ORCL Enterprise software · Mega cap · Cloud · AI infrastructure · Thesis updated July 12, 2026

Oracle’s AI backlog tests its balance sheet

Oracle’s story now centers on $638 billion of remaining performance obligations, meaning contracted revenue not yet booked.
Cloud infrastructure revenue grew 77% in fiscal 2026, making OCI the main growth engine.
The same build-out burned cash, with fiscal 2026 capital expenditures of $55.7 billion and free cash flow of $(23.7) billion.
Cloud and Software made up 87% of fiscal 2026 revenue, while Hardware and Services were much smaller.
The stock needs Oracle to turn backlog into real revenue without big delays, cost overruns, or painful dilution.

01 Running thesis

A giant backlog, a giant bill

Oracle has become a high-stakes AI infrastructure story. Remaining performance obligations, or contracted revenue not yet booked, reached $638 billion at May 31, 2026. That is the core bull case. If Oracle can turn that backlog into revenue, its cloud business could look very different in a few years.

The early signs are strong. Cloud infrastructure revenue grew 77% in fiscal 2026, and cloud infrastructure drove most of the growth in cloud revenue. Oracle is also pushing its AI Database, which helps customers use their private data with large language models inside Oracle Cloud Infrastructure, or OCI.

The bear case is the cost of getting there. Fiscal 2026 capital expenditures were $55.7 billion, up from $21.2 billion in fiscal 2025. Free cash flow was $(23.7) billion. That makes this less like a normal software story and more like a huge construction project tied to a few very large AI customers.

Finn’s view is cautious because both sides can be true. Oracle may have won one of the largest cloud opportunities in the market. It also must finance, build, and fill a large amount of data center capacity before investors can know how profitable the contracts really are.

Jun 2026▼Fiscal 2026 results raised the stakes. RPO reached $638 billion, but capital expenditures hit $55.7 billion and free cash flow fell to $(23.7) billion.

Mar 2026▼The 10-Q showed RPO of $552.6 billion and nine-month capital expenditures of $39.2 billion. It also made dilution risk explicit through preferred stock and the ATM program.

Mar 2026▲Management said it signed more than $29 billion of contracts using bring-your-own-hardware and upfront payment structures. That helped answer part of the funding concern, while execution risk remained high.

Dec 2025→The 10-Q confirmed the prior quarter’s story, with RPO at $523.3 billion and trailing four-quarter free cash flow of $(13.2) billion. The thesis stayed focused on backlog conversion.

Dec 2025→RPO rose to $523.3 billion, helped by contracts with Meta, NVIDIA, and others. Management also raised fiscal 2026 capital spending expectations by about $15 billion, keeping the risk and reward both high.

Sep 2025→The first-quarter 10-Q matched the earnings message. Cloud and software was 86% of revenue, and RPO stood at $455.3 billion.

Sep 2025▲Oracle reported a $317 billion sequential jump in RPO to more than $455 billion. The story shifted from whether Oracle can win cloud demand to whether it can build enough capacity.

Jun 2025▼The fiscal 2025 10-K showed the first clear cash strain from the cloud push. Capital expenditures were $21.2 billion and free cash flow was $(394) million.

02 Business model

Databases meet data centers

Oracle makes money by selling software and cloud services to businesses. Its base includes database software, enterprise apps, software support, and cloud infrastructure. The older software support base is slower, but it gives Oracle a large pool of customers to move into cloud services.

OCI is the faster part. Customers pay Oracle for computing power, storage, networking, and database services in Oracle’s data centers. AI customers also need large amounts of capacity, especially GPUs and related hardware, which makes growth possible but expensive.

Oracle’s AI Database is meant to link the old and new businesses. It lets companies turn private business data into a form AI models can search, called vectors. Oracle then hosts leading large language models inside OCI so customers can ask AI questions without moving sensitive data into less controlled systems.

The model breaks if the build-out outruns demand or cash. Oracle has said some newer contracts can use customer upfront payments or bring-your-own-hardware structures. Even so, the fiscal 2026 cash burn shows that investors need proof that these structures can scale.

03 Product portfolio

What Oracle sells

Growth engine

Oracle Cloud Infrastructure

OCI provides cloud computing, storage, networking, and AI infrastructure. It is the key driver, with cloud infrastructure revenue up 77% in fiscal 2026.

Option

Oracle AI Database

This product helps customers vectorize private enterprise data so AI models can use it for inferencing. The bet is that AI on private data becomes a large long-term use case.

Steady

Cloud applications

Oracle sells business software for finance, human resources, industry workflows, and other core functions. Cloud applications revenue grew 11% to $15.9 billion in fiscal 2026.

Cash cow

Software support

Support revenue comes from customers that keep using Oracle software. It grew 1% in fiscal 2026, so it is not the main growth engine, but it remains a large recurring base.

Steady

Software licenses

Traditional license sales fell 9% in fiscal 2026. This shows the business mix moving away from one-time software sales and toward cloud consumption.

Steady

Hardware and Services

These are smaller businesses tied to Oracle systems and customer support work. Together they made up 13% of fiscal 2026 revenue.

04 Business segments

Revenue mix is cloud heavy

Cloud and Software87%growing fast

Hardware5%flat

Services8%modest

The segment mix is from fiscal 2026 revenue. Cloud and Software was 87% of revenue, while Hardware was 5% and Services was 8%, but some OCI offerings may be concentrated in a small number of large customers.

05 Risk factors

What could go wrong

Backlog does not convert fast enough

High impact · Medium odds

The $638 billion remaining performance obligation is only valuable if Oracle can turn it into revenue on time. Data centers must be built, powered, staffed, and connected before customers can use them. Any delay can push revenue out while costs keep running.

We watchQuarterly RPO, cloud infrastructure revenue growth, and management comments on when new data center capacity comes online.

Cash burn stays too high

High impact · High odds

Oracle spent $55.7 billion on capital expenditures in fiscal 2026 and free cash flow was $(23.7) billion. The 10-K says the upward trend in capital spending is expected to continue in fiscal 2027 and later years. If cash flow does not recover, the balance sheet and shareholder returns could come under pressure.

We watchQuarterly capital expenditures, free cash flow, debt issuance, and any updated fiscal 2027 capital spending outlook.

Large customer concentration

High impact · Medium odds

Oracle says certain OCI offerings are more concentrated among a number of large customers. That matters because the backlog appears tied to very large AI contracts. If a top customer slows, renegotiates, or changes its AI plan, Oracle could be left with costly capacity that is not fully used.

We watchDisclosure on top customers, new large contract wins, contract cancellations, and customer-funded capacity details.

Dilution from financing

Medium impact · Medium odds

Oracle has warned that conversion of its Series D Mandatory Convertible Preferred Stock or issuance under its at-the-market program will dilute common shareholders. Dilution means each current share owns a smaller piece of the company. That risk rises if the cloud build-out needs more outside funding.

We watchShare count, preferred stock conversion updates, ATM program sales, and EPS impact.

Hardware supply and cost pressure

Medium impact · Medium odds

AI data centers need accelerators, networking gear, power equipment, and other scarce parts. Industry demand is intense. If Oracle cannot get hardware on time or at good prices, margins and delivery schedules can suffer.

We watchManagement comments on GPU supply, data center component availability, gross margin on AI capacity, and delivery timing.

Hyperscaler competition and AI shifts

Medium impact · Medium odds

Oracle competes with larger cloud platforms. Its RPO growth shows it is winning important work, but customers can still shift workloads over time. AI rules, model choices, or customer needs may also change in ways that hurt demand for today’s infrastructure plans.

We watchOCI win rates, multi-cloud database growth, AI regulation updates, and customer moves toward competing clouds.

06 Quick answers

In one breath

Why is Oracle spending so much on data centers?

Oracle has signed large cloud and AI infrastructure contracts, and it needs more data center capacity to serve them. Fiscal 2026 capital expenditures reached $55.7 billion mainly because of data center expansion.

What does Oracle’s RPO mean for investors?

RPO means remaining performance obligations, or contracted revenue that has not yet been recognized. Oracle’s $638 billion RPO gives strong revenue visibility, but it still has to be converted into actual services and revenue.

Is Oracle still mainly a database company?

The database is still central, but Oracle is now also a cloud infrastructure company. The AI Database strategy tries to use Oracle’s database strength to drive more OCI consumption.

What is the main debate around Oracle stock?

The bull case is that Oracle has locked in huge AI cloud demand. The bear case is that turning that demand into profit needs massive spending, strong execution, and possible shareholder dilution.