NextFin News - Alibaba Group's quarterly profit collapsed 76% in the three months ended June 30, even as revenue rose 9% and demand for artificial-intelligence cloud services accelerated. The divergence - a 75% surge in capital spending to 67.68 billion yuan ($9.98 billion) against a 45% jump in AI cloud revenue - is forcing investors to ask the question now echoing from Wall Street to Hangzhou: how much of the AI demand is real, and how much is money circling back to the same balance sheets that fund it?
The June-quarter results, released Thursday, landed as a deliberate stress test of that thesis. Net income fell to 10.54 billion yuan ($1.55 billion) from a year earlier, missing the 26.98 billion yuan analysts had expected. Revenue came in at 268.95 billion yuan ($39.64 billion), just ahead of the 268.88 billion yuan consensus compiled by LSEG. Free cash flow swung to an outflow of 44.67 billion yuan ($6.58 billion), more than double the 18.82 billion yuan outflow in the year-ago quarter. Yet within that cash squeeze sits the growth engine management is betting the company on: revenue from AI cloud and compute services rose 45% to 48.44 billion yuan ($7.14 billion), with AI-related product revenue of 12.38 billion yuan posting triple-digit growth for the 12th consecutive quarter.
The stakes rose beyond a single quarter last week, when CEO Eddie Wu committed Alibaba to spending at least 380 billion yuan ($53 billion) on AI and cloud infrastructure over the next three years - more than the company spent on AI and cloud across the entire prior decade. The shares fell 5.3% on the results, giving the skeptics their evidence while the bulls pointed to the cloud acceleration. The central tension is no longer whether AI is growing at Alibaba. It is whether the growth is being financed in a circle, and whether that circle can hold when the funding tightens.
The Spending Math: Profit and Cash Are Being Converted Into Capacity
The first-order read of the quarter is mechanical: Alibaba is converting current earnings and cash into future compute. Capital expenditure of 67.68 billion yuan in a single quarter is not an incremental budget line; it is a strategic reallocation. Against group revenue of 268.95 billion yuan, quarterly capex now runs at roughly 25% of sales - a ratio that belongs to a company building infrastructure for a decade, not defending a quarter.
The margin arithmetic shows the cost. Profit fell 76% while revenue rose 9%, meaning the incremental margin on that growth was deeply negative in the reported period. The AI Labs and Applications segment - housing model development, the Qwen consumer app, and enterprise agents - posted an adjusted EBITA loss of 13.86 billion yuan ($2.04 billion), wider than the 3.22 billion yuan loss a year earlier, as inference costs tied to the Qwen app climbed. Management's defense is that the spending is front-loaded and the revenue follows with a lag: depreciation on data-center assets lands before the customers fully occupy the racks. That is the classic infrastructure story, and it is coherent. It is also the same story every capex-heavy cycle tells on the way up.
What makes this quarter different from a normal investment cycle is the concentration of the bet. Alibaba Cloud's external revenue grew 45%, and AI-related product revenue is compounding at triple digits - an annualized run rate of roughly 50 billion yuan. In other words, the AI business is now large enough that a meaningful deceleration would be visible at the group level within two quarters. The bet is not hidden in a footnote.
"We delivered a strong quarter, driven by the improving commercialization of our full-stack AI capabilities," Wu said in a company statement, describing AI as the company's growth engine and committing to expand computing capacity to meet customer demand.
The cash-flow statement is where the rubber meets that ambition. A quarterly free-cash-flow outflow of 44.67 billion yuan means the buildout is consuming far more cash than the core commerce engine can throw off in the period. That is sustainable only if the capacity gets filled by paying customers whose demand does not depend on Alibaba's own financing. The core e-commerce business, still the cash cow, showed strain: China e-commerce revenue fell 8% to 110.90 billion yuan, even as China quick commerce surged 45% to 53.30 billion yuan. The company is spending the old business to buy the new one.
The Circular Question: Who Is Actually Buying the Compute?
This is where the Hangzhou story meets a global debate. Across the AI complex, the pattern is now familiar: chipmakers invest in AI labs, cloud providers fund AI startups, and those same companies sign multi-year compute contracts that show up as revenue. OpenAI agreed to purchase hundreds of billions of dollars of cloud capacity; Nvidia poured tens of billions into the very labs buying its GPUs. Critics call it circular financing - revenue that exists only because the buyer was funded by the seller's ecosystem.
Alibaba sits in the same structural position, with two channels for the concern. First, the company has invested across China's AI startup layer - model developers, application builders, and inference companies that are natural cloud customers. Second, its own AI models and products consume internal capacity while the external cloud business sells to the same ecosystem. If a meaningful share of that 45% cloud growth is coming from companies whose funding traces back to Alibaba's balance sheet, or to investors encouraged by Alibaba's commitment, then the revenue is not independent validation of demand. It is a form of self-reinforcing demand dressed as growth.
The concern is not that the transactions are fraudulent. They are not. The concern is that they are procyclical in the worst way: they amplify the up-cycle by creating the appearance of demand, and they accelerate the down-cycle when funding tightens. Jim Chanos, the veteran investor who has warned about the broader AI buildout, put the risk plainly: if AI demand in 2027 or 2028 falls short of today's projections, "you could see orders begin to be canceled." A circular revenue base is the first to cancel, because the funding that created it is the first to disappear.
There is also a second-order channel that a single-quarter earnings miss does not capture. Alibaba's capex is a revenue line for the semiconductor and equipment supply chain - the same companies whose earnings have carried global indexes higher. If Alibaba's spending slows because the circular demand proves illusory, the hit propagates outward: chip suppliers lose a customer, their earnings miss, their own capex plans shrink, and the AI revenue that depended on their expansion evaporates in turn. This is the cross-asset transmission that turns a China earnings story into a global AI-capex story.
Cyclical or Structural: Deciding Which Force Is Dominant
The cyclical-vs-structural call determines everything about how to read this quarter, and the honest answer is that both forces are present - but they operate on different horizons, and conflating them is the most common analytical error.
The cyclical leg is the capex cycle itself. History offers three clear analogs. In the late-1990s fiber-optic buildout, telecom companies laid cable on the assumption that internet traffic would grow exponentially - which it did. The mistake was not the demand forecast; it was the timing and the number of competitors racing to serve it. Capacity arrived before paying customers, margins collapsed, and the builders, not the users, lost everything. In the mid-2010s, US shale drillers borrowed heavily to expand production on the thesis that oil demand would keep rising. It did - but prices fell because supply grew faster, and the most leveraged producers were wiped out. In the 2020-2021 container-shipping boom, carriers ordered record fleets on visible demand; when the post-pandemic normalization hit, freight rates fell roughly 80% from peak and order books turned into liabilities.
The common pattern is not wrong demand. It is right demand, wrong price, wrong timing, and too many competitors sharing the same thesis. Alibaba's AI capex cycle fits this pattern in the near term: the demand for compute is real, but the returns depend on how much capacity China builds relative to what customers will pay for, and on whether the customers are creditworthy outside the ecosystem that funds them.
The structural leg is different, and it is why dismissing the whole buildout as a bubble is equally wrong. Three regime changes support the structural case. First, AI inference is not a one-time migration like moving email to the cloud; it is a continuous, compounding workload that grows with usage, which means capacity utilization can keep rising on the same hardware base. Second, China's push for domestic AI capability is a policy directive, not a market sentiment - the state will keep funding the stack even when private returns lag, which raises the floor under demand. Third, Alibaba's position as China's largest cloud provider gives it a distribution moat: once models are deployed on its infrastructure, switching costs for customers are high.
The correct read, then, is a structural trend wrapped in a cyclical financing wave. The structural trend - AI adoption in China - will outlast the cycle. The financing wave - the circular deals, the leverage, the multiple expansion - will not. Investors who treat the whole thing as structural will overpay at the peak. Investors who treat the whole thing as a bubble will miss the trend that survives it.
The Counter-Thesis: Why the Skeptics Could Be Wrong
The strongest case against the circular-financing concern is that Alibaba's cloud growth is being driven by customers with independent balance sheets and urgent, non-discretionary demand. The company has pointed to enterprise adoption, government AI initiatives, and large consumer-facing applications - none of which require Alibaba's funding to exist. If the 45% cloud growth is coming from state-owned enterprises, banks, and established internet companies migrating real workloads, then the circularity charge is a category error: it mistakes an ecosystem for a conspiracy.
There is also a valuation argument. Alibaba trades at roughly 19 times trailing earnings with a price-to-earnings-growth ratio near 0.52 - a discount to US peers running the same AI-cloud-commerce playbook. The market, in this read, has already priced in deep skepticism; the risk-reward favors the builders, not the short-sellers.
And there is a third rebuttal, the most important one: circularity is a spectrum, not a binary. Even in the United States, where the OpenAI-Microsoft and Nvidia-Anthropic relationships draw the most scrutiny, the underlying demand for inference is measurable in actual usage - tokens processed, active users, enterprise contracts. If Alibaba's AI revenue is similarly backed by real consumption, then the financing structure is a detail about who provided the capital, not a verdict on whether the demand exists. The falsifying test is not whether Alibaba invested in a customer. It is whether that customer's usage persists when the investment does not.
Even so, the counter-thesis carries one vulnerability that no multiple can hide: duration. A circular revenue base can look identical to a real one for quarters at a time. The difference reveals itself only at the turning point, when funding tightens and the customers who existed on paper stop renewing. By then, the capacity is already built and the depreciation is already due.
What Comes Next: Scenarios and the Signal That Would Break the Thesis
The forward look splits cleanly by time horizon. In the short term - the next two quarters - sentiment will dominate. The stock has already given back part of its post-announcement rally, and every capex line will be read as evidence for one side or the other. Liquidity flows, not fundamentals, will set the price.
Over the medium term - the next four to six quarters - the question is utilization. The base case is that Alibaba fills a large share of the new capacity as Chinese enterprises and the state absorb the compute, keeping cloud revenue growth above 30% and allowing margins to stabilize as depreciation rolls through. The upside case is that AI application adoption in China accelerates faster than expected, pushing utilization high enough that the 380 billion yuan pledge starts generating operating leverage before the cycle turns. The downside case is that a meaningful portion of cloud demand proves tied to funded startups, and as venture and state funding slow, orders cancel or fail to renew - the Chanos scenario - leaving Alibaba with stranded capacity and no earnings recovery to show for the spend.
In the long term, the structural case either validates or fails on one metric: whether AI becomes a durable, high-margin revenue stream for Alibaba Cloud rather than a capex sink. If China's AI ecosystem matures into paying, independent customers, today's spending will look like the Amazon Web Services buildout in retrospect - painful, dilutive, and obviously right. If it does not, it will look like the fiber-optic overbuild - right thesis, wrong economics.
The falsifying signal is specific and observable: watch Alibaba's cloud revenue growth and customer funding in the next two quarterly disclosures. If external cloud revenue growth decelerates below 25% year over year while capex remains above 20% of revenue, and management cannot show that major cloud customers have independent, non-Alibaba funding sources, the circular-financing thesis moves from concern to confirmation. Conversely, if cloud growth holds above 35% with expanding margins and disclosed customer diversification, the skeptics are wrong.
The market has priced Alibaba as a cheap e-commerce company with a speculative AI option. The next year will decide whether that option is free - or whether the company is paying for it with earnings that will not come back.
The real story of Alibaba's quarter is not the 76% profit drop. It is that China's AI boom is being financed on faith that the customers of today will still be paying tomorrow - and faith, unlike depreciation, does not appear on the cash-flow statement.
Explore more exclusive insights at nextfin.ai.
