this goal.

Alibaba's Organizational Restructuring Centered Around Tokens

If Baidu's organizational adjustment aims to break down barriers between technology and business and implement AI technologies in core business scenarios, then Alibaba's establishment of the ATH business group represents a more thorough organizational restructuring centered around the core production factors of the AI era.

In his internal letter, Wu Yongming clarified the core goal of the ATH business group: "Create Tokens, Deliver Tokens, Apply Tokens."

In his view, we are on the eve of an AGI explosion. In the future, a vast number of digital tasks will be supported by hundreds of billions of AI agents, all of which rely on model-generated Tokens. Tokens are becoming the core production factor in the AI era and the primary medium for human-digital world interaction.

Based on this judgment, Alibaba has completely reorganized its AI businesses.

The newly established ATH business group integrates five core business units: Tongyi Lab, MaaS business line, Qianwen Business Unit, Wukong Business Unit, and AI Innovation Business Unit. This covers the entire chain from basic model R&D to model service platforms and then to C-end personal AI assistants and B-end enterprise AI applications.

Tongyi Lab is responsible for "creating Tokens," developing leading multimodal basic large models, continuously pushing the limits of model capabilities, and providing the most advanced model capabilities for the entire business group, the entire group, and even the industry. It serves as the technological foundation of the ATH business group.

The MaaS business line is responsible for "delivering Tokens," building an efficient and open model service platform and technological system. It delivers the model capabilities developed by Tongyi Lab to various business lines within the group, as well as to external enterprise clients and developers through standardized interfaces, supporting the development of the entire AI ecosystem.

The Qianwen Business Unit, Wukong Business Unit, and AI Innovation Business Unit are responsible for "applying Tokens." They develop AI-native application products for C-end personal users, B-end enterprise clients, and innovative scenarios, respectively, transforming the value of Tokens into user value and commercial value.

This organizational structure completely breaks away from Alibaba's past model of dividing business groups by business lines.

In the past, Alibaba's AI capabilities were scattered across various business departments, including DAMO Academy, Alibaba Cloud, Taobao, Tmall, and DingTalk. Each department operated independently, leading to redundant resource investment, fragmented technologies, and an inability to rapidly reuse model capabilities.

For example, to implement the Tongyi large model developed by DAMO Academy in Taobao's e-commerce scenarios required cross-departmental coordination, resulting in significant information loss and low implementation efficiency. Similarly, DingTalk's use of large model capabilities required multiple rounds of communication with DAMO Academy and Alibaba Cloud, making rapid product iteration difficult.

The establishment of the ATH business group integrates all core AI resources into a unified organization, directly overseen by Group CEO Wu Yongming.

This way, the entire organization has a highly unified goal: maximizing the value of Tokens throughout their entire lifecycle. From basic model R&D to model service delivery and then to application scenario implementation, the entire chain is completed within the same organization, leading to higher decision-making efficiency, faster technology implementation, and more efficient resource utilization.

On March 17, the day after the ATH business group was established, Alibaba launched the enterprise-grade AI-native work platform "Wukong." This is the first product to be implemented after the establishment of the ATH business group.

Based on the capabilities of the Tongyi large model, the "Wukong" platform deeply integrates agent technology into enterprise workflows. It can coordinate multiple agents within a single interface to complete complex enterprise tasks such as document editing, spreadsheet updates, approval form filling, meeting audio transcription, and in-depth research. It can also natively integrate with DingTalk and will support integration with mainstream communication platforms like Slack, Microsoft Teams, and WeChat in the future.

The rapid implementation of the "Wukong" platform directly reflects the advantages of the ATH business group's organizational structure. From basic model R&D to enterprise-grade application product implementation, the entire process is completed within the ATH business group without the need for cross-departmental coordination. This significantly improves product iteration speed and implementation efficiency.

From Alibaba's series of actions, it is clear that Alibaba no longer views AI as a single product line but as the core infrastructure for the entire group in the future. It has redesigned its entire organizational structure around the core production factors of the AI era.

This organizational restructuring is more fundamental and critical than simple technology R&D and resource investment.

The Efficiency Paradox: Meta's Layoffs and AI Gamble

While domestic tech giants are promoting the implementation and integration of AI technologies through organizational adjustments, Meta across the ocean is exploring the organizational logic of the AI era in a more aggressive manner.

If this layoff plan is implemented, it will be Meta's largest personnel optimization action since its "Year of Efficiency" restructuring at the end of 2022, far exceeding the total number of layoffs in the two rounds conducted in 2022-2023.

After the layoff news emerged, the capital market reacted positively. The market generally believes that this layoff is a cost-cutting and efficiency-improving measure taken by Meta to cope with its massive investment in AI.

A $600 billion investment in AI infrastructure is equivalent to the total net profit of Meta over 12 years. On one hand, there are large-scale layoffs in non-core business units to cut costs and improve efficiency. On the other hand, there is an astronomical investment in the AI field. Meta's seemingly contradictory actions reflect the efficiency paradox of the AI era.

In traditional business logic, the core of efficiency is the return on investment, the profit created by each employee, and achieving maximum output with minimum input. According to this logic, large-scale layoffs are aimed at reducing costs and increasing profits, while astronomical investments would significantly increase costs and reduce profits, making the two completely contradictory.

However, in the AI era, the logic of efficiency has fundamentally changed.

For Meta, the current core goal is not short-term profit maximization but survival in the AGI era.

In an internal speech, Mark Zuckerberg clearly stated, "Projects that once required large teams can now be accomplished by a single highly talented individual."

Behind this statement is Zuckerberg's judgment on the organizational production methods in the AI era: AI will completely restructure organizational production methods. A large amount of repetitive and standardized work will be replaced by AI. Tasks that once required dozens or even hundreds of people to complete can now be done by just a few people with AI tools.

Based on this judgment, Meta's layoffs are actually a thorough restructuring of the organization.

The layoffs target non-core, inefficient positions that will be replaced in the AI era, retaining only top talents who can leverage AI tools and create core value.

At the same time, the costs saved from layoffs and the profits obtained from traditional advertising businesses are all invested in the AI field to build AI infrastructure, develop more advanced large models, and recruit top AI talents.

This logic is indeed the most efficient strategic choice for companies in the AI era. Because competition in the AI era is winner-takes-all. Once an intergenerational gap is formed in basic models and AI infrastructure, latecomers have almost no chance of catching up.

If all resources are not concentrated on AI now and a leading advantage is not established in core AI technologies, even if the current advertising business is highly profitable, it will eventually be eliminated by the times.

Just like Nokia in the feature phone era, which once dominated the global mobile phone market and generated substantial profits. However, because it failed to keep up with the technological transformation in the smartphone era, it quickly fell from the top of the industry to the bottom within a few years.

Meta's efficiency paradox is essentially a trade-off between short-term financial efficiency and long-term survival efficiency. In the face of intergenerational transformation in the AI era, short-term profits are no longer the most important indicator. Whether a company can establish a leading position in core AI technologies, complete its AI restructuring, and survive in the AGI era is what the giants care about most.

Organizational Design in the AI Era: Restructuring Production Relations

From Baidu's personnel rotation to Alibaba's establishment of the ATH business group and Meta's layoffs and AI gamble, a series of actions by global tech giants all point to the same core: Competition in the AI era has shifted from isolated technological breakthroughs to comprehensive competition in organizational capabilities.

A global competition in organizational design centered around AI is now fully underway in the tech industry.

Over the past few decades, every technological intergenerational transformation in the tech industry has brought about profound changes in organizational structures.

In the PC era, Microsoft's divisional system allowed it to dominate the PC software market. In the internet era, Google's flat organizational structure enabled it to quickly respond to market changes and establish a leading position in search and advertising. In the mobile internet era, Alibaba's "big middle platform, small front desk" model and ByteDance's flat, project-based organizational structure allowed them to rapidly incubate new products, seize mobile internet traffic dividends, and grow into industry giants.

Now, the arrival of the AI era is completely subverting the organizational structure logic that has been formed over the past few decades.

In the era of mobile internet, the core of organizational structure is 'traffic,' which involves how to quickly acquire and monetize traffic. Therefore, the design of organizational structure focuses on enhancing the speed of response to market changes and improving the efficiency of acquiring and monetizing traffic.

In the AI era, the core of organizational structure is 'intelligence.' It involves how to deeply embed AI capabilities into every aspect of the business, how to deeply integrate technology and business, and how to maximize value around the core production factors of AI.

This change has brought about a fundamental transformation in the underlying logic of organizational structure.

The previous architecture of 'technology platform - business front' has gradually become ineffective in the AI era. This is because AI's reconstruction of business involves a comprehensive reshaping of the entire business logic, requiring deep integration between technology research and development and product implementation, as well as full-process collaboration from technology research and development to product iteration.

If the technology platform and business front are separated, it will lead to a disconnect between technology and business, making it impossible to achieve true reconstruction of the business by AI.

Microsoft's transformation is the best example.

After Satya Nadella took office, his first action was to reform Microsoft's organizational structure, breaking away from the past Windows-centric divisional system and establishing a cloud-centric organizational structure. This transformed Microsoft from a closed software company into an open cloud services company.

In the AI era, Microsoft once again adjusted its organizational structure, deeply embedding OpenAI's technology into all business lines such as Office, Azure, and Bing. Each business department now has a dedicated AI team directly responsible for implementing AI technology in their respective business scenarios. This has allowed Microsoft to once again stand at the top of the technology industry in the AI era.

Now, global tech giants are exploring organizational structures for the AI era along this path.

ByteDance recently adjusted its organizational structure, splitting the AI department at the group level into various business groups such as Douyin, Toutiao, and Feishu. Each business group now has its own AI team directly responsible for implementing AI technology in their respective business scenarios, achieving deep integration between AI and business.

Tencent has also adjusted its large model team, splitting the large model R&D team into various business groups such as the WeChat Business Group, Interactive Entertainment Business Group, and Cloud and Smart Industries Business Group. This allows AI capabilities to be directly embedded into core businesses such as WeChat, gaming, and cloud services.

These organizational adjustments are essentially explorations of the production relations in the AI era. As a new productive force, AI will inevitably bring about changes in production relations. And organizational structure is precisely the core carrier of production relations. Only by establishing an organizational structure that matches AI's productive forces can we truly unleash AI's productive forces and realize the commercial value of technology.

The Finale of the Competition: Organizational Capability Determines the Future

Now, this competition around AI-driven organizational design has just begun.

Global tech giants are exploring organizational structures suitable for the AI era based on their own business characteristics.

Some choose to place core technology leaders on the business frontlines, breaking down barriers between technology and business; some choose to integrate all AI resources and establish a full-link organizational closed loop around core production factors; some choose to reconstruct the organization's production methods through layoffs and resource allocation, concentrating all resources on the core tracks of AI.

These different explorations do not have absolute right or wrong; they only differ in suitability.

But one thing is certain: In the AI era, technology can be replicated, computing power can be purchased, and talent can be recruited. However, organizational capability cannot be replicated; it requires long-term accumulation and refinement.

In the mobile internet era, there were countless companies with leading technologies, strong capital, and large user bases. However, due to rigid organizational structures, they were unable to keep up with the changes of the times and were ultimately eliminated by the market. In the AI era, the speed of this generational change will be faster, competition will be more intense, and the requirements for organizational capability will be higher.

For tech giants, every organizational adjustment, every personnel change, and every resource allocation now is a battle for survival in the AI era.

— END —