Oracle’s stock has been cut by more than half from its 52-week high, down near $153 from a peak above $330. Over that same stretch, the reported cloud business roughly doubled. That gap between the chart and the income statement is usually the setup I look for — a market still pricing the old story while the numbers improve. Here the gap closes for a concrete reason, and it changes the call.
By the numbers, the record kept improving. Cloud-infrastructure revenue grew 84% year over year in the third fiscal quarter to $4.9 billion, then 93% in the fourth.
| Quarter | OCI growth YoY | OCI revenue |
|---|---|---|
| Q3 FY2026 | 84% | $4.9B |
| Q4 FY2026 | 93% | not disclosed |
| Q1 FY2027 | pending | pending |
Total cloud revenue climbed 47% to $9.9 billion in the fiscal fourth quarter, and management raised its fiscal-2027 revenue forecast to roughly $90 billion. An operating record like that is what a rerating story is supposed to show. The trouble is not whether the growth is real. It is who buys it and who pays for the build.
Start with who buys it. Roughly half of Oracle’s backlog is tied to a single customer, OpenAI. The capacity being added to serve that demand is concentrated the same way: analysts expect about a gigawatt of new OCI capacity to come online this quarter, chiefly for that account, and OracleORCL-- and Crusoe are building an OpenAI data-center campus in Abilene, Texas, with two buildings already operational. The growth engine is real, and it is not diversified. Behind the doubling line, the legacy software business shrank 2% in the fourth quarter, so all the weight lands on one concentrated account that could slow or renegotiate at any time.
Then who pays for it. This is the part the accounting beat does not show. Fiscal 2026 burned $55.6 billion of capital spending and produced $23.7 billion of negative free cash flow, against trailing operating cash flow of about $32 billion. Management’s fiscal-2027 net-capex range of $70 billion to $95 billion sits above that current run-rate, so the cash burn points up, not down. To bridge the gap Oracle plans $45–50 billion of fiscal-year debt raises on a balance sheet that already carries more than $200 billion in total debt. Last quarter, an EPS beat without cash-flow evidence drove the stock down more than 10% — the market is already trading the cash-flow bridge, not the accounting line.
So the halving is not clean mispricing of a good record. The most defensible read is that the market is repricing the two things that could actually break the story — a single dominant customer and a leveraged, capex-heavy build. Both are already visible in the disclosed numbers, which is why an improving income statement and a falling stock can be true at once.
This is not about excitement. It is about a business that will look harder to dismiss once the free cash flow shows up — and right now it is not showing up, and guidance points to it staying negative. My method prices a stock off its forward free cash flow, and there is no positive bridge here to discount into a materially higher value today. So the honest call is hold: do not chase a falling knife on a doubling line paid for by one customer and a leveraged build.
I can be wrong again; the setup is what it is. What would change my mind is one thing nobody knows yet: the capex program topping out. If management shows spending peaking and free cash flow swinging toward positive while that concentrated backlog keeps converting, the bridge reappears and the repricing could be close to done. Until then, the divergence between a doubled cloud line and a halved stock is explained by concentration and balance-sheet risk — and I do not need to own that at this price.
Comments
No comments yet