LAIKA 

 2026/05/14 

The Turning Point for AI Commercialization Has Arrived

On the evening of May 13, Alibaba Group released its financial report for the fourth quarter of FY2026 (January-March 2026). Amid widespread concerns about sluggish growth among internet giants, Alibaba responded with a report card featuring 'AI exceeding expectations and e-commerce stabilizing the foundation.'

This quarter, Alibaba Cloud's AI-related product revenue exceeded RMB 9 billion in a single quarter, accounting for over 30% of external cloud business revenue for the first time, marking Alibaba AI's official entry into the 'harvest period' of large-scale monetization. Meanwhile, Taotian Group regained its e-commerce growth momentum through instant retail (flash sales) business.

01

Specific Businesses

1. AI Cloud

Alibaba Cloud's AI business annualized revenue has exceeded RMB 35.8 billion, maintaining triple-digit growth for 11 consecutive quarters. AI-related product revenue now accounts for 30% of Alibaba Cloud's external commercial revenue; AI revenue for this quarter alone was RMB 9 billion. It is expected that the revenue share of AI-related products will exceed 50% within the next year.

It is estimated that the annualized recurring revenue (ARR) for model and application services, including Bailian, will reach RMB 10 billion in the June quarter and exceed RMB 30 billion by the end of the year. Compared to November/December last year, token usage in May/June this year increased 10x, with the recent ARR for model and application services exceeding RMB 8 billion.

Revenue from model and application services consists of two parts: Bailian MaaS API + AI native software subscription value. The majority of Bailian's revenue comes from self-developed models. Growth in the MaaS business will positively impact gross margins, as MaaS business gross margins are higher than IaaS.

Compared to November/December last year, token usage for Bailian increased 10x in May/June this year, with the recent ARR for Bailian exceeding RMB 8 billion.

Currently, almost no server cards are idle. Future demand over the next 3-5 years can generate significant returns on current AI investments.

Currently, Alibaba Cloud's server deployment and launch costs have increased by 100% compared to two years ago. The increase in replacement costs has a pricing traction (pricing influence) effect on customer demand, positively boosting pricing for various future cloud services. Considering multiple factors, Alibaba Cloud's gross margin will see a significant improvement in the next 1-2 years, with changes beginning to be visible in the next 1-2 quarters.

The ROI for AI2B is easier to calculate, with stronger willingness from enterprises to pay. AI2C requires a certain investment cycle, with commercialization progress for 2C AI assistants expected within the next 1-2 years.

CAPEX: The scale to be built in the future is 10x that of 2022. Funds invested to acquire computing power centers are expected to far exceed RMB 380 billion over three years, but future funding methods will be diversified, such as through opex leasing and selling T-Head AI servers.

Currently, T-Head's self-developed GPU chips have achieved large-scale mass production, with over 60% of computing power serving external commercial customers, covering core scenarios such as internet finance and autonomous driving.

Self-developed chips currently account for a small portion of Alibaba Cloud's deployment. In addition to GPUs, T-Head also develops CPUs and storage network chips, with opportunities for full-stack self-development from GPUs to storage and network chips in the future. Increasing the proportion of self-developed chips can drive gross profit improvement. Considering the performance gap in domestic chips, there is room for cost-effectiveness improvement in future self-developed chips. The impact on gross margins will depend on capacity expansion and the proportion of Stock substitution (stock replacement).

2. E-commerce Business

This quarter, China's e-commerce group revenue reached RMB 122 billion, up 6% year-on-year; customer management revenue (CMR) increased by 8% year-on-year on a like-for-like basis (1% growth on a reported basis, mainly due to adjusting merchant subsidies from sales expenses to CMR deductions).

Adjusted EBITA for China's e-commerce group was RMB 24 billion, down 40% year-on-year, primarily due to investments in instant retail, user experience, and technology; excluding losses from instant retail, EBITA for China's e-commerce group remained stable year-on-year.

3. Instant Retail (Flash Sales) Business

This quarter, instant retail business revenue increased by 57% year-on-year, with overall order volume from January to March being 2.7 times that of the same period last year. Non-catering retail orders were three times those of last year, significantly increasing market share.

Unit economics (UE) continue to optimize, driven by improved logistics efficiency, optimized order mix, and increased average order value (AOV). UE is expected to turn positive in the new fiscal year, with overall profitability achievable at current scale and market share.

Significant business synergies are evident: Instant retail positively impacts new customer acquisition, user engagement, coverage of diverse consumption scenarios, transaction and commercialization growth, and logistics infrastructure improvement for physical goods e-commerce. It particularly drives growth in categories like food and fresh produce, promoting the development of related businesses such as Hema and Tmall Supermarket, serving as a key driver for GMV and CMR growth in physical goods e-commerce this quarter.

02

Performance Overview

In the fourth quarter of FY2026, the group's total revenue was RMB 243.4 billion. Excluding revenue from Sun Art and Intime, like-for-like revenue increased by 11% year-on-year.

GAAP net profit for this quarter was RMB 23.5 billion, up 96% year-on-year, primarily due to increased net gains from changes in fair value of equity investments and the base effect of losses from the disposal of Sun Art and Intime in the same period last year, partially offset by a decline in adjusted EBITA.

Net operating cash inflow was RMB 9.4 billion, while net free cash outflow was RMB 17.3 billion, mainly due to strategic investments in AI business.

As of March 31, 2026, the company held approximately USD 38 billion in net cash, with approximately USD 59 billion in net cash excluding debt maturing in over five years. The balance sheet remains robust, with strong capital market financing capabilities to support strategic investments.

The board approved an annual dividend of USD 1.05 per ADS.

Q&A

Q: What are the contributions from self-developed models (such as QianWen) and third-party models to the ARR related to AI MaaS and applications? How will the recent token price increase impact the profit margins of the MaaS business and cloud business?

A: Currently, revenue from model and application services mainly consists of two parts: API service revenue from the Bailian platform's MaaS and AI native software subscription revenue. The vast majority of this revenue comes from Bailian's MaaS API services. Alibaba Cloud Bailian is an open platform that hosts both self-developed models, third-party open-source models, and third-party closed-source models. The majority of current revenue comes from self-developed models, including QianWen base models, voice models, and video models. Recently, AI applications are shifting from conversational chatbots to Agents, which require completing complex reasoning tasks, driving a sharp rise in customer demand for model reasoning. Although token prices have increased, customer acceptance is very high, with sustained strong demand. Currently, supply cannot fully meet customer needs, with many customers still waiting in line for services. The MaaS business inherently has higher gross margins compared to the IaaS business. At the same time, reasoning technology continues to improve, with quarterly technical optimizations leading to sustained increases in token output per server and per card. As model capabilities continue to strengthen over the next 1-2 years, model prices will remain on an upward trajectory. It is expected that the rapid growth of the MaaS business in the coming quarters will have a very positive impact on the company's gross margins.

Q: How does the company assess the ROI of AI investments, and what management framework is used to balance large AI investments with profit stability?

A: The negative free cash flow this quarter is primarily due to significant AI investments over the past year. The company will remain committed to AI investments over the next two years, as this is a critical window for AI development. First, from a cash flow support perspective: 1) Taobao and Tmall serve as the core sources of the group's operating cash flow, with stable performance. Over the next two years, losses from flash sales will significantly narrow, and AIDC business will turn from losses to profits, with overall operating cash flow from consumer businesses maintaining positive growth. 2) Investments in cloud infrastructure will drive continued accelerated growth in AI cloud revenue, while gross margins in this segment will continue to improve, further enhancing operating cash flow for the cloud business and supporting ongoing cloud infrastructure investments. 3) The company's balance sheet remains very healthy, with approximately USD 38 billion in net cash as of March 31, 2026. Excluding debt maturing in over five years, net cash is approximately USD 58 billion. The company also has strong capital market financing capabilities, meeting strategic development needs through diversified financing methods. Solid financial reserves can fully support AI-related investments. From an investment return perspective, the return path for AI business is very clear. AI business development is similar to manufacturing, requiring the construction of two core facilities: AI training factories and AI reasoning factories, which correspond to AI data center construction. Such infrastructure investments are rigid. Currently, the company's B2B commercialization path is clear, with monetization achievable through cloud IaaS services, MaaS platforms, and AI native software. Currently, no server cards are idle, and the investment return for AI data center construction over the next 3-5 years is highly certain.

Q: What are the drivers for UE improvement in instant retail? Have there been any updates to the previous outlook for instant retail? What changes have there been in the assessment of the current market landscape?

A: After a year of investment, the market share of instant retail has undergone fundamental changes: Overall order volume from January to March 2026 was 2.7 times that of the same period last year, with non-catering retail orders being three times those of last year. Since April 2026, while maintaining order volume, UE has significantly improved due to enhanced logistics efficiency, optimized order mix, and further increases in AOV. The company is confident in achieving positive UE before the end of FY2027. While optimizing UE, the company will continue to enhance user and merchant experiences through innovation, maintaining long-term competitiveness in instant retail. The company is confident in achieving overall profitability in instant retail in the future at current scale and market share levels. This quarter, instant retail has demonstrated a definitive positive impact on physical goods e-commerce, including driving new customer acquisition, promoting user engagement, meeting diverse user consumption scenarios, boosting transactions and commercialization, and improving logistics infrastructure for the overall Taotian platform. It particularly drives growth in categories like food and fresh produce, also promoting the development of instant retail-related businesses such as Hema and Tmall Supermarket. This quarter, physical goods e-commerce has achieved strong growth momentum in both transactions and CMR.

Q: What are Alibaba's advantages in the MaaS field compared to other leading AI platform companies and AI startups in China?

A: The Alibaba Cloud Bailian platform is positioned as an open AI reasoning platform, with the majority of its current revenue generated by Alibaba's self-developed models. Compared to AI startups, Alibaba's core advantage lies in its larger-scale model investments and broader coverage of model types. It will deploy model R&D in multiple areas, including the QianWen base model prioritizing coding capabilities, Wanxiang and Happy House video models, future-oriented world models, and voice models, meeting comprehensive user needs for different model capabilities across various business scenarios. At the same time, Alibaba also believes that AI startups focused on vertical domains possess strong technical capabilities and commercial acumen, serving as partners rather than purely competitors for Alibaba Cloud Bailian.

Q: When will China's AI coding field experience similar rapid growth to that in the US?

A: China's AI coding field is already experiencing rapid growth similar to that in the US. Based on trends from the Alibaba Cloud Bailian platform and feedback from partnered AI startups, from November-December 2025 to May 2026, a significant portion of API demand growth among enterprises was driven by improved AI coding capabilities. AI coding not only replaces software engineers' work but also, when combined with agent operating environments and the integration of Harness engineering and data domains, drives the implementation of complex task agents across various digital work tasks. In both China and the US, this round of AI demand growth is primarily driven by improved AI coding capabilities. It is expected that this field will remain a very important growth trend for the next 2-3 years.

Q: Will the current low willingness to pay for SaaS among Chinese enterprises result in weaker commercialization potential for China's AI coding products compared to their US counterparts?

A: This situation will change in the era of large models. As long as AI model capabilities are strong enough to help users truly complete complex tasks at work, enterprises in both China and the US are willing to pay for corresponding intelligent capabilities. As long as the value created by models exceeds token costs, theoretically, API token demand can be unlimited. The growth of underlying AI demand has long-term certainty. Based on actual data from the Alibaba Cloud Bailian platform, growth has been very rapid, with data from May-June 2026 increasing over 10x compared to November-December 2025. Currently, platform ARR has exceeded RMB 8 billion, with exceeding RMB 10 billion in ARR in the fourth quarter of FY2026 being a highly certain event. This also verifies that Chinese enterprises are willing to pay for intelligent capabilities that help complete real work tasks, with AI coding's commercialization potential not weaker than that in the US.

Q: How does the company prioritize strategies and allocate resources for AI-related businesses in B2B and B2C segments? If enterprise-side development trends better in the future, will more resources be shifted from B2C offerings like the QianWen APP to B2B businesses such as cloud and MaaS?

A: AI is fundamentally a computing paradigm revolution, with the core goal of helping users complete tasks and solve problems. This logic applies to both B2B and B2C businesses. Currently, whether in global or Chinese markets, B2B clients have stronger willingness to pay, and ROI is easier to calculate. Therefore, the company allocates most of its reasoning resources to B2B commercialization. The B2C AI assistant business is still in the stage of cultivating customer acceptance and willingness to pay, requiring a certain investment cycle. However, as technology develops and customer acceptance increases, and products can help users complete more tasks, mature business models will gradually form. Currently, referenceable C-side payment models have emerged overseas. It is expected that there will be significant commercialization progress in the domestic B2C AI assistant field within the next 1-2 years. The company will continue to monitor development opportunities in related areas.

Q: What is the trend for EBT profit margins in Alibaba Cloud's business over the next few quarters? Will there be a similar profit margin expansion trend to overseas peers as business accelerates?

A: Currently, AI technology penetration across industries is still in its early stages. Alibaba Cloud and AI-related businesses' primary goal is to achieve growth, driving user token consumption, increasing market share, and striving to solidify absolute market leadership with growth exceeding the industry average. Profit margins are a secondary goal. In the long run, multiple factors will drive profit margin improvements for Alibaba Cloud: 1) There are physical bottlenecks on the supply side. Over the next 3-5 years, the construction cycle for AI data centers and capacity growth in related hardware such as chips and storage will struggle to match AI demand growth. Meanwhile, under the current supply-demand tight environment, the cost of deploying the same servers this year has increased by over 100% compared to two years ago. The rise in server replacement costs will positively influence cloud service pricing, driving long-term increases in cloud service asset pricing. 2) The MaaS business is a high-gross-margin business, with gross margins significantly higher than traditional IaaS businesses. At the same time, reasoning technology optimizations will continuously improve output per card. Currently, the same servers generate higher revenue and gross margins on the Bailian platform compared to traditional computing services. Future increases in MaaS revenue share will drive overall gross margin improvements. 3) The company possesses full-stack technological advantages. Future mass production of T-Head's self-developed AI chips will provide the most cost-effective reasoning platform domestically, forming strong synergies with the company's model business and further improving gross margins. Considering these factors, it is expected that Alibaba Cloud's gross margins will see significant improvements in the next 1-2 years, with related changes becoming visible in the next 1-2 quarters.

Q: To meet the long-term revenue goals of MaaS and cloud business, what level of capital expenditure does the company need to maintain? Additionally, what is the current penetration rate of T-Head chips in Alibaba Cloud, and how much improvement in the company's gross profit margin can be expected from the increase in their penetration rate?

A: Information related to capital expenditure: The company has set a goal of achieving a tenfold increase in Alibaba Cloud's revenue over the next five years compared to 2022, which requires the scale of data centers to also reach more than ten times that of 2022. Previously, the company announced a three-year plan to invest RMB 380 billion. In light of the goals for the next five years, the overall investment in computing power acquisition will far exceed RMB 380 billion. However, not all computing power will be acquired through self-built capital expenditures. Three approaches will be used to support expansion: first, partially through self-built capital expenditures; second, by acquiring computing power through operational leasing (OPEX); third, as T-Head chip production capacity expands, AI servers equipped with T-Head chips will be sold externally, and computing power centers will be jointly built with data center service providers. Information related to T-Head chips: Currently, the deployment ratio of T-Head chips in Alibaba Cloud is still relatively low. T-Head has achieved full-stack self-research in GPU, CPU, storage, and network chips. The core reason for the low penetration rate is the current limited domestic semiconductor production capacity. As the penetration rate of T-Head's full-stack self-researched chips increases, it will have a very significant positive impact on improving gross profit margins. Currently, there is still a gap between domestic semiconductor processes and advanced overseas levels, with equivalent chips having slightly lower energy efficiency than advanced overseas chips. However, the gross profit margin of mainstream overseas AI chips can reach 60%-80%. Even with deficiencies in performance and power consumption, domestic chips still have significant room for cost-effectiveness improvement. The specific extent of gross profit margin improvement depends on the subsequent production capacity expansion speed and the replacement ratio of existing computing power.

- END -