Alibaba's $10 Billion Quarter Fuels the Circular AI Debate

Alibaba's profit plunged after the company lifted quarterly capital spending to almost $10 billion on artificial intelligence, a ramp discussed on Bloomberg Tech on August 20, 2026 alongside Meta's emergence as one of Microsoft's largest AI customers.

Alibaba (BABA) has done what every large-cap technology company building artificial intelligence infrastructure eventually has to do: it has let the spending land on the income statement. The company's profit plunged after it raised quarterly capital expenditure to almost $10 billion for AI, a shift dissected by Ed Ludlow on the August 20, 2026 edition of Bloomberg Technology.

Investors did not punish it. Alibaba's shares closed at 130.53, up 1.26% on the day, after ranging between 121.88 and 130.62 — a spread that tells you the market spent the session arguing with itself before settling near the high. The prior close was 128.90. That is an unusual reaction to a profit collapse, and it is the single most interesting fact in the day's tape.

Why a profit plunge produced a share price gain

Capital expenditure of nearly $10 billion in a single quarter does not vanish. It becomes servers, data centre shells, power contracts and networking gear, and it becomes depreciation that grinds against earnings for years. When a company steps its quarterly capex up sharply, the near-term arithmetic is brutal and entirely predictable: costs arrive before revenue does.

The market appears to have decided that Alibaba's earnings decline is the price of admission rather than evidence of deterioration. That is a bet on the cloud and AI business converting spend into contracted, recurring demand fast enough to justify the depreciation schedule. It is the same bet U.S. investors have been making on American hyperscalers for two years running, and it is a bet that gets re-tested every quarter.

The intraday range is the tell. A move from 121.88 at the low to 130.62 at the high is roughly a 7.2% swing within a single session — the signature of a stock where two credible interpretations of the same numbers are competing in real time.

What 'circular AI financing' actually describes

The phrase now attached to nearly every AI earnings cycle deserves a plain definition. Circular financing describes an arrangement in which money flows between companies in a loop: a chip vendor or cloud provider invests in, lends to, or takes equity in a customer, and that customer then spends the proceeds buying the vendor's product. Revenue is recognised. Demand looks real. But some portion of it originated inside the same cluster of balance sheets rather than from an outside buyer.

The concern is not that such deals are improper. It is that they make demand look more diversified and more durable than it is, and that they can unwind quickly if the funding leg stops. Analysts watching AI infrastructure are increasingly trying to separate the share of cloud and chip revenue that comes from genuinely external customers from the share that traces back to a partner, an investee, or a vendor-financed arrangement.

Alibaba's situation is not the same thing — a company spending its own cash on its own capacity is straightforwardly funding itself. But the debate matters to how the spend is valued, because the multiple investors will pay for AI-driven cloud growth depends heavily on whether they believe the demand behind it is real and repeatable.

Meta buying from Microsoft rearranges the map

The second thread from the same segment may matter more over time. Meta (META) has quietly become one of the largest AI customers of Microsoft (MSFT). Read that slowly. Meta operates one of the world's largest private computing fleets and has been among the most aggressive builders of its own AI infrastructure. Microsoft is the cloud provider most closely identified with frontier model compute.

A company of Meta's scale buying capacity from a peer rather than exclusively building it tells you something about the constraint. When the binding limit is power, land, transformers and delivery schedules rather than capital, even the largest self-builders rent. It also means the boundary between hyperscaler and hyperscaler customer is dissolving. The same firms are simultaneously each other's suppliers, customers and competitors — which is precisely the structure that makes the circular financing question hard to answer from the outside.

Neither stock moved much on the day. Meta closed at 545.83, down 0.04% from a prior close of 546.03, with a day range of 539.31 to 549.99. Microsoft closed at 481.15, down 0.65% from 484.31, ranging 479.50 to 484.16. Both traded quietly against a soft broad market: the S&P 500 tracker (NYSEARCA: SPY) closed at $762.60, down 0.84%, and the Nasdaq 100 tracker (NASDAQ: QQQ) at $710.93, down 0.72%. The Dow 30 tracker (NYSEARCA: DIA) fell hardest, down 1.27% to $527.51. All figures as of the 20:00 GMT close on August 20, 2026.

Hypersonics as the other capital story

The same programme covered defense technology startup Castelion and its plans to increase production of hypersonic missile systems. It sits oddly next to AI capex until you notice the common thread: both are capital-intensive manufacturing bets where the constraint is physical throughput, not software. Castelion is private, and the lead gives no financial detail, so there is no valuation to discuss. What it signals is where venture capital is currently willing to underwrite factories rather than applications — defense hardware and AI compute are absorbing the money that once went to consumer software.

What to track from here

Three things will determine whether Alibaba's spend reads as investment or as a mistake. First, whether cloud revenue growth accelerates enough in coming quarters to outrun the depreciation now entering the cost base. Second, whether the capex run rate stabilises near the current level or keeps climbing — a one-off step up is a plan, a continuously rising number is an arms race. Third, whether the company can point to external, contracted demand rather than internal consumption.

For the American side of the trade, the Meta-Microsoft relationship is the one to watch in disclosure. If large cloud providers begin naming customer concentration, or if analysts force the issue, the market will finally be able to price how much of AI demand is genuinely exogenous. Until then, the circular financing debate stays exactly where it is: a legitimate question that no one outside the companies can settle with published numbers.

The reaction in Alibaha's shares suggests investors are, for now, extending the benefit of the doubt. That patience has a shelf life, and it is measured in quarters.

Key facts

• Alibaba (BABA) last close: 130.53, +1.26% as of 20:00 GMT, Aug 20, 2026
• Quarterly AI capex: Almost $10 billion, up from prior levels
• Meta (META) / Microsoft (MSFT) close: 545.83 (-0.04%) and 481.15 (-0.65%)
• Broad market: S&P 500 ETF $762.60 (-0.84%); Nasdaq 100 ETF $710.93 (-0.72%)

Frequently asked questions

How much did Alibaba spend on AI capital expenditure?

Alibaba raised its quarterly capital expenditure to almost $10 billion for artificial intelligence, according to reporting on the August 20, 2026 edition of Bloomberg Technology. That step up in spending on data centre capacity, servers and related infrastructure was the direct cause of the sharp decline in the company's reported profit for the period.

Why did Alibaba shares rise despite the profit plunge?

Alibaba closed at 130.53, up 1.26% on the day, after a wide intraday range of 121.88 to 130.62. The gain suggests investors treated the earnings decline as the expected cost of building AI capacity rather than as evidence of a deteriorating business, betting that cloud and AI revenue will eventually outrun the new depreciation load.

What is circular AI financing?

Circular AI financing describes arrangements where money loops between related companies: a chip maker or cloud provider invests in, lends to, or takes equity in a customer, and that customer then buys the provider's product. The revenue is real accounting revenue, but critics argue it can make underlying demand appear more diversified and durable than it actually is.

Why is Meta buying AI compute from Microsoft?

Meta has quietly become one of Microsoft's largest AI customers despite operating one of the world's biggest private computing fleets. When the binding constraint on AI infrastructure is power, land, equipment and construction schedules rather than capital, even the largest self-builders rent capacity from peers to meet near-term demand.

How did Meta and Microsoft shares perform that day?

Both were roughly flat. Meta closed at 545.83, down 0.04% from a prior close of 546.03, with a day range of 539.31 to 549.99. Microsoft closed at 481.15, down 0.65% from 484.31, ranging between 479.50 and 484.16. Both figures are as of the 20:00 GMT close on August 20, 2026.

What is Castelion?

Castelion is a defense technology startup that plans to boost production of its hypersonic missile systems, a plan covered on the same Bloomberg Technology segment. The company is private and no financial details were disclosed in the report. Its expansion reflects venture capital appetite for capital-intensive hardware manufacturing alongside AI compute.

Sources

• Alibaba's AI Spending Spree, Concerns of Circular AI Financing | Bloomberg Tech 8/20/2026 — Bloomberg Technology