Update, August 3, 2026: when this was published, the section below headed "Applying our own test to the name we skipped" refused to give a capex figure on the grounds that this repository did not hold one. That is no longer true — Oracle's capex has since been added to our stored Oracle data for all fourteen quarters, from SEC XBRL. The number the article declined to invent is now printed in that section, sourced. The refusal itself still stands for the two figures it also named, RPO and debt raised, which remain absent. Nothing else below has changed.
Three days ago we published a six-month update on the AI capex bet with a section headed "Free Cash Flow Is No Longer a Forecast — It's Actually Cratering." It named Alphabet, Meta, Tesla and Amazon. It went beyond the Mag 7 deliberately — it added SpaceX as "the eighth name in the chart," ran a Nvidia section, and pulled in Micron and Sandisk from the supply side.
Oracle is in none of it. Before this piece, Oracle appeared across the whole site exactly once, in a February piece about Robinhood, where the word is incidental, and no article carried ORCL as a ticker. This one does.
In the fiscal year that ended May 31, 2026, Oracle's free cash flow was −$23,686M. The year before it was −$394M. That is a $23.3 billion swing in twelve months, at a company whose quarterly revenue growth went from +6.4% (quarter ended 2025-02-28) to +21.7% (2026-02-28) over the same stretch.
What the series shows
Every figure below is from our stored Oracle data, which carries 14 quarters of revenue, EPS, free cash flow, gross margin and — since this piece was first published — capital expenditure. Oracle's fiscal year ends in May, so its "2026 Q4" is the quarter ended 2026-05-31, not a calendar Q4 — every row here carries its period end for that reason.
| Fiscal quarter | Period end | Revenue | Revenue YoY | Free cash flow | FCF margin |
|---|---|---|---|---|---|
| 2025 Q3 | 2025-02-28 | $14,130M | +6.4% | +$71M | +0.5% |
| 2025 Q4 | 2025-05-31 | $15,903M | +11.3% | −$2,923M | −18.4% |
| 2026 Q1 | 2025-08-31 | $14,926M | +12.2% | −$362M | −2.4% |
| 2026 Q2 | 2025-11-30 | $16,058M | +14.2% | −$9,967M | −62.1% |
| 2026 Q3 | 2026-02-28 | $17,190M | +21.7% | −$11,484M | −66.8% |
| 2026 Q4 | 2026-05-31 | $19,184M | +20.6% | −$1,873M | −9.8% |
The single number worth stopping on needs no comparison to anything: in the quarter ended 2026-02-28, Oracle spent two-thirds of its revenue more than it took in. Not two-thirds of profit. Two-thirds of revenue, as free cash outflow, in one quarter.
The four names our capex piece did cite are calendar-quarter filers, so their quarters and Oracle's do not line up and the dollar figures are not comparable side by side. The margin is, because it is a ratio of each company's own revenue. On the site's own series, Alphabet's first negative free-cash-flow quarter — −$5,855M for the quarter ended 2026-06-30 — was −4.9% of its revenue. Oracle's worst was −66.8%. Same measure, same source, more than thirteen times the intensity.
Gross margin gave ground while this ran, and this is the quieter line: 67.28% (period end 2025-08-31) → 66.53% (2025-11-30) → 64.56% (2026-02-28) → 65.22% (2026-05-31). Down 2.1 points across the fiscal year, with the trough in the same quarter as the worst burn.
A company growing 21% scores an 11
The Rule of 40 is revenue growth plus free-cash-flow margin — we are not going to re-derive it here, the explainer does that. For Oracle's quarter ended 2026-05-31:
+20.63% revenue growth + (−9.76%) FCF margin = 10.87
The site renders R40 11 for ORCL, and it reconciles. One caution if you check the arithmetic yourself: rounding the halves first gives +20.6 − 9.8 = 10.8, which disagrees with the unrounded 10.87 at one decimal. Both round to the 11 on the page, but the halves and the total are not the same number if you round before adding.
That score is the whole point of this piece. Oracle's growth went from +6.4% (quarter ended 2025-02-28) to +21.7% (2026-02-28) — the fastest growth this series records — and the score barely moved, because the cash half absorbed all of it. This is the formula doing exactly the job it exists to do: refusing to call growth free when it was bought.
Worth noting how the score is labelled now. The page carries R40 11 · 2026 Q4, and for Oracle that quarter ended in May. A reader who reads "2026 Q4" as December would be six months out.
Applying our own test to the name we skipped
The capex piece's central finding was not that spending is bad. It was that the market punishes capex without an independently verifiable demand signal attached and rewards capex with one — Microsoft's $678B backlog and Amazon's AWS reacceleration got roughly +8-9%; Alphabet and Meta got sold.
Oracle is the cleanest available test of that framework, because it is the borrower's side of the same buildout: a company funding datacenter construction that its customers, not it, will run. The test the framework demands is Oracle's remaining performance obligation — the contracted backlog that would show whether this spending is chasing signed demand or hope.
We are not going to give you that number, because this repository does not contain it. Our SEC filing index for Oracle lists 59 filings — including the FY2026 10-K, filed 2026-06-22 for the period ended 2026-05-31 — but it stores filing metadata only: form, dates, accession numbers. No line items. Oracle's RPO and the debt it raised are both in that 10-K and neither is in anything we hold, and inventing either would cost this piece more than leaving the question open.
Capex was in that same refusal when this published, and it no longer belongs there. The original passage said deriving a capex figure by subtracting our free-cash-flow line from operating cash flow and presenting it as reported would be inventing a number — which was right at the time. The series has since gained the real thing, from SEC XBRL, every point tagged PaymentsToAcquirePropertyPlantAndEquipment. So it can be printed rather than refused:
| Fiscal quarter | Period end | Capex | Free cash flow |
|---|---|---|---|
| 2026 Q1 | 2025-08-31 | $8,502M | −$362M |
| 2026 Q2 | 2025-11-30 | $12,033M | −$9,967M |
| 2026 Q3 | 2026-02-28 | $18,635M | −$11,484M |
| 2026 Q4 | 2026-05-31 | $16,493M | −$1,873M |
| FY2026 | $55,663M | −$23,686M |
The right-hand column is the check: it sums to −$23,686M, the same $23.7B this article leads on. The capex is not a new number sitting beside the burn — it reconciles against it, from the same file, on the same fourteen-quarter series.
And the ramp is the thing the section was reaching for. By fiscal year: $6,866M in FY2024, $21,215M in FY2025, $55,663M in FY2026 — roughly eight times the FY2024 figure two years later.
So the question stands open and specific: Oracle's RPO, from the FY2026 10-K, against the $23.7B it spent. That is the number that decides whether Oracle belongs in our piece's rewarded column or its punished one, and it is the thing to look up before forming a view.
Is the burn over?
The quarter ended 2026-05-31 came in at −$1,873M, up sharply from −$11,484M the quarter before. That looks like the end of it. Two reasons to hold off:
This series is lumpy in exactly this way. Look at fiscal 2025: +$5,124M (period end 2024-08-31) sat directly beside −$2,666M (2024-11-30), and +$71M (2025-02-28) beside −$2,923M (2025-05-31). A single quarter reversing tells you less in this series than it would in most.
The revenue line did not slow. Growth was +20.6% in that same quarter, on the highest revenue in the series. A burn that ends because the building stopped and a burn that ends because one quarter's payments landed differently look identical for one quarter and diverge over three.
What to watch
- Oracle's next print. Our earnings schedule carries a scheduled report for fiscal 2027 Q1, period ending 2026-08-31, with a publish date of 2026-09-09. Two consecutive quarters back near breakeven would make the snap-back real; one more −$9B quarter would make 2026 Q4 the outlier.
- The RPO line in the FY2026 10-K. Named above, still the number that settles the demand-signal question either way.
- Gross margin's direction. It ticked back up to 65.22% in the quarter ended 2026-05-31 after bottoming at 64.56%. Whether that recovers toward the 70% Oracle ran at through fiscal 2025 is a cleaner read on the economics of this business than the cash line, which is dominated by construction timing.
The bottom line
Oracle spent its entire free cash flow, and $23.3 billion more than the prior year, building capacity for the AI buildout, and its revenue growth more than tripled while it did. Our own capex retrospective built a section around free cash flow cratering and did not mention it — which is a coverage failure on a name this size, not a scope decision, and this piece is the correction.
The Rule of 40 caught what the coverage missed. A score of 11 on 21% growth is not the metric failing to notice a good quarter; it is the metric reporting, accurately, that the growth was paid for in cash the company no longer has. Whether that was a good trade depends on a backlog figure neither this article nor this site currently holds — and saying so is more useful than a number we would have had to make up.
Oracle figures are from our stored Oracle data: revenue, free cash flow, gross-margin and capital-expenditure series of 14 quarters through the period ended 2026-05-31, and the earnings-schedule entry scheduling fiscal 2027 Q1 for 2026-09-09. The capex series carries its XBRL concept on every point, PaymentsToAcquirePropertyPlantAndEquipment for all fourteen, and was added after this article was first published — see the note at the top. Do not assume any series in our data is sorted: exactly one is not. Tesla's gross-margin series runs 2026-03-31 back to 2010-03-31 across 65 points with three ascending steps in its early history, and it is the only unsorted series we store — Oracle's own six run oldest-first, as does every other series, and nothing is stored strictly newest-first. Figures here were read after sorting by period end, which costs nothing and is the only safe assumption. Alphabet's comparison figure is from our stored Alphabet data for the quarter ended 2026-06-30. The FY2026 10-K reference is the filing listed in our SEC filing index for Oracle, filed 2026-06-22 for the period ended 2026-05-31; no financial line item is quoted from it, because the index stores filing metadata only. The Rule of 40 definition is the one in our explainer.