The Database Company That Quietly Became AI's Landlord
The 48-year-old database company reinvented itself as the landlord of the AI boom - and a half-trillion-dollar order book to prove it.
In the spring of 1977, three engineers took on a job nobody else wanted: a lapsed database contract, worth about $50,000, for the Central Intelligence Agency. The project had a code name - Oracle. The company that grew out of it kept the name, and for most of the next half-century it did one thing better than anyone: store other people's most important data and refuse to lose it. That reputation for boring reliability made Oracle a fixture of the corporate back office. In 2026 it made the company something stranger - one of the most sought-after suppliers in the artificial-intelligence boom.
Oracle sells the machinery businesses run on. Not the app on your phone, but the database underneath the bank that holds your money, the airline that issues your ticket, the government office that files your taxes. Its flagship product, the Oracle Database, has anchored enterprise computing since 1979. Around it the company has stacked business applications for finance and human resources, the NetSuite suite for smaller firms, the Java programming language, and - the part Wall Street now obsesses over - Oracle Cloud Infrastructure, the data centers and GPU clusters where a growing share of the world's AI models are trained.
What Oracle actually does
Strip away four decades of acquisitions and the pitch is simple. A business generates data - orders, employees, inventory, transactions - and it needs somewhere to keep that data safe, query it fast, and never have it disappear. Oracle's database does that. Its Fusion Cloud Applications then sit on top, handing finance teams their ledgers and HR teams their payroll as software they rent rather than install. NetSuite, acquired in 2016, does the same job for startups and mid-sized companies. And OCI provides the raw computing power - the servers, storage and networking - that all of it runs on, whether the customer is closing the quarterly books or training a language model.
The company's real product, though, is trust in the dullest sense of the word. When a database holds the record of every trade a bank made today, "usually works" is not a feature. Oracle built its business on the assumption that the answer has to be right, and it has to be there at 3 a.m. That is unglamorous, and it is exactly why customers find the software so hard to leave.
Oracle will build more cloud infrastructure data centers than all of its infrastructure competitors combined.
Who pays for it
Oracle's customer list reads like an index of the modern economy: banks and telecoms, retailers and airlines, hospitals and tax authorities, plus a large share of the Fortune 500 and Global 2000. Smaller companies come in through NetSuite. Developers - millions of them - use Oracle Database, Java and MySQL whether or not they ever sign a contract. And in the last two years a new tier of buyer has arrived: AI companies that need enormous blocks of GPU capacity and are willing to commit years and billions of dollars to get it.
That last group is why Oracle's numbers stopped looking like a mature software company's. In its first quarter of fiscal 2026, Oracle signed four multi-billion-dollar cloud contracts and watched its remaining performance obligations - the value of business booked but not yet delivered - jump 359% to roughly $455 billion. Revenue from databases running on rival clouds, through its multicloud partnerships, grew 1,529% in the same period. Those are not the growth rates of a 48-year-old company. They are the growth rates of the thing every 48-year-old company wishes it had built.
The problems it solves
Every large organization faces the same two headaches Oracle sells against. The first is keeping data consistent and available across a business too big to run on spreadsheets - the province of the database and the Autonomous Database, which since 2018 has automated the patching, tuning and security work that used to require a room full of administrators. Oracle calls it "self-driving," and the claim is less marketing than it sounds: the point is to remove the humans from the routine maintenance that used to cause outages.
The second headache is newer. Companies want to use AI on their own data without shipping that data somewhere risky, and they do not want to bet the business on a single model vendor. Oracle's answer is to bring the models to the data. Its AI Database lets customers run their choice of large language model - Google's Gemini, OpenAI's GPT, xAI's Grok - directly on top of information that never leaves the Oracle database. For a bank or a hospital, "your data stays where it is" is not a detail. It is the whole decision.
How it's different
Oracle competes with the biggest names in technology, and it does not win by being the cheapest or the friendliest. Against Amazon Web Services, Microsoft Azure and Google Cloud, its edge is the database everyone already runs and the willingness to do something its rivals resisted for years: put Oracle's own database inside their clouds. Oracle Database@Azure, @Google Cloud and @AWS let a customer keep the Oracle database they depend on while using whichever cloud they prefer. It is a strange, almost cooperative move for a company famous for its combativeness - and it is working.
Against enterprise-application rivals like SAP, Salesforce and Workday, Oracle's argument is integration: buy the database, the applications and the infrastructure from one vendor and let them work as a set. Against the database upstarts - Snowflake, Databricks, MongoDB, PostgreSQL - its argument is 40 years of hardened reliability and the sheer cost of migrating away. None of this is subtle. Oracle's differentiation is that it is already inside the building, and prying it out is expensive enough that most customers decide not to try.
Products and services
The catalog is deep, but a handful of names carry the company. The Oracle Database remains the crown jewel, now in its AI-infused 23ai generation. The Autonomous Database is its self-managing form. OCI is the cloud platform underneath. Fusion Cloud Applications cover ERP, HR, supply chain and customer experience for large enterprises, while NetSuite serves smaller ones. From the 2010 purchase of Sun Microsystems came Java and MySQL, two of the most widely used pieces of software on earth. And Exadata, Oracle's engineered hardware, is built to run all of it at speed.
The business model
Oracle makes money four ways: cloud services and license support, software licenses, hardware, and services. The mix has shifted hard toward the first. What was once a business of selling perpetual licenses and charging for support has become a business of recurring subscriptions - Fusion and NetSuite billed monthly - and consumption-based infrastructure, where customers pay for the OCI capacity they use. The giant AI contracts sit inside that last bucket, which is why they show up as backlog long before they show up as revenue. The model has a catch worth naming: those data centers cost tens of billions of dollars to build up front, and Oracle is spending that money now against revenue it expects to collect over years.
Oracle's moat was never the software. It was that nobody wants to be the person who moved the database and broke the bank.
The expertise
Oracle is an engineering company run, at the top, by an engineer who never stopped competing. Larry Ellison co-founded it in 1977, took it public in 1986, and at 80-something still holds the chairman and CTO titles and shapes the technical direction. Safra Catz, CEO since 2014, supplies the financial discipline - the deal-making and margin obsession that turned Ellison's ambitions into a durable balance sheet. The company's deepest expertise is unfashionable and hard to copy: making software that does not fail at scale, and knowing exactly how much an enterprise will pay to be sure of it.
Where it fits
For years the conventional wisdom filed Oracle under "legacy" - a giant coasting on 1990s contracts and famous mostly for aggressive license audits. The AI era rewrote that. It turned out the least glamorous assets in technology - data centers, databases, the plumbing between them - were exactly what the boom needed, and Oracle had spent 30 years building them. The company now sits in an unusual spot: old enough to be trusted with a bank's core systems, and suddenly central to the newest thing in computing. In 2026 that combination briefly made co-founder Larry Ellison the second-richest person in the world. The database company, it turned out, had been standing on the right ground the whole time.