As we mentioned this morning, we’re still not sure what the market’s BABA narrative is these days. Is bad EBITDA okay because the company is spending aggressively and AI is growing triple digits? We honestly don’t know. But if you didn’t have time to sit through the conference call, we did—here are the five questions and answers we thought mattered most. We do suggest you read this one as Eddie really took the time to outline the spend, the payback and the future AI story at Alibaba!
BABA · Q1 FY2027 — Top 5 Earnings Q&A
2026-08-20
Total Revenue: RMB269B (+9% YoY) · ▲ est RMB268.34B (+0.2%)
Cloud External Revenue Growth: +45% YoY · est ~45% growth (in line)
Cloud Adjusted EBITA Margin: 12% (segment); 11.6% (mgmt. commentary)
AI-Related Product Revenue (Quarterly): RMB12.4B; ARR RMB49.5B (~$7.3B)
Total Adjusted EBITA: RMB27.3B (-30% YoY)
CapEx: RMB67.7B (vs. RMB18.8B FCF outflow prior year)
Q1 · Unidentified Participant · Citigroup
CapEx Surge, Three-Year Budget, ROIC Framework
Q: Can you comment on the drivers of the significant CapEx increase this quarter and what the trend looks like for coming quarters? Are there any updates to the existing three-year RMB380 billion CapEx budget? And can you provide a breakdown of that CapEx allocation across different use cases like training? What is management’s expected return on invested capital for these investments?
Eddie Wu, Chief Executive Officer: This is an important question and I want to explain our AI business model and CapEx expectations broadly. When we announced our three-year capital investment plan in February, we set total investment at RMB380 billion, and as of the end of the June quarter this year, we had already spent RMB190 billion, with progress broadly in line with our expectations. This quarter’s spending of RMB67.1 billion is somewhat higher, but that primarily reflects volatility in equipment delivery schedules — hardware deliveries follow different procurement cycles and there can be fluctuations in the cadence and pace of deliveries, so it is not evenly distributed across quarters. At the same time, we increased CPU procurement this quarter because we are witnessing a substantial surge in demand driven by the agent-centric era, and rising prices for semiconductor components have also contributed. So I don’t think you should take this quarter’s spending and multiply it by four to arrive at an annualized figure, or expect a steady linear progression. This is an asset-heavy business model — whether you monetize through software subscriptions, API calls, models as a service, training, or inference, in all of those respects you need compute centers to run and to monetize. That means we need to invest upfront in order to grow and monetize across all of those areas, which is why beginning in 2025 we entered a heavy investment cycle in hardware. On ROIC, there is an industry consensus that the current shortage in AI compute will not be resolved until at least 2030, which gives us high certainty in our AI compute investments. Based on our average gross margins today, we can roughly break even on AI-related CapEx in three years, and as average gross margins continue to rise, we expect to shorten that payback period to 2.5 years. Following the three-year payback period, these AI assets can achieve very positive and robust cash flow — a V100 purchased in 2018 is still running at full capacity today as an example. We have three levers to further enhance gross margin and ROIC: first, continuing to develop state-of-the-art models and expand higher-margin mass businesses while adapting the product mix to achieve higher gross margin across the portfolio — we’ve already seen a 4.4 percentage-point increase in Alibaba Cloud’s overall segment profitability, bringing it this quarter to 11.6%, which is initial validation of that thesis; second, deploying our own proprietary T-Head chips — since chips and storage are the most expensive components in AI data centers, as proprietary chips account for an increasing proportion of total chips and replace commercially procured chips, we can expect substantially higher gross margins and profitability; third, monetizing through co-building data centers with partners and pre-charging or receiving prepayments for compute-based services, which further enhances ROIC. Through these three methods we can shorten the payback period to 2.5 years or even two years. At our current AI product gross margin level and assuming a three-year payback period, theoretically, keeping growth below 33% would already enable positive cash flow — but that is not our strategic choice given that AI is still very early stage. As proprietary chip substitution rates increase and product gross margins improve, we expect the payback period to shorten to 2.5 years or less, at which point we can pursue growth of over 40% while also maintaining positive cash flow, and that is our long-term strategic direction.
