Full Text | Alibaba Q3 Earnings Call Transcript: Self-Developed Chip "Pingtouge" Does Not Rule Out IPO in the Future

Special Topic: Focus on U.S. Stock Market Q4 2025 Earnings Reports

Alibaba Group announced its Q3 fiscal year 2026 financial results, with total revenue of 284.843 billion yuan, up 2% year-over-year. Excluding the disposed businesses of Intime and Giga Retail, revenue grew 9%. Alibaba Cloud revenue accelerated 36%, and AI-related product revenue continued triple-digit growth for the tenth consecutive quarter. Adjusted net profit was 16.71 billion yuan; adjusted EBITDA was 34.06 billion yuan.

See: Alibaba: Q3 FY2026 Revenue of 284.84 Billion Yuan

Following the release of the earnings report, Alibaba Group Chairman Joe Tsai, CEO Daniel Zhang, CFO Maggie Wu, and CEO of Alibaba’s e-commerce business group Jiang Fan participated in a conference call to interpret the results and answer analyst questions.

Below are the main points from the Q&A session:

Bernstein Analyst Robin Zhu: Could you introduce how, with the establishment of Token Hub (ATH Business Group), the collaboration methods among different cloud and AI-related businesses might change from a design perspective? Additionally, from a strategic standpoint, what changes or new goals might this new structure bring? Is it more advantageous than previous arrangements? Also, could management elaborate on your priorities in voice and AI? For example, is the top priority market share and revenue growth, as you just announced, or is it developing the best model capabilities, or expanding consumer applications like AI agents? Please rank these priorities. Thank you.

Wu Yongming: Thank you very much for your question. I believe the purpose of establishing ATH, the Business Group, is related to the current era and technological transformation. Since the second half of 2025 through the first few months of 2026, AI has entered an era driven by AI Agents. The key difference in this era compared to early AI stages is the close integration of models with hardware applications. How to develop good models and applications is crucial. Early on, pre-trained data mainly came from static datasets.

Now, entering the AI Agent era, much of our focus is on enhancing model capabilities and improving applications through tight integration of models and applications, as well as real-time data tuning in customer scenarios.

From this perspective, considering the five levels of AI—application layer, model layer, AI infrastructure (like cloud computing), and chips—the close coupling of models and applications is the most critical in the agent era.

Let me also explain the internal relationships among our ATH business units. Based on current industry trends and future development, it’s clear that AI agents will form an ecosystem with abundant innovation at the application layer. Applications will be diverse and rich. We plan to launch the “Qianwen” app as a consumer-facing personal assistant at the application level, and aim to develop “Wukong” as a B2B personal assistant for Chinese enterprises.

Meanwhile, we see many industry-specific or vertical AI applications supporting various sectors. To support this, we need a strong MaaS (Model as a Service) business, which acts as a channel bridging models and applications. Besides supporting internal applications, MaaS will enable us to serve external industry applications, which we believe have huge market potential.

Therefore, the AI application layer will be the main distribution channel for AI tokens, and the strongest models will attract more applications. A more capable “Mars” product will better connect models and applications. This is the business logic behind our group’s structure.

In summary, our top priority is to build the most powerful AI models, because only the best models can expand application scenarios across industries and attract diverse industry applications to our MaaS platform. However, developing the strongest models requires integrating industry-specific applications, consumer and B2B applications, and MaaS to create a data flywheel and business loop. This iterative process with more scenarios, data, and users is why we established the ATH Business Group. Our core focus is on improving model capabilities, which depends on joint efforts across models, applications, and MaaS.

Bank of America Analyst Joyce Ju: Good evening, management, and congratulations on solid progress in cloud and AI. My question concerns CMR. We’ve observed a slowdown, especially in the December quarter, amid macro pressures on China’s online retail sector. Recently, we saw online retail growth accelerate again in January and February. Could you share your outlook for CMR trends in the March quarter? Are you seeing any consumer-side improvements?

Jiang Fan: I’ll answer this. As you mentioned, the December quarter faced challenges due to macroeconomic weakness, a warm winter, and a late Spring Festival. Additionally, extended promotional periods increased investments in consumer rights, which slowed both CMR and e-commerce profits. However, since the start of this year, we’ve seen a clear recovery in consumption. Driven by our instant retail strategy, physical e-commerce transactions and CMR growth have rebounded significantly, and e-commerce profits have improved notably.

Morgan Stanley Analyst Gary Yu: Thank you for taking my question. Regarding instant retail, I understand Alibaba has achieved some milestones recently in market share and user experience (UE) improvements. Looking ahead, what will be your priority? Market share, or further optimizing UE and narrowing losses? Also, how do you see the synergy between instant retail and traditional e-commerce, and how might these synergies accelerate CMR growth in the future? Thank you.

Jiang Fan: I’ll respond. Yes, we continue to see improvements in market share, driven by enhanced logistics efficiency, better monetization, and optimized order structures, which further improve UE. We believe UE will continue to improve. Over the past year, flash sales have significantly boosted the platform. Including flash sales, the number of active buyers on our platforms increased by 150 million, with physical e-commerce active buyers up by 100 million. Taobao’s physical e-commerce active buyers have grown more in a year than in the previous three years combined. New users tend to have lower short-term order values and purchase frequency compared to mature users, but we aim to increase ARPU and purchase frequency over time, which will be new growth engines for the platform. Flash sales have also notably driven categories like food, fresh produce, and health, accelerating development of instant retail businesses like Hema and Maicai.

Looking ahead, we maintain the goal of exceeding 1 trillion yuan in total transaction volume for FY2028. Based on this scale, we believe we can achieve positive cash flow, and expect FY2029 to be profitable overall.

Flash sales and instant retail are foundational to Taobao and Tmall, driving user acquisition, engagement, and supporting diverse consumption scenarios, as well as logistics and electronics. They are strategic for Alibaba’s long-term AI era development. Over the next two years, we will continue to invest to reach over 1 trillion yuan in scale and maintain market leadership.

Citi Analyst Alicia Yap: Thank you for the opportunity. I have a question about your chip business, “PingTouGe.” Recent reports suggest Alibaba plans to spin off PingTouGe for an IPO. Can management share any details? Also, could you provide some performance data? For example, beyond the 470,000 chips delivered, what is the revenue? What is the expected growth rate for the next year? You mentioned that 60% of demand comes from external clients—can you elaborate on whether these chips are mainly for inference, and whether internal use is for training or other purposes? How does PingTouGe compare to other domestic chips?

Wu Yongming: Thank you for your question. PingTouGe is a key part of Alibaba’s full-stack AI deployment. I’d like to take this opportunity to explain. Currently, in China’s AI chip ecosystem, PingTouGe’s technology and product capabilities are among the top tier. Our products cover the entire AI workflow from training and fine-tuning to inference. PingTouGe chips are already used extensively in Alibaba Cloud’s “Bailian” inference scenarios and in large-scale training. Over 60% of PingTouGe chips are used by external commercial clients across internet, autonomous driving, and manufacturing sectors.

These clients use PingTouGe chips for both training and inference. Our software stack is well compatible with CUDA ecosystems, enabling quick migration for clients. PingTouGe’s importance to Alibaba also lies in its ability to deepen integration with Alibaba Cloud infrastructure and models like Tongyi Qianwen, improving cost-performance ratio. Unlike other domestic chips, PingTouGe aims to deliver higher AI capability and lower inference costs, supporting our Bailian platform.

Furthermore, in China’s current AI industry context, PingTouGe’s self-developed chips are crucial for supply security, especially as global AI compute resources are expected to be tight over the next 3–5 years, particularly in China. As the only domestic cloud provider with self-research chip capability, PingTouGe is vital for Alibaba’s AI strategy, providing more compute power and driving growth in cloud and MaaS businesses.

Looking ahead, PingTouGe has already shipped over 470,000 chips, generating annual revenue in the hundreds of millions. We expect production to continue expanding in FY2026 and FY2027, providing strong compute support and growth momentum for Alibaba’s AI initiatives. Its value is not only cost optimization but also supply assurance. Given the scarcity of AI compute resources, PingTouGe is strategically important for Alibaba’s AI future. An IPO is possible, but no specific timeline has been set.

CITIC Bank Analyst: Good evening, management. Thank you for taking my question. It concerns Alibaba’s AI strategic business goal of surpassing $100 billion in revenue over the next five years. Could you share more details, such as the compound annual growth rate (CAGR) and growth drivers? How should we interpret the timeline for profitability improvements in Alibaba Cloud, given the scale of revenue growth?

Wu Yongming: Thank you. Achieving over $100 billion in AI and cloud revenue in five years is feasible based on current market growth potential, our foundation, and product lineup. The key driver will be breakthroughs in large AI models. By early 2026, we see clear signs that large models are capable of handling complex B2B workflows.

As more companies deploy AI Agents powered by large models, the traditional IT budget market is fundamentally changing. Many enterprises now see tokens not just as IT expenses but as part of their production and R&D costs, which is a long-term growth factor for AI.

Our three main growth engines are: first, large models driving MaaS as a core growth engine; second, expanding AI cloud scenarios, including public MaaS and enterprise internal inference and training markets; third, transforming traditional CPU-based cloud computing to better support AI Agents, which requires significant infrastructure upgrades.

As AI business scales, resource sales will shift from merely providing infrastructure to offering intelligent capabilities, boosting profitability. The scale effects and PingTouGe chip expansion will further enhance margins. Overall, we are confident in reaching our five-year revenue target, though growth may not be perfectly linear—initial investments may take a year or two to yield significant returns.

JPMorgan Analyst Alex Yao: Thank you. Shifting to e-commerce, we previously mentioned a three-year investment cycle, partly driven by opportunities like flash sales and delivery. If no adjustments are made, are we already in the mid-phase of this cycle? Are we approaching a period of financial stabilization and returns by the end of this cycle? Could you share your strategic positioning and outlook for this stage? Thank you.

Jiang Fan: As I mentioned, we’ve made significant investments in instant retail this year, seeing it as a key opportunity. We plan to continue investing over the next two years to reach over 1 trillion yuan in transaction volume, which we believe will generate positive economic returns for our e-commerce segment.

Regarding AI, as previously discussed, AI will profoundly impact e-commerce. Although three years is a long horizon, AI evolves rapidly—weekly and monthly improvements. We are actively investing in AI to enhance user experience and merchant operations, launching new features and AI-driven business models. AI will bring substantial upgrades across many e-commerce segments, especially in B2B, creating new opportunities. We are committed to seizing these opportunities. (End)