NextFin News - Oracle raised its fiscal 2027 revenue forecast to $90 billion and said it can "comfortably meet and likely exceed" that growth rate, after third-quarter cloud infrastructure revenue jumped 84% to $4.9 billion and its backlog of contracted AI work swelled to $553 billion. The real story is not the beat — it is the financing model behind it. Oracle is building one of the world's most capital-intensive AI infrastructure stacks while insisting that customer prepayments, not its own balance sheet, will foot most of the bill. The quarter proved the demand is real; the next 18 months will prove whether the funding structure survives the build-out.
The Quarter: A Beat Built on AI Contracts
For the fiscal third quarter ended Feb. 28, Oracle reported total revenue of $17.2 billion, up 22% in U.S. dollars and 18% in constant currency, with non-GAAP earnings per share of $1.79, up 21% and ahead of the $1.70 consensus. Cloud revenue reached $8.9 billion, up 44%, and cloud infrastructure revenue — the AI-facing engine — accelerated to $4.9 billion, up 84%, after growing 68% and 55% in the two prior quarters. Management called it the first quarter in more than 15 years where organic total revenue and non-GAAP earnings per share both grew at 20% or more.
The guidance was the bigger move. Oracle lifted fiscal 2027 total revenue guidance to $90 billion, roughly 4% above the street consensus of about $86.6 billion, and kept fiscal 2026 revenue at $67 billion with capital expenditures at $50 billion. For the fiscal fourth quarter, it called for revenue growth of 19% to 21% in U.S. dollars and non-GAAP EPS of $1.96 to $2.00. After the release, shares jumped as much as 10% in extended trading, trading above $162 — a sharp reversal for a stock that had fallen from a record near $328 in September 2025 amid doubts about its AI spending.
The quarter's defining number sits off the income statement. Remaining performance obligations — contracted revenue not yet recognized — ended at $553 billion, up 325% from a year earlier and up $29 billion from the prior quarter. Oracle was explicit about what drove it:
Most of the increase in RPO in Q3 related to large scale AI contracts where Oracle does not expect to have to raise any incremental funds to support these contracts as most of the equipment needed is either funded upfront via customer prepayments so Oracle can purchase the GPUs, or the customer buys the GPUs and supplies them to Oracle.
That sentence is the key to the entire investment case. Oracle is not betting its own cash flow that AI demand will hold. It is asking customers to prepay for the GPUs that will serve them, turning a capital expenditure problem into a working-capital advantage. If that structure holds, Oracle can scale capacity without crushing free cash flow. If it breaks, the company is left holding $50 billion of annual data-center spending against a backlog that may convert slower than expected.
Why the Funding Model Matters More Than the Revenue Number
The mechanism here is simple but easily overlooked. A cloud infrastructure build-out has two bottlenecks: physical capacity — chips, data centers, power — and financial capacity, or who pays for it before revenue arrives. Most hyperscalers fund the first with their own balance sheets and hope the second follows. Oracle has inverted that sequence for a large share of its AI contracts. Customer prepayments arrive before Oracle buys the GPUs; in other cases the customer buys the GPUs and hands them to Oracle to operate. Either way, Oracle's cash leaves later than the customer's.
This matters because the spending is enormous. Oracle plans to raise up to $50 billion in debt and equity financing in calendar 2026 and said it does not expect to issue additional bonds beyond that this year. Within days of the announcement it raised $30 billion through investment-grade bonds and mandatory convertible preferred stock, with what it called a record, substantially oversubscribed order book. The at-the-market equity portion had not yet begun. Operating cash flow over the trailing twelve months was $23.5 billion, up 13% — healthy, but not nearly enough to self-fund a $50 billion capex year without the prepayment structure.
The second-order effect is what the market is not fully pricing. If Oracle's model works, it sets a new competitive standard for the entire AI infrastructure market: the winner is not the company with the most chips, but the one that can finance capacity at the lowest cost of capital while shifting deployment risk onto customers. That would compress margins for rivals that fund build-outs internally and force them to match Oracle's prepayment terms or accept slower growth. The third-order implication is even sharper: a prepayment-funded backlog is more durable than an internally funded one in a downturn, because customers with skin in the game are less likely to walk away — but only if the capacity actually comes online when promised.
Oracle is also trying to reduce the labor cost of delivering that capacity. The company said its own AI coding capabilities have advanced to a point where they are "enabling us to build more software in less time with fewer people." In a business where deployment speed determines how fast a backlog converts into revenue, shaving engineering headcount from the build cycle is not just a margin story — it is a race against the clock on lease commitments.
The Counter-Thesis: Backlog Is Not Revenue, and Capacity Is the Constraint
The strongest case against Oracle is not that demand is fake. It is that a $553 billion backlog is a promise, not a paycheck, and that the conversion rate depends on factors Oracle does not fully control: GPU supply, data-center construction, power availability, and customer credit. A backlog that grows faster than capacity can be deployed becomes a liability, not an asset — it raises expectations, locks in pricing, and then disappoints on timing.
There is already evidence of friction. Reports in March said Oracle and OpenAI had scrapped plans to expand their flagship Stargate data center in Texas after financing negotiations dragged and OpenAI's needs changed; Oracle disputed that report as "false and incorrect." Separately, credit rating agencies have carried negative outlooks on Oracle following its $18 billion bond issuance in late 2025, citing the scale of spending, counterparty concentration, and the risk that leverage rises faster than AI revenue scales. The company has disclosed roughly $248 billion in long-term lease commitments tied to data centers and cloud capacity, with 15- to 19-year terms commencing between fiscal 2026 and 2028.
Then there is the customer concentration question. A large share of the AI backlog is tied to a small group of the largest AI compute buyers. If even one of them slows deployment, renegotiates, or — in a severe scenario — fails, the conversion math changes quickly. Michael Burry, the investor known for the "Big Short," said in January that he was betting against Oracle:
I do not like how it is positioned or the investments it is making.
He is not alone in worrying that the capital intensity of the AI build-out will outrun the returns.
The industry-wide context makes the risk concrete. The five largest hyperscalers — Amazon, Microsoft, Alphabet, Meta, and Oracle — are on track to spend between $700 billion and $900 billion on capital expenditures in 2026, a 36% increase over 2025, according to CreditSights estimates. Roughly 75% of that, about $450 billion, is tied directly to AI infrastructure. Sequoia Capital's David Cahn has calculated that there is approximately a $600 billion annual revenue gap between what hyperscalers are spending on AI infrastructure and what the AI ecosystem is actually generating in sales. That gap is the silent partner in every AI infrastructure deal: if the end-market revenue does not materialize, the customers signing these contracts eventually face their own reckoning, and Oracle's prepayment-funded backlog becomes only as strong as the credit of the buyers backing it.
The falsifying signal is specific: watch the ratio of RPO growth to cloud infrastructure revenue growth over the next two quarters. If RPO keeps rising by $25 billion to $30 billion a quarter while cloud infrastructure revenue growth stays near or above 80%, the conversion engine is working and the thesis holds. If RPO growth stalls below $15 billion per quarter while capex stays at $50 billion annually, the backlog is hitting a deployment wall and the funding advantage turns into stranded capacity. A second warning sign: any move by Oracle to delay or restructure lease commitments, or a credit downgrade that raises borrowing costs above the margin it earns on AI contracts. A third: non-GAAP operating margin, now at a level that produced $7.4 billion of quarterly operating income, contracting for two consecutive quarters while capex remains elevated.
Cyclical or Structural: A Structural Shift With a Cyclical Financing Risk
The demand for AI compute is structural. Enterprise migration to cloud databases, the shift from on-premises software to SaaS, and the need for training and inference capacity are not mean-reverting trends; they are a change in how computing is consumed. Oracle's position as a database and applications incumbent with a built-out cloud gives it a durable claim on that demand, and the 325% year-over-year jump in RPO is the footprint of a regime change, not a cycle.
But the financing leg is cyclical. Debt markets, credit spreads, and investor appetite for capital-intensive growth stories all revert. Oracle's ability to raise $30 billion quickly at investment-grade terms is a function of current market conditions that will not necessarily hold through a multi-year build-out. The company's judgment — that it can fund $50 billion of annual capex largely through customer prepayments — is a bet that the structural demand will keep the cyclical funding markets on its side. If credit tightens and customers slow prepayments at the same time, Oracle faces the worst of both worlds.
This separation matters for the conclusion. The structural case says Oracle wins share and grows revenue for years. The cyclical case says the stock will be volatile, and the valuation will compress whenever funding costs rise or deployment slips. Both can be true simultaneously, and the market is pricing them that way: a strong buy rating from some analysts, a skeptical short book from others, and an average price target that sits well below the September highs.
What Comes Next: Three Horizons
Short term, the stock reaction is a relief rally. Oracle entered the quarter with a damaged reputation after the December revenue miss and the September peak near $328. The beat and the raised guidance repair that narrative, and the after-hours jump of 8% to 10% reflects investors repricing the probability that Oracle can execute. The next catalyst is the fiscal fourth quarter, due in June, where the market will watch whether cloud infrastructure growth stays above 80% and whether the $90 billion FY27 number starts to look conservative rather than ambitious.
Medium term, the battleground is conversion. Oracle needs to turn the $553 billion RPO into recognized revenue at a pace that justifies $50 billion of annual capex. Base case: capacity comes online on schedule, prepayments keep funding the build, and revenue grows into the low twenties. Upside case: AI demand accelerates further, Oracle raises guidance again, and the prepayment model lets it outspend rivals without diluting shareholders. Downside case: deployment delays, a credit downgrade, or a major customer renegotiation forces Oracle to fund more internally, compressing margins and free cash flow.
Long term, the question is whether Oracle becomes a permanent pillar of AI infrastructure or a cautionary tale about overbuilding. The structural demand supports the first outcome; the capital intensity threatens the second. The differentiator is execution discipline — bringing capacity online fast enough that the backlog converts before financing conditions turn.
Analysts remain divided. Jefferies reiterated a $320 price target after the report, calling it a "clean beat across the board," while Guggenheim's John DiFucci maintains a $400 target and calls Oracle his best software idea for 2026. The average Wall Street target sits around $256, implying significant upside from current levels but well below the highs. That dispersion is the market's way of saying the story is not settled.
Oracle has built a formidable position at the center of the AI infrastructure build-out, and this quarter showed the demand is real and the funding model is clever. The clever part is also the risky part: a business model that depends on customers paying upfront for capacity that does not yet exist works beautifully until it doesn't. The $90 billion forecast is not the ceiling of Oracle's AI opportunity — it is the first test of whether the company can finance its way there without the bill coming due first.
Explore more exclusive insights at nextfin.ai.
