Key Takeaways
Oracle's fiscal year 2025 (ended May 31, 2025) marked the definitive pivot from database incumbent to cloud infrastructure contender — a transformation driven by the surprise emergence of Oracle Cloud Infrastructure (OCI) as the preferred AI training cloud for large language model developers who found AWS, Azure, and GCP capacity constrained during the 2024-2025 GPU procurement surge. Total revenue reached approximately $57-58B, growing approximately 11-12% from FY2024's $52.9B, as Cloud Services and License Support (the recurring revenue base) grew approximately 12-14% to approximately $44B and Cloud License and On-Premise License grew modestly. Non-GAAP EPS reached approximately $6.30-6.60, growing approximately 10-12% from FY2024's $5.93, with non-GAAP operating margins at approximately 42-44% as OCI's rapid revenue ramp began amortizing the fixed infrastructure investment more efficiently. The defining inflection of FY2025 was the Remaining Performance Obligation (RPO) backlog reaching approximately $130B+ — a tripling from FY2023 levels — driven by multi-year GPU cloud contracts with xAI (Elon Musk's AI company), Cohere, and other LLM developers who signed Oracle GPU capacity agreements of $1B-$10B+ over 3-5 years. FY2026 tests whether this RPO converts to recognized revenue at the pace management has guided, and whether Oracle's database cloud migration — moving the on-premise Oracle Database installed base to Oracle Autonomous Database and MySQL HeatWave on OCI — generates the database modernization revenue that justifies the $65B+ Cerner acquisition in terms of total cloud portfolio.
Oracle was founded in 1977 by Larry Ellison, Bob Miner, and Ed Oates in Santa Clara, California, as Relational Software Inc., building the first commercial relational database management system based on IBM's System R research papers. Oracle Database became the dominant enterprise RDBMS through the 1980s-2000s by running on every major enterprise hardware platform and building an ecosystem of ERP, CRM, and analytics software that integrated tightly with Oracle Database — creating vendor lock-in that generated extraordinarily durable maintenance and license revenue. Larry Ellison remains executive chairman and CTO and retains approximately 40%+ of Oracle shares, making him the primary decision-maker for all major strategic moves. CEO Safra Catz has managed Oracle's financial discipline since 2014 — the aggressive share buybacks, the capital structure, the pricing strategy — while Ellison drives the technology roadmap. The $28.3B acquisition of Sun Microsystems (2010) and the $9.3B acquisition of PeopleSoft (2005) were the foundational cloud-precursor moves; the $28B acquisition of Cerner (2022) and the $15.7B acquisition of Fusion (pre-IPO, internal development) represent Oracle's bet that healthcare EHR cloud and HCM/ERP SaaS can complement the infrastructure cloud.
Business Structure
Oracle reports three primary revenue streams.
Cloud Services and License Support (~$44B, ~77% of total in FY2025): The largest and most recurring component, comprising Oracle Cloud (Infrastructure + Applications SaaS) and legacy on-premise database/software maintenance contracts. Cloud Infrastructure (OCI) — IaaS and PaaS — is growing at approximately 50%+ annually as AI workload capacity comes online. Oracle Cloud Applications (Fusion ERP, NetSuite ERP, Healthcare/Cerner) grow approximately 15-20% annually. Legacy on-premise support contracts (approximately $18-20B annually) are extraordinarily stable but slowly declining as customers migrate to cloud.
Cloud License and On-Premise License (~$5B, ~9%): One-time or term license fees for Oracle Database, middleware, and applications. This segment has been in structural decline since the cloud migration began; the revenue is valuable (it feeds the support stream) but not a growth driver.
Hardware (~$3B, ~5%): Engineered systems (Exadata Cloud@Customer, OCI Dedicated Region) and hardware support. Oracle has strategically used Exadata as a cloud migration bridge — Exadata Cloud@Customer brings Oracle's cloud operating model to customer data centers, enabling compliance-constrained customers to access Oracle cloud economics without moving data off-premise.
Services (~$5B, ~9%): Consulting, managed cloud services, and education. Lower margin than the license/support business.
Key Core Metrics Performance
Revenue and Cloud Growth (FY2021–FY2025)
Oracle's revenue growth re-accelerated from the FY2021-FY2022 stagnation as OCI's AI cloud capacity build attracted GPU demand.
| Fiscal Year | Total Revenue | Cloud Services+Support | Cloud License | YoY Growth |
|---|---|---|---|---|
| FY2021 (ended May 2021) | $40.5B | ~$28.7B | ~$5.4B | +4.1% |
| FY2022 (ended May 2022) | $42.4B | ~$30.9B | ~$5.5B | +4.7% |
| FY2023 (ended May 2023) | $49.6B | ~$36.7B | ~$5.1B | +17.1% (Cerner full year) |
| FY2024 (ended May 2024) | $52.9B | ~$39.4B | ~$4.9B | +6.7% |
| FY2025 (ended May 2025) | ~$57.5B | ~$44.0B | ~$5.0B | ~+8.7% |
The FY2023 revenue step-up reflects Cerner's first full consolidation year (~$6B revenue). OCI cloud infrastructure has grown from approximately $1.5B in FY2022 to approximately $8-9B in FY2025.
Non-GAAP Profitability (FY2021–FY2025)
Oracle's business model generates exceptional cash margins from the legacy database installed base, which subsidizes OCI's infrastructure investment.
| Fiscal Year | Revenue | Non-GAAP Op. Margin | Non-GAAP EPS | FCF |
|---|---|---|---|---|
| FY2021 | $40.5B | 44.4% | $4.67 | ~$13B |
| FY2022 | $42.4B | 42.6% | $4.90 | ~$8B |
| FY2023 | $49.6B | 40.7% | $5.28 | ~$10B |
| FY2024 | $52.9B | 42.7% | $5.93 | ~$11B |
| FY2025 | ~$57.5B | ~43.0% | ~$6.45 | ~$12B |
Oracle's non-GAAP operating margins have sustained above 40% through the OCI build-out, reflecting the cross-subsidy from the legacy database maintenance annuity. FCF remains strong despite elevated capex (approximately $8-10B annually for OCI data center build-out).
Remaining Performance Obligation (RPO) — AI Cloud Backlog (FY2022–FY2025)
The RPO explosion is the most important leading indicator for Oracle's future revenue trajectory.
| Fiscal Year | Total RPO | Cloud RPO | YoY Growth |
|---|---|---|---|
| FY2022 | ~$36B | ~$24B | +15% |
| FY2023 | ~$64B | ~$43B | +78% |
| FY2024 | ~$98B | ~$73B | +53% |
| FY2025 | ~$130B+ | ~$100B+ | ~+33% |
The RPO tripled in two years as xAI, Cohere, and other AI-native companies signed multi-year OCI GPU capacity agreements — Oracle's ability to deliver GPU clusters faster than hyperscalers (due to its earlier infrastructure commitment and more flexible contracting) created a unique competitive window in FY2024-FY2025.
Market Evaluation
Oracle trades at approximately 22-28x forward non-GAAP EPS — above its historical 15-18x range, justified by the RPO-driven revenue acceleration narrative. The bull case is RPO conversion: if the $130B+ backlog converts at the guided 2-3 year revenue recognition timeline, Oracle's cloud revenue alone reaches $25-30B by FY2027, pushing total revenue toward $70-75B and EPS toward $8-9 as OCI's fixed cost base is better amortized. The legacy database installed base provides a natural OCI migration funnel — Oracle Database customers migrating to Autonomous Database on OCI are "moving within the Oracle ecosystem" rather than re-platforming, dramatically lowering migration risk. The bear case is that RPO is not revenue: multi-year GPU cloud contracts can be renegotiated, reduced, or cancelled if AI infrastructure demand plateaus (as it may if hyperscalers bring their own training capacity online faster than expected). Cerner's EHR business has underperformed Oracle's acquisition thesis — healthcare cloud is a slower-growth, lower-margin market than anticipated — and represents an approximately $10B+ revenue overhang that compresses blended cloud growth rates.
OCI AI Cloud and the GPU Cluster Advantage
Oracle Cloud Infrastructure built its AI compute capacity through a counterintuitive strategy: while AWS, Azure, and GCP were constrained by GPU allocation systems that prioritized existing customers and long-term contracts, Oracle offered custom GPU cluster configurations with no minimum commitment minimums and delivered them in record time by building bare-metal AI clusters that could be spun up in weeks rather than months. xAI's Colossus cluster — the world's largest GPU cluster at the time of deployment — was built on OCI at Oracle's Memphis data center, giving Oracle a landmark reference customer that validated the company's ability to deliver at hyperscale AI infrastructure.
Oracle's technical differentiation in AI infrastructure centers on RDMA network fabric — ultra-low latency, high-bandwidth interconnects that allow GPU-to-GPU communication at speeds comparable to NVLink within a node but at cluster scale across thousands of GPUs. This is critical for training large frontier models where communication overhead between GPUs is a primary bottleneck. Oracle's cluster network achieves approximately 1.6 Tbps per GPU with microsecond latency, competitive with NVIDIA's DGX SuperPOD reference architecture. Combined with Oracle's flexible contracting (customers can take partial clusters rather than entire rack commitments) and its AI-native startup customer base (which prioritizes speed-to-compute over cloud ecosystem lock-in), OCI has created a defensible niche in AI infrastructure that justifies the RPO explosion and positions Oracle for continued enterprise AI cloud growth.