Key Highlights

  • Cloud infrastructure revenue rose 93% to US$5.8 billion in fiscal Q4 2026, with remaining performance obligations of US$638 billion.
  • The year produced negative free cash flow of US$23.7 billion on US$55.7 billion of capex, and S&P cut Oracle to BBB-.
  • Shares closed at US$119.90, down roughly 51% over twelve months, versus an analyst consensus target of US$248.15.
  • The dominant risk is counterparty: if OpenAI cannot fund its commitments, roughly half of a US$638 billion backlog becomes questionable while the depreciating assets built to serve it stay on Oracle's books.

Oracle (NYSE:ORCL)'s fiscal 2026 delivered the fastest infrastructure growth in its history and a contracted backlog larger than most national economies. Investors responded by cutting the shares roughly in half, and S&P lowered the credit rating to one notch above high yield.

A Growth Story the Market Stopped Paying For

Oracle Corporation (NYSE: ORCL) has spent the past year proving it can sell artificial-intelligence computing capacity faster than almost anyone expected, and demonstrating how expensive that success is. Its fiscal 2026 results, reported, were the strongest in its modern history by nearly every growth measure, yet the shares now sit near the bottom of a 52-week range topping out at US$345.72. Oracle is no longer judged on whether demand exists, but on whether the balance sheet can survive meeting it.

The Rating Agency Moves the Goalposts

S&P Global Ratings lowered Oracle's issuer credit rating to BBB- from BBB, the lowest rung of investment grade. The rationale: S&P now projects fiscal 2027 capital expenditure at US$90 billion to US$95 billion, against a prior assumption near US$60 billion, and free cash flow near negative US$42 billion, nearly double its earlier estimate, with adjusted leverage reaching the mid-4x area.

Press coverage also reported a negative outlook from Moody's, whose rating stood at Baa2 as of Oracle's February 2026 bond sale; that outlook is treated here as reported, not confirmed by a primary agency release.

S&P's capex projection sits well above Oracle's own guidance: on the fourth-quarter call, Chief Financial Officer Hilary Maxson guided to roughly US$70 billion of net capital expenditure for fiscal 2027.

What Fiscal 2026 Actually Delivered

For fiscal fourth-quarter 2026, total revenue was US$19.2 billion, up 21%. Cloud revenue reached US$9.9 billion, up 47%, of which cloud infrastructure (OCI, the IaaS line) contributed US$5.8 billion, the 93% figure. Cloud applications grew a far more sedate 10% to US$4.1 billion, and legacy software licence and support revenue fell 2%.

For the full year ended revenue was US$67.4 billion, up 17%, with OCI at US$18.1 billion, up 77%. GAAP operating income was US$20.6 billion at a 31% margin; non-GAAP operating income US$28.9 billion at 43%. GAAP net income rose 36% to US$17.0 billion, giving GAAP EPS of US$5.83 and non-GAAP EPS of US$7.63.

The cash-flow statement tells the other half. Operating cash flow grew 54% to US$32.0 billion, but capital expenditure of US$55.7 billion, above the roughly US$50 billion guided, left free cash flow at negative US$23.7 billion.

Remaining performance obligations, the technical term for what headlines call "backlog", stood at US$638 billion at year-end, up from US$553 billion three months earlier and more than four times the prior-year level. Because Oracle's fiscal year ends 31 May, its next results (fiscal 2027 first quarter, ending ) were unavailable at this article's cut-off.

Enormous Bookings, Concentrated in Few Hands

The RPO number is both the bull case and the bear case. Management disclosed on the fourth-quarter call that four customers had each signed contracts exceeding US$8 billion, and that only about 12% of RPO converts to revenue within twelve months, with roughly 34% inside three years.

It is also concentrated: S&P and multiple financial outlets reported that OpenAI accounts for approximately half of Oracle's RPO, reflecting the multi-year cloud agreement reported in September 2025 at around US$300 billion. Oracle publishes no customer-by-customer breakdown, so that share is an agency and press estimate, not company disclosure.

Oracle stated in March 2026 that leasing for its 4.5-gigawatt OpenAI commitment had been "completed" and the flagship Abilene, Texas site remained on schedule, after Bloomberg reported a separate 600-megawatt expansion lease there had been abandoned. Roughly 42% of Abilene's Stargate capacity had been delivered. A March 2026 filing valued Oracle's TikTok US joint-venture stake just over US$2 billion.

The Information, citing internal documents, reported that Oracle's AI cloud deals carried an average gross margin near 16%, ranging between 10% and 20%, with Nvidia Blackwell capacity running at a loss in one period. Oracle declined to comment; management argues OCI margins expand with scale.

Fastest Grower, Smallest Cushion

Oracle's OCI growth rate comfortably exceeds that of Amazon Web Services, Microsoft Azure and Google Cloud, but it grows from a far smaller base and without their financial ballast. The hyperscalers fund AI capacity from established, cash-generative franchises; Oracle funds it from a database and applications business whose licence revenue is slightly shrinking, plus capital markets.

Against neoclouds such as CoreWeave and Nebius, Oracle holds advantages in enterprise relationships, sovereign-cloud positioning and its applications estate. But its model has drifted closer to theirs, leasing GPU capacity on long contracts, financed by debt, than to a software company's, which explains much of the multiple compression.

Funding the Build

Oracle announced a calendar-2026 financing plan targeting US$45 billion to US$50 billion in gross proceeds, split roughly evenly between equity and debt, including a US$20 billion at-the-market equity programme and mandatory convertible preferred securities, which it said "reflects Oracle's commitment to maintaining an investment-grade rating."

The debt half executed : a US$25 billion, eight-tranche senior unsecured offering spanning three to forty years, priced at spreads of 95 to 195 basis points, with order books reported near US$155 billion. Demand was not the problem; price was, those spreads came 40 to 58 basis points wider than Oracle's September 2025 deal.

On the most recent reported balance sheet, Oracle carried roughly US$167.4 billion of total debt against US$31.9 billion of cash. Its five-year credit default swap spread was reported around 198 basis points in April 2026, described then as a company record; more recent levels were not verifiable at the cut-off. The US$0.50 quarterly dividend continues.

What Could Re-Rate the Shares

Management guided to fiscal 2027 revenue of roughly US$90 billion and non-GAAP EPS of US$8.05, with first-quarter cloud growth of 57% to 63%. These are company forecasts, not results; delivering against them is the most powerful catalyst available.

Four other developments would matter: OCI gross margins climbing as capacity scales; RPO conversion accelerating beyond the 12% twelve-month rate; backlog diversification away from OpenAI; and a credible path back to positive free cash flow. Press reports of roughly 30,000 job reductions in 2026, plus the leadership reset installing Clay Magouyrk and Mike Sicilia as co-CEOs in September 2025 with Safra Catz as executive vice chair, suggest management is optimising for cash.

Where the Risk Actually Sits

The dominant risk is counterparty: if OpenAI cannot fund its commitments, roughly half of a US$638 billion backlog becomes questionable while the depreciating assets built to serve it stay on Oracle's books. Second is refinancing, at BBB-, one further notch pushes a very large debt stack out of investment-grade indices, forcing mechanical selling and higher funding costs. Third is asset life: GPU generations turn over quickly, and the Abilene episode showed how architecture transitions can strand planned capacity. Fourth is dilution, as the ATM programme issues equity at whatever price the market sets, now far below last year's.

Short interest is modest at roughly 2.9% of float, suggesting the repricing came from long holders exiting rather than aggressive shorting.

Valuation Reflects Genuine Disagreement

At US$119.90, Oracle's market capitalisation was roughly US$345.4 billion and enterprise value near US$480.5 billion. Trailing P/E was about 19.7, forward P/E about 14.3, price-to-sales about 5.1 and EV/EBITDA about 16.1. For a company growing revenue 17% and infrastructure revenue 77%, those are not demanding multiples.

Analyst opinion is unusually dispersed. Across 44 analysts polled, the average target was US$248.15 and the median US$240, ranging from US$110 to US$400, a 3.6x spread on a mega-cap. Guggenheim's John DiFucci maintained a Buy at US$400 ; Bernstein's Mark Moerdler a Buy at US$325 ; CLSA's Bhavtosh Vajpayee initiated at Hold with a US$145 target.

Conclusion

Oracle has answered the demand question emphatically: a US$638 billion contracted backlog and 93% infrastructure growth are not accounting artefacts. What it has not answered is whether it can bridge the gap between spending now and collecting later without a credit accident or heavy dilution.

The equity market has priced meaningful probability of damage; the credit market, via a BBB- rating and wider spreads, has priced it explicitly. If Oracle converts backlog on schedule at improving margins, today's multiples will look conservative. If OpenAI stumbles, or capex runs to S&P's US$90 billion-plus projection while cash flow stays deeply negative, the current price will not have been the floor. Both outcomes remain live, and the next data point lands with fiscal 2027 first-quarter results in September 2026.

This article is for informational purposes only and does not constitute investment advice or a recommendation to buy, sell, or hold any security. Share prices can rise or fall, and past performance does not guarantee future results. Investors should conduct independent research and consult a licensed financial adviser before making investment decisions.