Alibaba's AI cloud hits US$5B run rate as operating earnings collapse to zero
Alibaba's March quarter 2026 results show AI-related cloud revenue at RMB9B, an annualized run rate of approximately US$5.2B and the 11th consecutive quarter of triple-digit growth.
AI-related cloud revenue reached RMB9B in the quarter, a US$5.2B annualized run rate and 11th consecutive quarter of triple-digit YoY growth.
Cloud Intelligence Group external revenue accelerated to 40% YoY, the fastest growth rate among China’s three major cloud providers.
Adjusted EBITA collapsed 84% YoY to RMB5.1B, with non-GAAP net income falling to RMB86M from RMB29.8B a year ago.
Free cash flow turned negative at -RMB17.3B, compared to +RMB3.7B in the same quarter of 2025.
Quick commerce revenue grew 57% YoY to RMB20B, scaling fast and consuming capital at pace.
Alibaba reported March quarter 2026 results on May 13. The headline revenue figure of RMB243.4B represents a 3% increase YoY. The number is misleading. Excluding divested businesses Sun Art and Intime, Alibaba grew 11% on a like-for-like basis. Reported revenue is depressed by the absence of assets Alibaba chose to shed.
The more consequential numbers sit elsewhere. Cloud Intelligence Group AI-related products generated RMB8.97B in revenue, an annualized run rate of approximately US$5.2B. The segment posted its 11th consecutive quarter of triple-digit YoY growth. At the same time, adjusted EBITA fell 84% YoY to RMB5.1B.
Free cash flow turned to a RMB17.3B outflow from a RMB3.7B inflow one year earlier. Non-GAAP net income, which strips out investment gains and non-cash items to approximate operational earnings power, reached RMB86M, effectively zero, against RMB29.8B in the prior year.
Alibaba is funding three simultaneous bets: AI model development and cloud infrastructure, quick commerce market share, and Qwen consumer app user acquisition.
The quarterly cost of that combination is now visible. The strategic question is whether one bet (AI cloud) justifies the cost of running all three at once.
Cloud is the proof point, and the scale is now globally meaningful
Cloud Intelligence Group is the only segment where Alibaba’s AI investment is generating commercial validation this quarter. Revenue reached RMB41.6B (US$6.0B), up 38% YoY. External customer revenue accelerated to 40% YoY growth, driven by AI-related product adoption within the public cloud.
AI-related product revenue reached RMB8.97B, representing approximately 30% of Cloud Intelligence Group revenue. Annualized, the AI run rate sits at approximately US$5.2B.
According to Alibaba’s earnings release, the segment posted its 11th consecutive quarter of triple-digit YoY growth in AI-related cloud products. Compounding triple-digit growth across 11 quarters implies the base level in early 2023 was negligible. The current quarterly figure is now large enough to reshape competitive dynamics.
Global calibration matters. Microsoft disclosed an AI revenue run rate of approximately US$37B in late April, growing 123% YoY, according to Microsoft’s Q3 FY2026 release. AWS reported an AI run rate exceeding US$15B with triple-digit growth.
Google Cloud crossed US$20B in quarterly revenue at 63% YoY growth, with AI as the primary driver. Alibaba’s AI cloud business at approximately US$5.2B annualized is smaller than each of these. The growth rate is comparable to the hyperscalers, and the trajectory of 11 consecutive triple-digit quarters has no peer.
Within China, Alibaba Cloud is the unambiguous leader. Huawei Cloud’s external cloud revenue actually declined 3.5% in 2025 to RMB32.2B, according to Omdia, even as the Chinese cloud infrastructure market grew 26% YoY in Q4 2025. Alibaba is capturing share in a market where the #2 player is shrinking.
Cloud Intelligence Group adjusted EBITA grew 57% YoY to RMB3.8B. The segment is growing and becoming more profitable at the same time. EBITA margin expanded meaningfully as revenue scale began absorbing fixed infrastructure costs. For fiscal year 2026, Cloud Intelligence Group revenue reached RMB158.1B, up 34% YoY, with adjusted EBITA growing 35% to RMB14.3B.
The Model Studio platform, Alibaba’s MaaS offering, saw its customer base expand 8x YoY as of March 2026. The platform hosts Qwen3.6-Plus, enterprise agents including Wukong and Meoo, and industry-specific models. The breadth of that offering is a direct response to enterprise demand for flexible deployment across model sizes and use cases.
Qwen’s positioning: reasoning, coding, and multimodal expansion
Alibaba launched Qwen3.6-Plus during the quarter. The model delivers what the company describes as state-of-the-art results across front-end web development and complex repository-level tasks.
It features a native context window of up to 1 million tokens, enhanced multimodal reasoning, and improved stability. The context length positions it competitively for long-document enterprise workflows.
Complementing the core Qwen family, Alibaba introduced two specialized models. HappyOyster is a world model enabling real-time creation and interaction. HappyHorse handles video generation.
Both are in phased commercial rollout. The multimodal expansion reflects Alibaba’s assessment that enterprise AI applications will increasingly require vision, video, and world-simulation capabilities alongside text.
T-Head silicon: a strategic hedge against domestic chip dependency
The competitive context for Alibaba’s chip strategy has shifted dramatically. Nvidia’s share of the Chinese AI accelerator market has collapsed to zero, confirmed publicly by CEO Jensen Huang and reported by Tom’s Hardware.
Huawei has filled the vacuum. Its Ascend chip business is targeting US$12B in revenue in 2026, up from US$7.5B in 2025, with major Chinese tech companies including Alibaba, Tencent, and ByteDance among the buyers.
The dependency risk is real. DeepSeek V4 was optimized specifically for Huawei’s Ascend architecture, with Huawei engineers collaborating directly on kernel-level integration.
Alibaba Cloud deployed V4 inference services within hours of release. Morgan Stanley projects Chinese suppliers will capture 86% of China’s AI chip market by 2030. Alibaba is exposed to single-vendor concentration risk on its most critical input.
T-Head, Alibaba’s proprietary chip design subsidiary, is the strategic hedge. Over 100,000 Zhenwu PPUs have been deployed on Alibaba Cloud’s public cloud platform. More than 30 automakers and autonomous driving companies are using these chips for intelligent driving R&D.
The Zhenwu chips, combined with Alibaba Cloud infrastructure and Qwen models, form a vertically integrated stack that reduces both inference costs and reliance on Huawei silicon. Proprietary silicon is a medium-term competitive advantage. The strategic value has increased materially in the past 12 months as Nvidia’s China business closed and Huawei’s domestic dominance solidified.
The profitability collapse: two bets running simultaneously
The 84% collapse in adjusted EBITA reflects two distinct investments running at scale in the same quarter. Separating them is analytically important.
Quick commerce is the first. Revenue from quick commerce reached RMB20B, up 57% YoY, driven by Taobao Instant Commerce, which launched in April 2025. The growth rate is high. So is the burn. Quick commerce requires logistics subsidies, delivery network investment, and merchant incentives. The segment is not yet profitable on a standalone basis.
Qwen consumer app user acquisition is the second. Sales and marketing expenses rose 47% YoY to RMB53.4B, reaching 21.9% of revenue from 15.3% a year ago. A significant portion of this increase was allocated to acquiring users for the consumer-facing Qwen app.
Alibaba integrated Taobao and Tmall e-commerce directly into the Qwen app during the quarter, creating a closed-loop AI shopping experience. The consumer Qwen app is a direct response to competitors including ByteDance and Baidu who have deployed conversational AI surfaces with large user bases.
The China E-commerce Group adjusted EBITA fell 40% YoY to RMB24B. The e-commerce core is profitable and growing. Customer management revenue grew 8% on a like-for-like basis. Its earnings are being consumed to fund the two expansion bets.
The All Others segment posted an adjusted EBITA loss of RMB21.2B, compared to a RMB3.4B loss in the same quarter of 2025. The sharp widening reflects investment in technology businesses and Qwen app user acquisition, which are classified within this segment.
Cash position and the funding gap
Free cash flow turned to a -RMB17.3B outflow in the March quarter, against a +RMB3.7B inflow a year earlier. For the full fiscal year 2026, free cash flow was a -RMB46.6B outflow against +RMB73.9B in fiscal year 2025. The swing is RMB120.5B across two annual periods.
Capital expenditure for fiscal year 2026 reached RMB126.1B, up from RMB84.3B the prior year. The bulk of that increase reflects data center and cloud infrastructure investment.
Alibaba retains substantial liquidity. Cash and other liquid investments stood at RMB520.8B as of March 31, 2026. The balance sheet can absorb continued investment at current pace for multiple years without structural stress.
The board declared an annual regular cash dividend of US$1.05 per ADS for fiscal year 2026, approximately US$2.5B in aggregate. Maintaining dividends while generating negative free cash flow signals management’s confidence in the liquidity position. It also signals a desire to preserve institutional shareholder support during the investment cycle.
The strategic logic of running three bets simultaneously
A reasonable investor reading the financials might conclude Alibaba is over-extended. Three concurrent investment cycles are running at scale: AI cloud infrastructure, quick commerce expansion, and Qwen consumer user acquisition. Each could absorb full corporate attention. Running all three at once concentrates risk.
The strategic logic is defensible. Each bet defends a different flank. AI cloud is the offensive bet, where Alibaba’s category leadership translates into compounding revenue.
Quick commerce defends the China e-commerce core against Meituan, which has been eroding share through Instashopping. Qwen consumer app responds to ByteDance’s Doubao and Baidu’s Ernie Bot, which are building large consumer AI surfaces that could become alternative distribution channels for AI-mediated commerce.
The temporal logic also matters. Each investment cycle has a finite peak. Quick commerce burn rate will fall as unit economics improve. Management explicitly highlighted improving order economics and rising average order value this quarter.
Qwen consumer app user acquisition is most expensive in the land-grab phase, before competitive equilibrium. AI cloud capital expenditure follows a multi-year infrastructure cycle. Running them concurrently compresses the pain period rather than extending it.
The risk is execution complexity. Three simultaneous bets demand attention from senior management and capital from a single balance sheet. If any one underperforms, the strategic logic fragments. The investor question is whether Alibaba can sustain disciplined execution across all three until at least one generates positive cash contribution.
Risks and signals to monitor
The financial trajectory creates near-term investor risk on three fronts.
First, the quick commerce bet carries execution risk that cloud does not. Quick commerce requires physical logistics density, merchant network effects, and sustained subsidy capacity. Ele.me and Taobao Instant Commerce are competing against Meituan’s established network in a market where logistics unit economics are hard to compress quickly.
Second, Qwen’s consumer traction is self-reported and unverified by independent data. The 8x growth in Model Studio customers is a meaningful enterprise signal. The consumer app user base is not disclosed with comparable specificity.
Third, the free cash flow outflow trajectory implies Alibaba is in a period of maximum investment intensity.Management has not signalled when the investment cycle peaks. Investors who require near-term cash generation have limited visibility on the timeline to positive free cash flow resumption.
Non-GAAP net income of RMB86M, effectively zero, means Alibaba’s operational earnings power is temporarily negligible. Investment gains inflated the GAAP net income figure to RMB23.5B. Strip those out and the operational business generated almost nothing after expenses.
Cloud is what matters. The rest is noise around the central bet.
Alibaba entered fiscal year 2026 needing to prove its AI investment thesis was real, scalable, and ultimately profitable. The March quarter answers the first two questions with hard data.
AI-related cloud revenue at a US$5.2B annualized run rate makes Alibaba a globally relevant AI cloud business. The 11th consecutive quarter of triple-digit growth and 8x expansion in Model Studio customers confirm enterprise demand is converting into cloud billings.
The profitability question remains open. The answer is increasingly visible in the segment data. Cloud Intelligence Group adjusted EBITA grew 57% YoY this quarter, with margin expansion as fixed infrastructure costs absorb across a larger revenue base.
The cloud business is profitable today and becoming more profitable with scale. The group-level earnings collapse is a consequence of quick commerce and Qwen consumer spending, not of the AI cloud business.
For institutional readers, the analytical view is this: Alibaba is the only Chinese tech company with a credible path to a top-tier global AI cloud business at scale. The cost of acquiring that position is visible and painful. The position itself is becoming durable.
Investors who require near-term cash generation will avoid the name through fiscal year 2027. Investors with multi-year horizons are watching the only Chinese company where AI cloud revenue is large enough to matter on the global comparison set. The next two quarters will reveal whether quick commerce unit economics improve fast enough to ease the cash flow pressure, allowing the cloud thesis to deliver without distraction.