Hyperscaler capex rose 81% to $566bn while free cash flow fell 38% to $126bn
Trailing-twelve-month operating cash flow, capital expenditure, free cash flow and free cash flow after stock-based compensation, summed across MSFT, AMZN, GOOGL, META, ORCL, quarterly since 2015 (latest 2026 Q2, calendar-aligned).
Capex against free cash flow, five hyperscalers combined
$bn, trailing twelve months
Free cash flow = operating cash flow − capital expenditure. FCF after SBC also deducts stock-based compensation.
Capex as a share of operating cash flow
TTM to 2026 Q2, by company
What it shows · fact
Combined TTM capex is $566bn (+81% on a year earlier) against operating cash flow of $692bn, leaving free cash flow of $126bn (-38%). After $91bn of stock-based compensation, it is $35bn.
Capex absorbs 82% of operating cash flow, up from 61% a year earlier. ORCL, AMZN now spend more than their operating cash flow.
Why it matters · inference
The AI build-out is absorbing operating cash flow that could otherwise fund buybacks and dividends. The question for valuation is whether the spending earns a return above the cost of capital, which free cash flow will not show for several years.
Capex is a cash measure; depreciation of the new assets will flow through earnings for years, so reported margins lag the cash picture.
What would change this read
Capex growth slowing below operating-cash-flow growth would let free cash flow recover without any change in AI demand.
Disclosed AI revenue growing faster than the depreciation of the new assets would be the evidence that the spending is earning its keep.
Method
- Quarterly cash-flow statement items as reported (operating cash flow, capital expenditure, stock-based compensation), summed over four quarters per company, then across companies.
- Capex is the vendor capital-expenditure line (cash purchases of property and equipment); equipment obtained under finance leases is not a cash capex outflow and is not included.
- Fiscal quarters are calendar-aligned on their period-end date (Microsoft and Oracle have non-calendar fiscal years).
Sources
- MCH research warehouse, quarterly statements (Alpha Vantage, as restated).
Universe
| Names in the published feed | 886 |
| Excluded: held, suspended, withdrawn, deferred, deal-pending or pending review | 53 |
| Reader-safe universe | 833 |
| Used in this exhibit | 5 |
Names whose rating is held, suspended, withdrawn, deferred or under a pending deal are excluded from every exhibit, so no held output appears here.
General and impersonal investment research. Not personal financial advice. Forecasts, valuations and model outputs are estimates and may be wrong. See Investment Disclaimer.
MCH Analysis is a general investment-research publication of MCH Advisory (Pty) Ltd. It is not personal financial advice and does not take your circumstances into account. MCH is not a licensed financial services provider or a registered investment adviser. Past performance, whether actual or simulated, does not predict future results. Subscriber Agreement | Privacy and Cookie Notice | Research Governance Policies