The quarterly reports published between 21 and 29 October 2026 settle one debate and open another. Spending on AI infrastructure is not slowing: at the four largest US cloud and platform companies it roughly doubled year on year. What has changed is how it is paid for: three of the four reported negative free cash flow, spending more on capital projects than their operations generated. Every figure below comes from the companies’ filings with the US Securities and Exchange Commission unless stated otherwise, checked on 31 October 2026.

The numbers, company by company

Alphabet. Its third-quarter release shows purchases of property and equipment of $51.4bn, against $24.0bn a year earlier, and free cash flow of minus $9.1bn for the quarter. Revenue rose 25% to $127.6bn and Google Cloud revenue more than doubled to $30.7bn. According to CNBC, the company kept its 2026 capital spending guidance at $195bn to $205bn and said spending would “increase significantly in 2027”.

Microsoft. Cash additions to property and equipment were $39.4bn, against $19.4bn a year earlier, in its fiscal first-quarter release. Unlike the others, it stayed cash-positive: operating cash flow was $58.0bn. Azure revenue grew 48%, and commercial remaining performance obligations, contracted revenue not yet recognised, reached $677bn, up 73%. On the earnings call, the chief financial officer said capital spending in the current quarter would be “over $50 billion”.

Meta. Capital expenditure including finance-lease principal was $34.9bn and free cash flow was minus $2.1bn in its third-quarter release, published on 21 October. Operating margin fell to 20% from 40%, partly because of a $9.89bn charge for a settlement with state attorneys general. Meta narrowed its 2026 capital spending range to $135bn to $145bn.

Amazon. Purchases of property and equipment were $77.3bn in the quarter, against $35.1bn a year earlier, according to its third-quarter release. Free cash flow over the past twelve months was an outflow of $42.4bn, compared with an inflow of $14.8bn a year before. AWS revenue grew 39.3% to $46.0bn. Amazon said it still expects 2026 cash capital expenditure of about $220bn.

One detail deserves a question on the next call. Amazon’s nine-month spending is already $175.7bn. Reaching about $220bn for the year implies roughly $44bn in the fourth quarter (our arithmetic), far below the $77.3bn just spent. Either the guidance is conservative, timing is lumpy, or spending really will slow sharply. The filing does not say which.

Where the money lands

The spending shows up as revenue at suppliers. Nvidia’s second-quarter fiscal 2027 results, for the quarter to 26 July 2026, showed revenue of $96.2bn, up 106%, with data centre revenue of $89.0bn. Oracle, which builds capacity for AI customers, reported remaining performance obligations of $664bn in its fiscal first quarter, alongside negative free cash flow of $5.4bn.

Model developers are a growing share of demand, and their funding is concentrated. SoftBank said on 1 October 2026 that it had completed a $30bn follow-on investment, bringing its cumulative investment in OpenAI to $64.6bn for roughly 13% ownership. In other words, a large part of the demand that justifies the build-out is itself funded by a small number of investors.

Why negative free cash flow matters now

Negative free cash flow is not a crisis for companies of this size. Alphabet’s trailing-twelve-month free cash flow was still about $19.7bn positive at the end of the quarter, and all four have strong balance sheets. But it changes the conversation in three ways.

  • Funding mix. When capital spending exceeds operating cash flow, the gap is covered by cash reserves, debt or leases. That makes AI infrastructure more sensitive to interest rates and credit markets than it was when it was paid for from profits.
  • Investor patience. Shareholders accepted rising capex while free cash flow grew. Now the return case rests on future revenue, such as Microsoft’s $677bn of contracted obligations, turning into cash on schedule.
  • Supplier concentration. For the chip makers, data centre builders and power suppliers that sell to these companies, a handful of buyers decide the order book. A pause by one of them would be felt quickly.

What it means for founders and investors

For start-ups selling into the AI stack, today’s conditions are unusually favourable: buyers are spending heavily and competing for capacity. The risk is in the assumptions. A business plan that extrapolates 2026 spending growth into 2027 and 2028 is betting that the funding gap stays open. Founders should know which of their customers’ budgets are funded from operating cash and which depend on capital raised elsewhere, and should stress-test revenue against a year in which one big buyer holds its spending flat.

For investors, the October numbers argue for watching cash conversion more closely than headline revenue growth. Microsoft’s ability to keep spending while staying cash-positive is a different risk profile from Amazon’s twelve-month outflow, even though both are growing their cloud businesses quickly.

What to watch

  • 10 November 2026: SoftBank Group’s results for the quarter to September (SoftBank IR), including the effect of its OpenAI investments.
  • 17 November 2026: Nvidia’s third-quarter fiscal 2027 results, after guidance of $108.0bn revenue plus or minus 2% in its August release.
  • Alphabet’s fourth-quarter call (date not yet set): the company has said it will give more detail on 2027 capital spending.

Sources

  1. Alphabet: Q3 2026 results, SEC exhibit 99.1
  2. Microsoft: FY27 Q1 results, SEC exhibit 99.1
  3. Meta: Q3 2026 results, SEC exhibit 99.1
  4. Amazon: Q3 2026 results, SEC exhibit 99.1
  5. Oracle: FY27 Q1 results, SEC exhibit 99.1
  6. Nvidia: Q2 FY27 financial results
  7. SoftBank Group: completion of follow-on investment in OpenAI (1 October 2026)
  8. SoftBank Group: investor relations
  9. Microsoft FY27 Q1 earnings call transcript (The Motley Fool)
  10. CNBC: Alphabet Q3 2026 earnings (28 October 2026)