$638 billion of signed business. Negative $23.7 billion of cash. Oracle is building the AI economy on borrowed time.
Oracle Is No Longer a Software Company. It Is an AI Landlord.
For thirty years the pitch was simple. Oracle sold the database that ran the enterprise, then sold the support contract that renewed forever. That business still exists, and in fiscal 2026 it produced $24.5 billion in software revenue, down 1%. It is not the business that moves the stock anymore.
Cloud infrastructure is. Oracle Cloud Infrastructure revenue grew 93% in Q4 to $5.8 billion, and $18.1 billion for the full year, up 77%. Management has told investors that OCI reaches $166 billion by fiscal 2030, which would be roughly 75% of total company revenue. In fiscal 2025 it was about 18%. There is no precedent for a company this large re-basing its revenue mix this fast.
The mechanism is compute leasing. Oracle builds gigawatt-scale data centres, fills them with Nvidia systems, and rents capacity under multi-year contracts. That is closer to a utility or a REIT than to enterprise software, and it should be valued accordingly.
From Relational Database to Hyperscaler: 49 Years in Five Acts
Larry Ellison founded the company in June 1977 and ran it as CEO until 2014. He remains Executive Chair and CTO at 81, per the fiscal 2026 10-K. The arc since then breaks into five phases.
The leadership change matters for how the next two years get executed. Clay Magouyrk and Mike Sicilia took over as CEOs in September 2025, with Safra Catz moving to Executive Vice Chair. Hilary Maxson joined as CFO in April 2026, hired specifically for experience scaling capital-intensive global businesses. Read that hire as a signal. Oracle knows its problem is funding, not demand.
The single largest commitment is with OpenAI. In July 2025 the two companies agreed to develop up to 4.5 gigawatts of additional Stargate capacity, a partnership OpenAI has described as exceeding $300 billion over five years, with revenue beginning in earnest in fiscal 2027.
The Q4 FY2026 Print Was a Record and a Warning
The June 10 report was the strongest quarter Oracle has ever produced and the market sold it anyway. Both reactions were rational.
Metric | FY2026 | Change |
|---|---|---|
Total revenue | $67.4B | +17% |
Cloud revenue | $34.0B | +39% |
Cloud infrastructure | $18.1B | +77% |
Non-GAAP EPS | $7.63 | +27% |
Operating cash flow | $32.0B | +54% |
Capital expenditure | $55.7B | +162% |
Free cash flow | negative $23.7B | negative |
RPO | $638B | +363% |
Two numbers deserve attention. Capex of $55.7 billion against operating cash flow of $32.0 billion is the entire bear case in one line. And of the $638 billion backlog, $75 billion is already covered by customer prepayments or customer-supplied GPUs, which is the underrated bull case. Customers are funding part of the build. That is not what a desperate vendor gets.
Company Health: The Balance Sheet Is Where the Story Turns
At May 31, 2026, Oracle carried $129.5 billion of borrowings against $31.9 billion of cash and securities, so net debt of roughly $97.6 billion. Stockholders' equity was $43.1 billion. Property, plant and equipment more than doubled to $100.0 billion. Depreciation nearly doubled to $7.6 billion, and interest expense reached $4.6 billion for the year with the Q4 run-rate closer to $5.8 billion annualised.
Funding is now the operating constraint. Oracle raised $43 billion in debt and $5 billion in equity in fiscal 2026, and plans roughly $40 billion more in fiscal 2027 including a previously announced $20 billion at-the-market equity programme. It has said it does not expect to issue more debt in calendar 2026.
Every dollar of the build-out above the green bar is borrowed or issued. That is the whole credit story in one chart.
The rating agencies responded. S&P downgraded Oracle to BBB- from BBB on July 9 and forecasts the fiscal 2027 free operating cash flow deficit widening toward $42 billion, with leverage in the mid-4x area. Moody's holds a negative outlook. Bloomberg reported Oracle's ten-year bonds yielding around 6.5%, well wide of comparable BBB paper. One more notch takes Oracle to high yield, which would trigger mechanical selling from investment-grade-only mandates. That is the tail risk equity holders are actually pricing.
Why Did the Stock Fall 66% While the Backlog Rose 363%?
Because a backlog is a promise and cash flow is a fact. The market has decided it will not pay a growth multiple on contracted revenue it cannot yet see converted, financed with debt at 6.5%, concentrated in a handful of counterparties, with OpenAI widely estimated to represent around half of the relevant cloud exposure.
There is also a mechanical component. Credit spreads widened across the AI cloud complex through late July, and equity followed. When the bond market re-rates the funding cost of a build-out, the equity re-rates the terminal value. Nothing about Oracle's demand picture changed between $345 and $118.
What Wall Street Expects Next
Guidance for fiscal 2027 is $90 billion of revenue, up roughly 34%, and $8.05 of non-GAAP EPS, up 18% adjusted for one-off gains. Q1 guidance calls for revenue growth of 27% to 29% and EPS of $1.72 to $1.76. Long-term targets are unchanged at a 31% revenue CAGR and 28% EPS CAGR through fiscal 2030, landing at $225 billion of revenue and $21 of EPS.
Analyst positioning is unusually split. MarketBeat's tally of 39 analysts shows a $265.03 average target, with 30 buys, 8 holds and 1 sell. BTIG carries $400. CLSA initiated at hold with $145 on July 20. RBC sits at $190. Stephens is at $164. A $255 gap between high and low target on a $343 billion company is not a forecast, it is a coin flip with a spreadsheet attached.
The next hard data point is Q1 fiscal 2027 earnings, confirmed for September 14, 2026.
Valuation: What Are You Actually Paying For?
Start with the arithmetic. At $117.77 on 2.914 billion diluted shares, market capitalisation is $343.2 billion. Add net debt of $97.6 billion and enterprise value is $440.8 billion.
Multiple | Calculation | Result |
|---|---|---|
P/E on FY27 guidance | 117.77 / 8.05 | 14.6x |
P/E on FY26 non-GAAP | 117.77 / 7.63 | 15.4x |
EV / FY27 revenue | 440.8 / 90 | 4.9x |
EV / RPO | 440.8 / 638 | 0.69x |
P/E on FY30 target EPS | 117.77 / 21 | 5.6x |
Two observations. First, 14.6x forward earnings for a business guiding 34% revenue growth is a multiple normally reserved for value traps or cyclicals at peak. Second, the leverage cuts both ways in valuation. With $97.6 billion of net debt against a $343 billion equity value, every one turn of EV/sales on $90 billion of revenue is worth about $30 per share to equity holders. Every one turn of P/E is worth $8.05. Small changes in sentiment produce large changes in price, which is exactly what the last ten months demonstrated.
Price Targets and the Math Behind Them
Three scenarios, each anchored to an explicit EPS figure and an explicit multiple. Assume roughly 3.0 billion shares after the $20 billion equity programme, which is about 5.7% dilution at current prices.
The amber line is spot. Everything above and below it is somebody's spreadsheet.
Bear case reasoning: an 11x multiple is what the market pays for a levered utility with an uncertain returns profile, and it is roughly where Oracle would trade on a downgrade to high yield. Base case reasoning: 17x on delivered guidance is still a discount to the broad market, which is fair for a business with negative free cash flow. Bull case reasoning: 24x on fiscal 2028 earnings requires visible free cash flow inflection and a stable rating, which S&P currently projects for around 2029.
Cross-check on enterprise value. At 5.5x FY27 revenue of $90 billion, enterprise value is $495 billion. Subtract expected net debt of roughly $115 billion after further borrowing, divide by 3.0 billion shares, and you get $127. That brackets the base case from below.
Long-horizon check. If management delivers $21 of fiscal 2030 EPS and the market applies a 15x multiple, the 2030 price is $315. Discounting four years at a 12% cost of equity gives $200 today. That is the number to keep in mind: taking management fully at its word still implies a wide range against a bear case in the $70s.
Probability-weighting the three scenarios at 25% bear, 50% base and 25% bull produces an expected value of $148, or 26% above spot. Note how much less that is than the $265 consensus. The gap is not disagreement about demand. It is a disagreement about what a multiple should be when free cash flow is negative and the rating is one notch from junk.
So Is Oracle Expected to Grow?
Yes, and that was never the question. Growth is contracted, disclosed, and audited. The question is whether shareholders capture it, and that depends on three things.
Conversion speed. Management stated that most recent bookings sit where near-term capacity is already available, which shortens the path from RPO to revenue.
Margin defence. AI infrastructure gross margins have been framed at 30% to 40%. Depreciation on $100 billion of PP&E and rising interest expense both work against reported earnings.
Funding cost. Every basis point of spread widening raises the hurdle rate on the entire build.
Get all three right and the bull case is conservative. Miss one and the base case is generous.
Key Risks
Customer concentration, with OpenAI estimated near half of relevant cloud exposure.
A further downgrade to high yield, forcing index and mandate-driven selling.
Equity dilution from the $20 billion at-the-market programme.
A broader AI capex digestion phase compressing multiples across the complex.
Execution risk on power, land, and GPU delivery timelines.
What to Watch Next
September 14: Q1 fiscal 2027 results, with RPO conversion the only line that matters.
The next S&P and Moody's reviews, which are the highest-beta events on the calendar.
Credit default swap levels as a real-time sentiment gauge, currently near eighteen-year wides.
Progress on the $40 billion fiscal 2027 funding plan, and whether the debt promise holds.
How to Trade a Stock With a $255 Target Spread
Oracle is not a conviction trade in one direction. It is a volatility trade with a credit overlay, and the September print is a binary. That calls for an instrument that lets you take either side, size precisely, and manage exposure around a known date.
You can do exactly that on ApeX Omni with the ORACLE-USDT RWA perpetual, with up to 50x leverage, long or short, no KYC and full self-custody. RWA pricing runs on Chainlink Data Streams, so your fill tracks the underlying rather than a thin internal book. Backlog converts, you are long into $247. Rating gets cut, you are short into the $70s. Either way you are not waiting on a broker to open at 9:30.
Trade the print, not the narrative. Position sizing still matters more than direction, and leverage magnifies losses as fast as gains.
This article is for educational and informational purposes only and does not constitute financial, investment, or legal advice. Traditional finance (TradFi) markets and financial products involve risk, and past performance does not guarantee future results. Market conditions, valuations, and other financial information referenced are accurate to the dates and sources cited but may change without notice. Always conduct your own research and consult a licensed financial professional before making any investment decisions.