Oracle's $638B RPO Is Not a Revenue Number. It's a Commitment Machine.

The Number That Matters Is Not Revenue

Oracle reported 93% year-over-year growth in cloud infrastructure revenue on June 10, 2026. The more consequential figure is $638 billion — Oracle's Remaining Performance Obligations at the close of FY2026 Q4, up 363% year-over-year and up $85 billion in a single quarter.

RPO is contracted future revenue: money customers have legally committed to spend. Oracle's RPO finished at $638 billion, up 363%, and management described it as providing "exceptional visibility" into future revenue growth, supported by long-term contractual customer commitments reflecting strong demand across both AI infrastructure and cloud services. That is a backlog nearly ten times Oracle's quarterly total revenue. It signals something deeper than a strong earnings beat: enterprises have moved from buying compute on demand to reserving it years in advance.

What Drove the Surge

Most of the RPO increase in both Q3 and Q4 came from large-scale AI contracts where customers either prepaid Oracle for the purchase of GPUs, or the customer bought and supplied the GPUs to Oracle. The prepaid and customer-supplied hardware portions of those AI contracts now total $75 billion, which substantially reduces the capital Oracle must raise to build out its AI data centers.

This inverts the traditional cloud model. Customers are not consuming capacity and paying monthly. They are financing Oracle's build-out in exchange for guaranteed access. Cloud infrastructure revenue of $5.8 billion reflects customers choosing Oracle Cloud Infrastructure for both AI workloads and database migration, with 97.5% global GPU utilization confirming that supply is still falling well short of demand. When utilization runs that tight, spot capacity evaporates. Long-term contracts become the only path to guaranteed compute.

Three forces collided: GPU scarcity made capacity reservation a competitive necessity. Enterprise AI deployment timelines accelerated, pushing procurement from "when we need it" to "before someone else claims it." And Oracle structured its contracts to attract prepayment through commitment-based pricing that spot buyers cannot access.

The OpenAI Concentration Problem

Bank of America analysts noted that over 50% of Oracle's remaining performance obligation comes from OpenAI. This reshapes the risk profile. A $638 billion backlog built predominantly on one counterparty is not diversification — it is leverage on OpenAI's capital access, revenue trajectory, and continued reliance on Oracle infrastructure. Wall Street's focus remains squarely on Oracle's massive RPO, contracted future revenue largely anchored by its partnership with OpenAI for the $500 billion "Stargate" supercomputer initiative, but anchoring so much future revenue to a single cash-burning AI startup creates massive concentration risk if the broader AI boom cools.

The implication cuts both ways. It reveals Oracle's exposure and signals where the largest AI infrastructure commitments are flowing. If OpenAI consumes more than half of Oracle's contracted backlog, then a meaningful portion of frontier AI training compute runs through OCI.

The Conversion Timeline Is the Real Question

Oracle expects 12% of its RPO to be recognized in the next 12 months, and another 34% between 13 and 36 months, with both percentages expected to accelerate over coming quarters based on current long-term outlook. Do the math: 12% of $638 billion equals roughly $77 billion in revenue Oracle expects to recognize in the next year alone. That dwarfs Oracle's entire FY2026 revenue of $67.4 billion. The conversion rate, not the headline RPO figure, is what matters quarter to quarter.

Management expects a massive acceleration in capacity delivery, with Q1 FY2027 alone approaching one gigawatt — nearly equaling the total capacity delivered in the prior four quarters — and revenue and earnings are anticipated to accelerate in the second half of fiscal 2027 as more megawatts at large-scale data center sites come online.

The Capex Paradox

Oracle's revenue visibility is extraordinary. Its cash position is not. Free cash flow was negative $23.7 billion for fiscal year 2026 as Oracle continued to execute on investments to support the growth of its cloud infrastructure business. On the earnings call, the company said it expects to spend $70 billion in net capex next year, with a total capex of $90 billion to $95 billion after including customer prepayments of $20 billion to $25 billion.

Oracle raised $48 billion in FY2026 via debt and equity and plans approximately $40 billion more in financing in FY2027. Oracle uses fixed-price contracts when costs are known, but when supply-chain or future cost uncertainty is too high, it uses mechanisms that can pass costs through — a provision that helps prevent Oracle from absorbing reduced margins when component costs rise. For customers negotiating today, the terms embedded in multi-year agreements — particularly cost pass-through clauses — will determine whether those contracts remain economically rational three years hence.

What This Means for Operators

Enterprises that have not yet signed long-term cloud infrastructure agreements operate in a different market than those who locked in capacity a year ago. Cloud revenues combining IaaS and SaaS increased 47% to $9.9 billion, driven by 93% growth in cloud infrastructure, while software revenues fell 2% to $6.8 billion, reflecting customers' continuing migration from on-premise software to the cloud. On-premise is contracting. Cloud capacity is constrained. The window for favorable multi-year terms may be narrowing as Oracle's uncommitted capacity shrinks.

The 363% RPO growth also reframes Oracle's competitive position against AWS and Azure. Amazon Web Services, the leader in the market, generated $37.59 billion in revenue in the March quarter — roughly 6.5x Oracle's quarterly cloud infrastructure revenue. Raw revenue scale differs from committed backlog growth, though. If Oracle's RPO trajectory reflects a genuine structural shift in enterprise AI compute commitments, hyperscalers will face pressure to match or lose enterprise anchors to providers willing to lock in capacity now.

One critical unknown: the breakdown of Oracle's RPO between GPU-intensive AI infrastructure contracts and traditional database, ERP, and SaaS workloads. Oracle has not disclosed that split publicly. Until it does, the $638 billion figure carries the caveat that the AI-driven portion may be smaller — or more concentrated — than the aggregate suggests.

What to Watch

1. RPO conversion rate, Q1 FY2027. The first test of whether 12% of $638 billion actually recognizes in the next twelve months. A miss would force a hard revaluation of the backlog thesis.

2. AWS and Azure RPO disclosures. Neither hyperscaler reports RPO with Oracle's granularity. Any shift in their disclosure practices or long-term commitment metrics in upcoming earnings matters.

3. OpenAI financial stability. If more than 50% of Oracle's backlog traces to a single customer, OpenAI's fundraising trajectory and revenue growth become leading indicators for Oracle's RPO durability.

4. Gross margin trajectory in H1 FY2027. Management said FY2027 gross margin will step down because of data center ramp timing and mix, and expects infrastructure margins to improve as data centers reach full contractual revenue contribution. The spread between when capacity costs hit and when contracted revenue recognizes will define Oracle's near-term financial health.

5. Contract terms in competitor bids. If AWS and Azure begin offering prepayment or customer-supplied hardware structures matching Oracle's model, the competitive moat narrows. Watch enterprise procurement patterns in Fortune 500 Q3 capex disclosures.