Alibaba Group announced its financial results for the third quarter of fiscal year 2026. The group reported revenue of RMB 284.843 billion, representing a 2% year-over-year increase. Excluding the revenue from disposed businesses like Intime and Sun Art Retail, revenue grew by 9% year-over-year. Revenue from Alibaba Cloud accelerated, growing by 36%, while AI-related product revenue achieved triple-digit growth for the tenth consecutive quarter. Adjusted net profit was RMB 16.71 billion, and adjusted EBITDA was RMB 34.06 billion.

Following the earnings release, Alibaba Group's Executive Chairman Joseph Tsai, CEO Eddie Yongming Wu, CFO Toby Xu, and CEO of Taobao and Tmall Group, Fan Jiang, attended a conference call to discuss the results and answer questions from analysts.

The following are key excerpts from the analyst Q&A session:

**Bernstein Analyst Robin Zhu:** Could you elaborate on how the establishment of the Token Hub (ATH business group) changes the collaborative dynamics, from a design perspective, between the various cloud and AI-related businesses? Strategically, what changes or new goals does this new structure facilitate that are superior to the previous arrangement? Also, could management outline the priorities regarding AI and voice? For instance, is the top priority market share and revenue growth, as mentioned in your targets, or is it possessing the best model capabilities, developing more consumer-facing applications, or advancing AI agents? Could you please rank these priorities?

**Eddie Yongming Wu:** Thank you for your question. The purpose of establishing the ATH business group is closely related to the current era of technological transformation. From the second half of 2025 through the first few months of 2026, we have observed that AI is entering an era driven by AI Agents. The key distinction in this AI Agent-driven era, compared to the earlier stages of AI, is the tight integration required between models, applications, and hardware. This integration is crucial for developing effective models and applications. In the early stages of AI, much pre-training data came from static sources. Now, as we enter the AI Agent era, a significant portion of the capability to enhance models and improve applications comes from the close coupling of models and applications, and from iterative data tuning within customer usage scenarios.

Therefore, looking at the five layers of AI—the application layer, the model layer, followed by AI infrastructure/cloud computing, and then the chip layer—we believe that in this Agent era, the tight integration between models and applications is critically important.

Next, let me explain the interrelationships among the businesses within ATH. Based on observed industry trends, it is becoming very clear that a rich ecosystem of AI agents will emerge at the application layer, fostering immense innovation. The variety of applications at this layer will be vast. Our own layout at the application layer includes the "Qianwen" App as our consumer-facing personal assistant. We also aim to develop "Wukong" into a leading enterprise-facing personal assistant for the Chinese market. Simultaneously, we anticipate a multitude of industry-specific, vertical, and scenario-based AI agent applications supporting various sectors. Beyond the application layer, we also require a strong MaaS (Model-as-a-Service) business.

The MaaS business acts as a conduit between the model and application layers. In addition to supporting our internal applications, a robust MaaS business will enable us to support a wide array of external AI industry applications. We believe the market value and potential here are substantial.

From this perspective, we view the AI application layer as the primary distribution channel for future AI models and tokens. The most powerful model capabilities will attract more applications to use our models, and a more efficient, capable MaaS product will better connect these two layers. This is the business logic behind the design of this business group.

Therefore, regarding the model and application layers, our priority is unequivocally to build the most intelligent models. Only the strongest models can drive the expansion of application scenarios across various industries and attract diverse applications to utilize our MaaS business.

However, I want to emphasize that building the strongest models requires collaboration with various industries, our own B2C and B2B applications, and connecting with industry applications via MaaS to engage more users. This creates a data flywheel and business闭环 (closed loop). Through more scenarios, more data, and more users iterating on our models, we can form this data flywheel, continuously enhancing model capabilities. This is a key reason for establishing the ATH business group. To summarize, our priority is enhancing model capability, but achieving this requires the collective effort across the model, application, and MaaS domains within ATH to drive long-term model improvement.

**Bank of America Merrill Lynch Analyst Joyce Ju:** Good evening management, and congratulations on the solid progress in cloud and AI. My question concerns CMR (Customer Management Revenue). We observed a slowdown in growth, particularly in the December quarter, given macro pressures in China's overall online retail sector. However, we recently saw online retail growth re-accelerate in January and February. Could you share your perspective on CMR trends entering the March quarter? Have you begun to see improvements on the consumer side?

**Fan Jiang:** I will address this. As you mentioned, the December quarter faced challenges due to weak macro consumption, a warm winter, and the later timing of the Chinese New Year this year, which impacted our growth that month. Additionally, with an extended promotional season, our investment in consumer benefits increased compared to previous years. Consequently, both CMR and e-commerce profits saw a slowdown in the December quarter. However, since the beginning of this year, we have observed a very clear consumption recovery. Coupled with the pull from our instant retail strategy, the growth rates for both physical goods GMV and CMR have significantly rebounded, and our e-commerce profits have also shown marked improvement.

**Morgan Stanley Analyst Gary Yu:** Thank you for taking my question. It's regarding instant retail. I understand Alibaba has achieved some phased results in market share and UE (Unit Economics) improvement over the past few months. Looking ahead, what will be the priority: gaining market share, or leveraging this opportunity to further optimize UE and narrow losses? Also, how should we view the synergy between instant retail and traditional e-commerce, and how might this synergy translate into faster CMR growth in the future?

**Fan Jiang:** I'll answer this. Yes, we continue to see that alongside market share gains, improvements in logistics efficiency, enhanced commercialization capabilities, and optimization of order structure are driving continuous UE optimization. We believe our UE will improve further. We have also seen significant platform-wide pull from flash sales over the past year. The annual active buyers for the overall e-commerce platform, including flash sales, increased by 150 million, and annual active buyers for physical goods e-commerce grew by 100 million. The growth in annual active buyers for Taobao's physical goods e-commerce exceeded the total growth of the past three years combined. While new users typically have lower average order value and purchase frequency compared to mature users in the short term, we aim to continuously improve the ARPU and purchase frequency for this segment, which we see as a new growth engine for the platform in the coming years. Furthermore, flash sales have significantly boosted related categories, particularly food, fresh produce, and health products, and have also accelerated the development of instant retail businesses like Hema and Tmall Supermarket.

Regarding future expectations: Firstly, we maintain the target for the overall GMV of the instant retail business to exceed RMB 1 trillion by fiscal year 2028. At that scale, we believe we can achieve scaled positive cash flow. Additionally, we anticipate the instant retail business segment will achieve overall profitability by fiscal year 2029.

Flash sales and instant retail have become foundational businesses for the Taobao and Tmall e-commerce板块 (sector), contributing significantly through new customer acquisition, boosting user activity, fulfilling diverse consumption scenarios, driving GMV and commercialization, and enhancing logistics and infrastructure. They play a crucial strategic role in the long-term development of Taobao and Tmall in the AI era. Over the next two years, we will continue to invest firmly to achieve the goal of surpassing one trillion in total scale while maintaining a leading market position.

**Citi Analyst Alicia Yap:** Thank you for taking my question. I have some questions regarding your chip business, "Pingtouge." Recent reports suggest Alibaba is considering spinning off Pingtouge for an IPO. Could management provide related information? Also, could you share some performance data for Pingtouge? For instance, beyond the 470,000 chips delivered, what is the revenue? What is the expected growth rate for the coming year? You mentioned that about 60% of demand comes from external customers. Could you elaborate on whether these chips supplied to external customers are primarily used for inference, while internal use is for model training or other purposes? How does Pingtouge's chip compare to other domestic chips?

**Eddie Yongming Wu:** Thank you for this question. Pingtouge is a vital part of Alibaba's comprehensive AI strategy. Within the current domestic AI chip ecosystem, we believe Pingtouge's technical and product capabilities are in the first tier. Our products cover the complete AI workflow from training and fine-tuning to inference. Pingtouge AI chips are already deeply and extensively used within Alibaba Cloud, both in training scenarios and in the "Bailian" inference scenario. Furthermore, over 60% of Pingtouge chips deployed in Alibaba Cloud's public cloud and hybrid cloud products are used by external commercial customers. These external customers span industries such as internet, autonomous driving, and smart manufacturing.

These external commercial customers utilize Pingtouge chips for both their training and inference scenarios. Our Pingtouge software stack offers good compatibility with the CUDA ecosystem, allowing customers to migrate their systems and projects with minimal time investment. Another key point regarding Pingtouge's importance to Alibaba is that, given the relative lag of domestic chips compared to foreign chips in manufacturing processes and performance, we aim for deeper co-design with Alibaba Cloud infrastructure and the Tongyi Qianwen models to achieve better cost-performance. This is a distinguishing feature of our Pingtouge chips compared to other chip companies.

Therefore, our primary goal is to create higher AI capability per unit cost, making it a key product for reducing inference costs on our future Bailian platform.

Beyond overall AI efficiency improvement and cost reduction, Pingtouge holds significant value in the specific context of China's AI industry: ensuring AI computing power supply. Over the next 3 to 5 years, global AI computing power will be in a very tight supply, especially within the Chinese market. As the only cloud computing company in China with self-developed chip capabilities, Pingtouge is crucial for Alibaba Group. Providing additional AI computing power supply will help fuel higher growth momentum for our cloud and AI businesses, including our MaaS business.

Finally, let me share some future expectations. Pingtouge has successfully achieved commercial application over the past two years, with cumulative chip shipments exceeding 470,000 units and annualized revenue reaching the tens of billions of RMB level. For 2026 and 2027, we expect the scale of high-quality AI chips produced by Pingtouge to continue expanding. This will provide sufficient computing power assurance for the group's overall AI business and serve as a strong growth driver. It will also significantly contribute to future profitability improvements. In summary, Pingtouge's value to Alibaba lies not only in cost optimization but also, critically, in supply assurance. In an era of computing scarcity, this is vital for Alibaba's AI strategy. Therefore, a future IPO for Pingtouge is not ruled out, but there is no specific timeline at present.

**China Citic Bank Analyst:** Good evening management. Thank you for taking my question. It concerns the commercial targets of Alibaba Group's AI strategy mentioned, specifically the goal for AI-related revenue to exceed $100 billion over the next five years. Could management provide more details on this target? For instance, the implied compound annual growth rate (CAGR) over five years and the key growth drivers? How should we understand the relationship between this scaled revenue growth and when we might see an improving trend in Alibaba Cloud's profit margins?

**Eddie Yongming Wu:** Thank you for your question. Regarding the target for AI and cloud-related business revenue to exceed $100 billion in the next five years, we have relatively strong visibility on the path to achieving this goal, based on current market growth potential, our existing foundation, and product base.

We believe the primary growth driver will be breakthroughs in large language model capabilities. In the first few months of 2026, we have already observed clear trends indicating that large models are beginning to possess the ability to handle complex B2B workflows. As more companies internally deploy AI Agent-driven models to assist with tasks, the fundamental nature of the market addressing IT budgets for AI and cloud is changing significantly. When enterprises consume tokens, they are increasingly viewing them not as part of the IT budget but as production or R&D costs—part of their means of production. This is a fundamental, long-term intrinsic factor for future AI growth.

We see three major growth drivers. First, MaaS business driven by large models will be a core growth engine. The growth of MaaS will be fueled by its users, including our own applications, their customers, and the diverse array of AI applications and scenarios within various industries, including AI application software. We believe MaaS-driven growth will be a core factor for future AI and cloud revenue.

Secondly, for AI cloud computing, another significant incremental scenario exists. We anticipate a substantial market for public MaaS services. Additionally, a market for enterprise-level internal inference and training will persist long-term among many medium and large enterprises. Depending on their business form, security requirements, or specific application scenario needs, some scenarios will use public MaaS API services, while many others will require on-premises deployment. These scenarios represent a major incremental opportunity for Alibaba Cloud's AI infrastructure.

A third growth driver, often overlooked, involves traditional CPU-centric cloud computing. In the AI Agent era, there is immense growth potential here. Traditional cloud computing was designed for IT engineers—a potential customer base in China numbering in the millions, perhaps up to 10 million. However, the AI agents created by large models could number in the billions. Their operating environments will require massive support from traditional, CPU-centric cloud computing—CPUs, databases, storage, and substantial memory to support their long-term operation. Therefore, the transformation of the traditional cloud computing market from a product for human IT engineers to one highly suitable for agent interaction presents a huge growth space. This transformation is a key focus for Alibaba Cloud's upgrade this year.

As company revenue scales and the AI business transitions more from selling resources to selling intelligent capabilities, it represents a significant business model upgrade. Combined with the cost-saving and efficiency gains from our self-developed Pingtouge chips, we believe that as our AI and cloud revenue continues to grow, the profit margin of the cloud business should see visible improvement. This will likely be a continuous, though not necessarily linear, process. It could involve step-changes due to scale effects or significant increases in Pingtouge chip production volume, related to the manifestation of overall product scale effects.

Regarding the CAGR from 2026 to 2031, it is quite clear if calculated. However, the progression of R&D investment and market growth will not be linear. Investments made today may yield greater growth after one or two years. Nonetheless, we are very confident in achieving the overall five-year target.

**J.P. Morgan Analyst Alex Yao:** Thank you for the question. Shifting focus to e-commerce development: previously, it was mentioned that e-commerce would enter a three-year investment cycle. Is this primarily due to the opportunities seen in flash sales/instant retail, leading to some adjustments? If not adjusted, would we be roughly in the middle of this three-year cycle? Could you share your perspective on the阶段性定位 (phase positioning) and considerations within this track, leading towards financial improvement and a relatively stable financial harvest period by the end of the three years?

**Fan Jiang:** As I mentioned earlier, we are making significant investments in instant retail this year, which we see as a major opportunity. We will continue to invest over the next two years to achieve the goal of instant retail scale exceeding one trillion. We believe that after two years, the investments in instant retail will bring positive economic returns to the entire e-commerce sector.

Regarding AI's impact on e-commerce, which has been discussed extensively, we believe AI will have a very significant influence. Of course, three years is a long timeframe in AI, which evolves weekly and monthly. We are actively investing in AI. We will continuously introduce new user experiences and upgrade AI-based business operation models for merchants this year. The emergence of AI will bring substantial upgrades to many sub-fields within e-commerce and create significant opportunities, particularly when combined with our B2B business. We will strive to seize these new opportunities.