Oracle Corporation

Oracle Corporation
Type	Public
Traded as	NYSE: ORCL; S&P 500 component
Industry	Enterprise software; Cloud computing; Databases
Founded	16 June 1977, as Software Development Laboratories, Santa Clara, California
Founder	Larry Ellison; Bob Miner; Ed Oates
Headquarters	Austin, Texas, United States
Key people	Larry Ellison (Chairman and chief technology officer); Clay Magouyrk and Mike Sicilia (Co-chief executive officers); Safra Catz (Executive vice chair)
Products	Oracle Database, Oracle Cloud Infrastructure, Fusion Cloud Applications, NetSuite, Java, MySQL, Oracle Health
Employees	Approximately 141,000 (2026)
Website	oracle.com

Oracle Corporation is an American multinational technology company headquartered in Austin, Texas. It is among the largest software companies in the world and has been since the 1990s, built on the relational database that carries its name and extended over five decades into enterprise applications, hardware, and most recently into cloud infrastructure for artificial intelligence.

The company was founded on 16 June 1977 in Santa Clara, California, as Software Development Laboratories by Larry Ellison, Bob Miner and Ed Oates. It renamed itself after its principal product, which was the first commercially available relational database system to implement SQL, and which became the standard on which a very large proportion of the world's corporate and governmental record-keeping was built.

Oracle's history is unusually adversarial for a software company. It has conducted a long sequence of hostile takeovers, most famously of PeopleSoft in 2004 and 2005, litigated aggressively over intellectual property, most consequentially against Google over the Java application programming interfaces in a case that reached the Supreme Court, and competed publicly and personally with Salesforce, SAP, IBM and Amazon Web Services. Ellison's practice of attacking competitors by name from the stage at the company's conferences is a defining feature of its public identity.

Since 2024 the company has been reshaped by demand for artificial intelligence computing. It is a principal participant in Stargate, the data centre venture announced in January 2025 with OpenAI, SoftBank and MGX, and signed a contract with OpenAI in September 2025 to supply US$300 billion of computing over five years. Its remaining performance obligations, the value of contracted work not yet delivered, reached approximately US$130 billion in the second quarter of its 2026 financial year.

History

Founding and the relational database, 1977 to 1986

Larry Ellison was working at Ampex on a project for the Central Intelligence Agency, code-named Oracle, when he read Edgar F. Codd's 1970 paper "A Relational Model of Data for Large Shared Data Banks", published at IBM's San Jose laboratory.

Codd's proposal was that data should be organised as relations, that is tables, and queried by describing what was wanted rather than how to retrieve it, which separated the logical structure of information from its physical storage. IBM understood the significance and built System R to explore it, but was reluctant to commercialise a product that would undermine its existing IMS hierarchical database, which was extremely profitable.

Ellison, together with his Ampex colleagues Bob Miner and Ed Oates, concluded that the reluctance was an opportunity. They founded Software Development Laboratories on 16 June 1977 with about US$2,000, of which Ellison contributed US$1,200. Miner was the principal engineer and wrote much of the early code; Ellison sold.

Version 2 was released in 1979, there being no version 1 on the reasoning that nobody buys a first release, and it was the first commercially available relational database implementing SQL, IBM's own SQL/DS following in 1981. The company was renamed Relational Software Inc., then Oracle Systems in 1982, then Oracle Corporation.

The product's early advantage was portability. Oracle was written to run on many operating systems and many machines at a time when database software was generally tied to a manufacturer's hardware, and the strategy positioned the company to benefit from every shift in the underlying platform. Version 3, released in 1983, was rewritten in C, which made this possible.

Oracle listed on Nasdaq on 12 March 1986, one day before Microsoft.

Near-failure and recovery, 1990 to 1995

Growth through the late 1980s was extremely rapid and, as it emerged, partly fictitious. Oracle's sales organisation booked revenue on contracts before delivery and in some cases before the products worked, a practice encouraged by compensation arrangements that rewarded bookings rather than collections.

In 1990 the company restated its results, reported its first quarterly loss, lost roughly 80 per cent of its market value, and came close to defaulting on its debt. The Securities and Exchange Commission investigated and the company settled without admitting wrongdoing. Ellison has described the period as the closest Oracle came to failure and has said that the fault was his, in that he had built an organisation that measured the wrong thing.

The recovery was conducted largely by Raymond Lane, who joined as president in 1992 and professionalised the sales and consulting organisations, and by Jeff Henley as chief financial officer. Oracle 7, released in 1992, was the product that established the company's technical dominance for the following decade, introducing stored procedures, triggers, declarative referential integrity and a substantially better optimiser.

By the mid-1990s Oracle held the largest share of the relational database market and had begun to build the applications business that would eventually rival it in size, entering enterprise resource planning against SAP and gradually adding financials, manufacturing, human resources and customer relationship management.

Ellison's public positioning during this period centred on the network computer, a thin client he promoted from 1995 as the successor to the personal computer and as a direct attack on Microsoft, whose dominance he opposed with an intensity that was personal as well as commercial. The network computer failed commercially and the underlying argument, that computing would move to servers accessed over networks, was correct about fifteen years early.

The acquisition decade, 2004 to 2014

Oracle's transformation from a database company into a software conglomerate was accomplished by purchase, and the strategy was articulated explicitly: enterprise software customers do not switch suppliers easily, so the way to grow is to buy the supplier.

The campaign opened with PeopleSoft, which Oracle pursued from June 2003 in an 18-month hostile takeover that is among the most contested in technology. PeopleSoft's board resisted, adopted a customer assurance programme promising refunds of several times licence fees if the company were acquired and support withdrawn, which was designed to make the target expensive, and dismissed its own chief executive Craig Conway during the fight. The Department of Justice sued to block the deal on competition grounds and lost at trial in September 2004. Oracle completed the purchase in January 2005 for approximately US$10.3 billion and cut about 5,000 jobs.

Siebel Systems, the customer relationship management company founded by the former Oracle executive Tom Siebel, followed in 2006 for about US$5.85 billion. Hyperion came in 2007, BEA Systems in 2008 for approximately US$8.5 billion after a public dispute over price, and Sun Microsystems in January 2010 for approximately US$7.4 billion.

The Sun purchase was the most consequential and the most debated. It brought hardware, which Oracle had never sold, the Solaris operating system, the MySQL database, which competed with Oracle's own at the low end, and Java, which by then underpinned an enormous proportion of enterprise software and of the Android operating system. Oracle's handling of the open-source assets was contentious: the OpenOffice community forked to LibreOffice, the Hudson project forked to Jenkins, and MySQL's creator Michael Widenius forked it to MariaDB, each out of concern that Oracle would not maintain the projects in the community's interest.

Further purchases included Taleo and RightNow in 2012, Responsys and MICROS Systems in 2014, NetSuite in 2016 for about US$9.3 billion, a company Ellison had funded at its founding and in which he held a large stake, which drew shareholder objection over the conflict of interest, and Cerner in 2022 for approximately US$28.3 billion, the largest acquisition in Oracle's history, which brought electronic health records and a very large position in American hospitals.

Safra Catz and Mark Hurd, the latter having arrived from Hewlett-Packard in 2010 after resigning there over an expenses matter, were appointed co-chief executives in September 2014 when Ellison moved to executive chairman and chief technology officer. Hurd died in 2019 and Catz continued alone.

Cloud, 2015 to 2023

Oracle was late to cloud computing and initially dismissive of it. Ellison mocked the term publicly in 2008, saying that the computer industry was the only one more fashion-driven than women's fashion and that cloud was simply a new word for what the industry already did. The remark is frequently quoted against him and was, in substance, a defensible observation about marketing that was mistaken about the direction of the business.

The company reversed and invested heavily. Its first generation of infrastructure was not competitive, and a second generation, Oracle Cloud Infrastructure, was built from 2016 under engineers recruited principally from Amazon Web Services, including Clay Magouyrk and Don Johnson. The design decisions distinguished it from its rivals: a flat, non-oversubscribed network; bare metal instances; and simpler, lower egress pricing, which addressed a specific and widely resented aspect of competitors' billing.

The applications business moved to software as a service under the Fusion Cloud banner, and NetSuite served smaller organisations, with the combination eventually becoming the larger part of the company's cloud revenue.

Oracle's other prominent activity in this period was litigation. Oracle America, Inc. v. Google LLC ran from 2010 to 2021 over Google's reimplementation of 37 Java application programming interfaces in Android. Oracle claimed billions of dollars. The case turned on two questions: whether the declaring code of an API is copyrightable, and if so whether Google's use was fair. The Federal Circuit twice found for Oracle; the Supreme Court held in April 2021, by six to two, that Google's use of the declarations was fair use, declining to decide the copyrightability question. The outcome was greeted with relief across the software industry, since a contrary ruling would have made reimplementation of interfaces a copyright matter.

Oracle also joined the consortium that took a stake in TikTok's American operations from 2020, becoming its cloud provider under the arrangement known as Project Texas, in which TikTok's American user data was moved to Oracle infrastructure and its code reviewed by Oracle personnel.

Artificial intelligence and the expansion, 2024 to 2026

The demand for computing to train and operate artificial intelligence transformed Oracle's infrastructure business from a distant fourth to a central supplier.

The advantages were partly accidental. Oracle's network design, built for database workloads that are sensitive to latency, suited the collective communication patterns of large model training, and the company had spare capacity and a willingness to sign very large contracts at a time when the established providers were allocating capacity to their own products. NVIDIA clusters at Oracle were used by OpenAI, xAI, Cohere, Meta and others.

In April 2024 Oracle announced that it would move its world headquarters from Austin to a new complex in Nashville, Tennessee, without giving a timeframe, and that the Nashville hub would be its largest office with about 8,500 jobs. In June 2024 it announced a US$1 billion investment in Spain, creating a cloud region in Madrid with Telefónica.

On 21 January 2025 President Donald Trump announced Stargate at the White House, a joint venture of Oracle, OpenAI, SoftBank and MGX committing US$500 billion over four years to artificial intelligence infrastructure in the United States, with Oracle building and operating much of the physical capacity. The first site at Abilene, Texas, is among the largest computing facilities ever built. Oracle also announced approximately US$5 billion of investment across the United Kingdom, Germany and the Netherlands in 2025.

In September 2025 Oracle signed the contract to supply OpenAI with US$300 billion of computing over five years, an agreement large enough to move Oracle's own share price by tens of per cent on disclosure and to make its backlog the most closely watched figure in enterprise software.

Also in September 2025 the company promoted Clay Magouyrk, president of cloud infrastructure, and Mike Sicilia, president of industries, to be co-chief executives, with Safra Catz moving to executive vice chair after eleven years in the role.

In January 2026 TikTok finalised the change of ownership of its American operations, with Oracle, MGX and Silver Lake each holding 15 per cent and ByteDance retaining 19.9 per cent. Oracle's stake was estimated at about US$2 billion in March 2026.

In March 2026 Oracle signed a voluntary ratepayer protection pledge under which data centre operators commit to building or buying the energy their facilities need at rates distinct from those charged to ordinary consumers, a response to political pressure over rising household electricity bills in areas with heavy data centre construction. In the same month it announced plans for 72 multicloud data centres embedded within Microsoft, Amazon and Google cloud regions, and reported more than 200 live or in-progress cloud regions, more than any competitor. Multicloud database revenue grew 817 per cent in the second quarter of its 2026 financial year.

The expansion was accompanied by large reductions in staff. In April 2026 it was reported that nearly 12,000 employees had been let go in India, and in May 2026 the company's annual filing showed 141,000 full-time employees, down from 162,000 a year earlier, a reduction of approximately 13 per cent.

For the second quarter of financial 2026, ending November 2025, Oracle reported revenue of US$14.1 billion with cloud infrastructure revenue growing 68 per cent year on year to US$4.1 billion.

Products

Oracle Database

The Oracle Database remains the company's technical foundation and, for most of its history, the largest single source of its profit.

Major releases have run from Version 2 in 1979 through Oracle 7 in 1992, Oracle 8 and 8i, which added internet and object features, 9i, 10g in 2004, whose g denoted grid computing, 11g, 12c for cloud, 18c and 19c, 21c, 23ai in 2024, and the AI Database 26ai in 2025. The naming has followed whatever the company was promoting at the time, which is a reasonable summary of its marketing.

Technically significant features include Real Application Clusters, which allows several servers to operate against one database; partitioning; Data Guard for replication; Exadata, a purpose-built hardware and software combination that offloads parts of query processing into the storage layer and which has been among Oracle's most successful products; the in-memory column store; multitenant architecture, which allows many databases to share one instance; and the Autonomous Database, announced in 2018, which automates patching, tuning, backup and scaling and which Ellison promoted on the argument that most security failures follow from patches not being applied.

Version 23ai and the AI Database 26ai integrate vector search and retrieval directly into the database, on the argument that organisations would rather query their existing data with new methods than copy it into a separate vector store.

The database's commercial character is as important as its technical one. Oracle licences are expensive, its licensing terms are complicated, particularly in virtualised and cloud environments, and its licence audits are a well-known feature of the enterprise software landscape and a persistent source of customer resentment. This has driven a long migration toward PostgreSQL, MySQL and cloud-native alternatives, which Oracle has slowed by improving the product and by pricing concessions rather than by simplifying the terms.

Applications

Oracle Fusion Cloud Applications covers enterprise resource planning, human capital management, supply chain and customer experience, assembled from Oracle's own development and from PeopleSoft, Siebel, JD Edwards and others. NetSuite serves mid-sized organisations and was the first substantial cloud enterprise resource planning system, founded in 1998 with Ellison's backing.

Industry-specific applications have grown in importance. Oracle Health, built on Cerner, supplies electronic health records to a large share of American hospitals and to the United States Department of Veterans Affairs and Department of Defense, the latter contracts having been troubled and subject to repeated pauses. OPERA Cloud, under Oracle Hospitality, is a property management system available in 236 countries and territories, adopted by Accor in 2025 and by IHG Hotels and Resorts in 2026. MICROS supplies point-of-sale systems to restaurants and retail, and Oracle also serves financial services, utilities, construction and telecommunications with specialised products.

Cloud infrastructure

Oracle Cloud Infrastructure provides compute, storage, networking, database and artificial intelligence services. Its distinguishing features are bare metal instances, a non-oversubscribed network, consistent pricing across regions and substantially lower egress charges than its principal competitors.

Two strategies distinguish it from the larger providers. Dedicated Region places a complete Oracle cloud region inside a customer's own data centre, which suits governments and regulated industries that cannot use shared infrastructure. Multicloud places Oracle database services physically inside Microsoft Azure, Google Cloud and Amazon Web Services data centres, under arrangements announced from 2023, which accepts that customers will use several providers and positions Oracle's database rather than its infrastructure as the product. The approach would have been unthinkable under the company's earlier posture and has grown very rapidly.

The generative artificial intelligence service runs models from Cohere, from Meta's Llama family and, from 2025, xAI's Grok, and Oracle has integrated generative features across its applications, including the Clinical Digital Assistant announced in September 2024, which produces consultation documentation and proposes orders automatically.

Java and open source

Oracle acquired Java with Sun Microsystems and has stewarded it since, moving to a six-month release cadence in 2017 and changing the commercial licensing of its own distribution several times in ways that customers found difficult to track, including a 2023 move to per-employee pricing that prompted substantial migration to alternative builds such as those from Eclipse Adoptium, Amazon and Microsoft.

MySQL, also from Sun, remains widely used and is developed by Oracle alongside its commercial database, an arrangement the open-source community regarded with suspicion at the outset and which has been maintained without the neglect that was predicted, though MariaDB's fork has taken a substantial share.

Oracle also maintains VirtualBox, GraalVM and, until it was transferred to community stewardship, OpenOffice.

The relational database and why it mattered

Oracle's commercial position rests on a piece of computer science whose significance is easy to underrate because it has been universal for so long.

Before the relational model, data was stored in hierarchical or network databases in which the physical arrangement of records and the pointers between them were part of the application's logic. A programme that read customer records knew where they were and how they linked to orders. Changing the structure meant rewriting every programme that touched it, and asking a question the designers had not anticipated could require new code and a new physical layout.

Codd's proposal separated the two. Data would be stored as relations, that is as tables of rows and columns, with relationships expressed by matching values rather than by pointers. Queries would state what was wanted, and a component called the optimiser would decide how to obtain it. The consequence is that the physical arrangement can change without the queries changing, that new questions require no new code, and that the correctness of a query can be reasoned about independently of the machine.

The objection at the time was performance, and it was serious: a system that decides how to execute a query at run time will generally be slower than one whose access path was chosen by a programmer. The counter-argument, which prevailed over about fifteen years, was that optimisers improve faster than programmers and that hardware improves faster than either.

Oracle's specific commercial insight was not the model, which was Codd's, nor SQL, which was IBM's, but the recognition that portability mattered more than integration. By writing a database that ran on many manufacturers' machines, Oracle made itself the beneficiary of every hardware transition rather than a casualty of it. Every time customers moved from mainframes to minicomputers, from minicomputers to Unix workstations, from Unix to Windows and Linux, and from on-premises servers to cloud, the database went with them and the hardware supplier changed. IBM, which had the model first, was constrained by exactly the integration that made it money.

The database is also the reason the applications strategy worked. Enterprise resource planning, customer relationship management and human resources systems all sit on a database, and a supplier that owns the layer beneath the application has a permanent commercial advantage over one that does not.

How enterprise software is sold

Oracle is among the clearest illustrations of a business model that is poorly understood outside the industry and that explains much of the company's reputation.

Enterprise software is not bought the way consumer software is. A decision to install a database or an enterprise resource planning system commits an organisation for ten to twenty years, because the system becomes the record of the business, because the surrounding processes are built around it, and because the cost of migration includes not only the software but the retraining, the data conversion, the integration with dozens of other systems and the risk of interrupting operations. The switching cost is therefore very high and is borne by the customer rather than the supplier.

This has three consequences that are visible throughout Oracle's history.

First, the initial sale is worth far more than its price, which is why enterprise software is sold at very large discounts to win the position and why maintenance, typically 22 per cent of list licence value annually, is where the profit accumulates. Maintenance revenue is close to pure margin, is very rarely cancelled, and rises with the list price rather than with the discounted price actually paid.

Second, the supplier's incentives after the sale are weaker than in most markets. A customer who cannot leave need not be delighted, which is why enterprise software is consistently rated poorly by its users and why the industry's customer satisfaction scores bear no relationship to its market shares.

Third, acquisition is a rational growth strategy. Buying a competitor delivers its installed base and its maintenance stream immediately, and the acquirer can reduce the target's costs substantially because the customers cannot leave in response. Oracle's acquisition programme after 2004 was an explicit application of this reasoning, and Safra Catz's financial discipline in executing it is why it produced returns where similar programmes elsewhere destroyed value.

The check on the model is the arrival of a genuinely cheaper alternative at a moment when a customer is replacing the system anyway, which is what cloud and open-source databases have provided and what has made the past decade harder for Oracle than the preceding two.

TikTok and Project Texas

Oracle's involvement with TikTok began in September 2020, when the Trump administration sought to force ByteDance to divest the application's American operations on national security grounds.

Oracle, partnered with Walmart, was selected over Microsoft in an arrangement that was never completed in its original form after the executive orders were challenged and then withdrawn. What survived was Project Texas, under which TikTok's American user data was migrated to Oracle cloud infrastructure, Oracle reviewed the application's source code and its recommendation algorithm, and a separate entity, TikTok US Data Security, was established with oversight arrangements.

The arrangement was criticised from several directions: by security specialists who argued that data residency does not address the question of who can access data or influence the recommendation system; by ByteDance's opponents who regarded it as insufficient; and by observers who noted that Oracle had become both the auditor and the commercial beneficiary.

Legislation requiring divestiture passed in April 2024 and was upheld by the Supreme Court in January 2025. After repeated extensions, the transaction completed in January 2026, with Oracle, MGX and Silver Lake each taking 15 per cent of the American venture and ByteDance retaining 19.9 per cent, below the threshold the legislation set. Oracle's stake was estimated at approximately US$2 billion in March 2026, and it continues to provide the infrastructure and to hold responsibility for security review.

Oracle Health

The acquisition of Cerner in June 2022 for approximately US$28.3 billion made Oracle one of the two dominant suppliers of electronic health records in the United States, alongside Epic Systems.

The strategic argument was that healthcare is among the largest industries with the least modernised information systems, that Cerner's installed base in hospitals was substantial, and that Oracle's cloud and database technology could modernise a product built on older architecture. Ellison described an ambition to build a national health records database, a statement that drew immediate privacy objections.

Execution has been difficult. The United States Department of Veterans Affairs implementation, begun under Cerner in 2018 and one of the largest health information technology projects ever attempted, was paused repeatedly after patient safety incidents, congressional criticism and cost increases, and resumed on a revised schedule. The Department of Defense implementation proceeded more smoothly. Large numbers of Cerner staff were let go after the acquisition, and hospital customers reported degraded support during the transition.

Oracle has since rebuilt the product on its cloud infrastructure and added generative features, including the Clinical Digital Assistant announced in September 2024, which produces documentation from a recorded consultation and proposes orders, a category of product that has been among the more genuinely useful early applications of language models because clinical documentation consumes a very large share of physicians' time.

The March 2025 claim of a breach of six million Oracle Health records, and the company's handling of its disclosure, damaged confidence among hospital customers.

Energy and sustainability

The expansion of data centre capacity has made Oracle, like its competitors, a very large consumer of electricity and a participant in the political arguments that follow.

The company has committed to powering its operations entirely with renewable electricity and reports having reached that for a large majority of its cloud regions, and has set a 2050 net zero target with interim milestones. Its cloud facilities have been designed for relatively high power usage effectiveness by industry standards.

The ratepayer protection pledge signed in March 2026 is the more consequential commitment. Under it, data centre operators undertake to build or procure the generation their facilities require on terms separate from those applying to residential and small business customers, so that the cost of new capacity is not socialised through general tariffs. The pledge followed political pressure in several American states where household bills rose sharply in areas of heavy data centre construction, and it is a voluntary anticipation of regulation rather than a response to it.

The larger question, which applies to the whole sector, is whether generation can be added at the rate at which computing demand is growing, and Oracle's construction programme for Stargate and its other artificial intelligence capacity is constrained more by interconnection and turbine availability than by capital.

Corporate affairs

Leadership

Larry Ellison has been chairman and chief technology officer since 2014, having been chief executive from 1977 to 2014 and chairman from 1990 to 2004. He owns a very large proportion of the company, reported at around 42 per cent in 2021, and is among the wealthiest people in the world, with holdings including 98 per cent of the Hawaiian island of Lanai, the Indian Wells tennis tournament, and substantial investments in Tesla. His public persona, built on yacht racing, jet aircraft, and a readiness to attack competitors by name, is unusual for the leader of an enterprise software company and has been a deliberate marketing instrument.

Safra Catz was chief executive from 2014, initially alongside Mark Hurd, and moved to executive vice chair in September 2025. She had joined from Donaldson, Lufkin and Jenrette in 1999, became chief financial officer and co-president, and was the architect of the acquisition strategy and of the financial discipline that made it work. She was ranked tenth on Fortunes Most Powerful Women list in 2016.

Clay Magouyrk and Mike Sicilia became co-chief executives in September 2025. Magouyrk had come from Amazon Web Services and built Oracle Cloud Infrastructure; Sicilia ran the industry applications businesses.

Headquarters and workforce

Oracle moved its headquarters from Redwood Shores, California, to Austin, Texas, in 2020, and announced in April 2024 a further move to Nashville, Tennessee. The Redwood Shores campus, with its cylindrical towers beside an artificial lake, remains a large site and is among the most recognisable corporate buildings in Silicon Valley.

The workforce stood at approximately 141,000 in May 2026, down from 162,000 a year earlier following large reductions, including the roughly 12,000 positions reported cut in India in April 2026. Oracle has historically expanded through acquisition and reduced through subsequent consolidation, and the pattern following PeopleSoft, Sun and Cerner was consistent.

Finances

Oracle's financial year ends on 31 May. Revenue has grown from about US$10 billion in 2004 to approximately US$50 billion in 2023 and higher since, with the composition shifting sharply from perpetual software licences to cloud subscriptions and infrastructure.

The company's operating margin has historically been among the highest in enterprise software, in the mid-thirties as a percentage of revenue on a reported basis and considerably higher on the measures the company prefers, sustained by maintenance revenue on installed software that costs very little to provide.

The artificial intelligence expansion has changed the financial character of the business substantially and not entirely favourably. Cloud infrastructure requires very large capital expenditure, which Oracle has funded partly with debt, and the margins on renting computing capacity are far below those on licensing software. The remaining performance obligation of approximately US$130 billion reported for the second quarter of financial 2026 is very large, and analysts have observed that its value depends on customers, principally OpenAI, being able to pay, and that Oracle is therefore exposed to the financing of the artificial intelligence sector in a way that a software company would not normally be.

Oracle pays a dividend and has historically conducted very large share repurchases, which have been reduced as capital expenditure has risen.

Hardware and engineered systems

Oracle became a hardware company by accident when it bought Sun Microsystems in January 2010, and the outcome was mixed.

Sun had been among the defining companies of the workstation and server era, with the SPARC processor architecture, the Solaris operating system, the NFS file system, Java, and the ZFS file system. It had been badly damaged by the collapse of the dot-com boom, whose infrastructure it had largely supplied, and by the rise of commodity Intel servers running Linux. IBM had negotiated to buy it and withdrawn, and Oracle bid approximately US$7.4 billion.

Ellison's stated intention was to build integrated systems in which hardware and software were designed together, on the analogy of Apple. The strategy produced Exadata, which combines database servers, storage servers with database-aware software, and InfiniBand networking, and which offloads parts of query execution into the storage layer so that only relevant data crosses the network. Exadata has been genuinely successful, is the platform on which a large proportion of Oracle's largest customers run, and now underpins the company's cloud database services.

Other engineered systems followed, including Exalogic for applications, Exalytics for analytics, the SPARC SuperCluster and the Oracle Database Appliance.

The general-purpose hardware business did not survive. SPARC development was wound down after 2017 with large reductions in the Solaris and SPARC engineering groups, and the market moved entirely to x86 and subsequently to Arm. Oracle now designs its cloud hardware around commodity processors and NVIDIA accelerators and sells engineered systems principally where the database is the workload.

The European Commission's review of the Sun acquisition, which delayed it by several months over concerns about MySQL, produced commitments from Oracle about the database's continued development that it has broadly observed.

Litigation

Oracle has been an unusually litigious company, both as plaintiff and as defendant, and several of its cases have shaped software law.

Oracle v. Google, described above, ran for eleven years over the reimplementation of Java application programming interfaces in Android and concluded in the Supreme Court's April 2021 holding that Google's use was fair. The decision is now the principal American authority on the copyright status of software interfaces and is generally regarded as having preserved the industry's practice of reimplementing interfaces for compatibility.

Oracle v. SAP concerned TomorrowNow, an SAP subsidiary that provided third-party support for Oracle applications and that was found to have downloaded very large quantities of Oracle software and documentation without authorisation. A jury awarded US$1.3 billion in 2010, which the judge set aside as excessive; the matter was finally resolved in 2014 for US$356.7 million. SAP closed TomorrowNow.

Oracle v. Hewlett-Packard arose when Oracle announced in 2011 that it would cease developing software for Intel's Itanium processor, on which HP's high-end servers depended. HP argued that a settlement agreement reached when Mark Hurd left HP for Oracle obliged continued support. A California court agreed, ordered Oracle to continue, and in 2016 a jury awarded HP approximately US$3 billion, later reduced.

Rimini Street, a third-party support provider for Oracle software, has been in litigation with Oracle since 2010 across several actions concerning the copying of software for support purposes, with findings substantially in Oracle's favour and with injunctions restricting Rimini's practices. The cases matter because third-party support is the principal competitive check on maintenance pricing in enterprise software.

The company has also defended shareholder litigation over the NetSuite acquisition, given Ellison's ownership stake in the target, which was settled in 2022 for US$115 million paid by insurers, and a long-running derivative action over the Cerner purchase.

Competition

SAP is Oracle's oldest and closest competitor in enterprise applications, and the rivalry has been conducted with public hostility for thirty years. SAP is stronger in manufacturing and in Europe, Oracle in financial services and in the United States, and both are engaged in migrating very large installed bases from on-premises software to cloud versions, a transition each has found slower than planned.

Salesforce was founded in 1999 by Marc Benioff, a former Oracle executive whom Ellison had mentored and personally funded, and the relationship between the two men has alternated between friendship and public antagonism. Salesforce established software as a service as a category and took the customer relationship management market Oracle had bought Siebel to win.

Amazon Web Services is both Oracle's principal competitor in infrastructure and, historically, one of its largest database customers, and Amazon's public removal of Oracle databases from its own operations, completed in 2019, was a marketing exercise as much as a technical one. AWS's Aurora and Redshift are designed explicitly to displace Oracle.

Microsoft competes across applications, database with SQL Server, and infrastructure with Azure, while simultaneously partnering with Oracle on the multicloud arrangements that place Oracle databases inside Azure data centres. The combination of competition and cooperation is characteristic of the present market.

Google competes in infrastructure and remains Oracle's most prominent litigation opponent, while also hosting Oracle database services under the multicloud programme.

In databases the more significant long-term competition is from PostgreSQL, an open-source system whose capability now approaches Oracle's for a large majority of workloads at no licence cost, and from cloud-native systems including Snowflake, Databricks and the hyperscalers' own offerings.

Oracle CloudWorld

Oracle's annual customer conference, held as Oracle OpenWorld from 1996 and renamed CloudWorld in 2022, has been among the largest technology events in the world, at its peak drawing more than 60,000 people to San Francisco and closing several city blocks.

Ellison's keynote is the event's defining feature and has followed a consistent form for three decades: a technical claim, a demonstration, a direct attack on a named competitor, and a benchmark comparison whose methodology the competitor then disputes. The tradition has produced a considerable number of quotable remarks, including his 2008 characterisation of cloud computing as fashion, and has been an effective instrument for a company whose products are otherwise difficult to make interesting.

The event moved to Las Vegas in 2022 and has been held in several cities since.

Philanthropy

Larry Ellison signed the Giving Pledge in 2010. His giving has concentrated on medical research, through the Ellison Medical Foundation, which funded work on ageing and infectious disease from 1997 until it wound down its grant programme in 2013, and through the Ellison Institute for Transformative Medicine, established at the University of Southern California in 2016 with a US$200 million gift and subsequently expanded, including the Ellison Institute of Technology at Oxford announced in 2023 with a commitment reported at more than £1 billion.

He gave US$100 million to the University of California, Los Angeles for cancer research in 2007 as part of a settlement of a shareholder suit over his stock trading, an origin the university did not emphasise.

Oracle operates education programmes including Oracle Academy, which provides computing curriculum and software to schools and universities worldwide, and the Oracle Education Foundation.

Timeline

1970 Edgar Codd publishes the relational model at IBM.
1977 Larry Ellison, Bob Miner and Ed Oates found Software Development Laboratories on 16 June.
1979 Oracle Version 2 ships, the first commercial SQL relational database.
1983 Version 3, rewritten in C, makes the product portable across machines.
1986 Lists on Nasdaq on 12 March.
1990 Restates revenue, reports a loss and nearly fails.
1992 Oracle 7 establishes technical dominance; Raymond Lane joins as president.
2004 Wins the antitrust trial over PeopleSoft; releases 10g.
2005 Completes the PeopleSoft acquisition for about US$10.3 billion.
2006 Acquires Siebel Systems.
2008 Acquires BEA Systems; Ellison mocks cloud computing.
2010 Acquires Sun Microsystems, gaining Java, MySQL and Solaris; sues Google over Java.
2014 Ellison becomes chairman and chief technology officer; Safra Catz and Mark Hurd become co-chief executives.
2016 Acquires NetSuite; begins building the second-generation cloud infrastructure.
2018 Launches the Autonomous Database.
2020 Moves its headquarters to Austin; joins the TikTok arrangement as cloud provider.
2021 The Supreme Court decides Oracle v. Google in Google's favour.
2022 Acquires Cerner for approximately US$28.3 billion, its largest purchase.
2024 Announces a move to Nashville; settles the privacy class action for US$115 million; releases Database 23ai.
2025 Stargate is announced in January; the US$300 billion OpenAI computing contract is signed in September; Clay Magouyrk and Mike Sicilia become co-chief executives.
2026 Takes a 15 per cent stake in TikTok's American operations; announces 72 embedded multicloud data centres; the workforce falls to 141,000.

Controversies

Foreign Corrupt Practices Act

Between 2011 and 2024 Oracle was investigated for bribery across subsidiaries in India, Turkey, the United Arab Emirates and South Africa. In each instance it settled with the Securities and Exchange Commission. The 2022 settlement concerned schemes in which excessive discounts were approved and the difference used to fund slush accounts that paid foreign officials, and followed an earlier 2012 settlement over similar conduct in India, which the Commission noted in observing that the company had not implemented adequate controls after the first case.

Privacy

A class action filed in 2022 alleged that Oracle's advertising technology business had assembled detailed profiles on approximately five billion people without consent, drawing on data from a very large number of sources, and characterised the company as operating a surveillance apparatus. Oracle settled in July 2024 for US$115 million and subsequently wound down much of the advertising business, which had been assembled from acquisitions including BlueKai, Datalogix and AddThis and which had never been a large contributor to revenue.

Security

In March 2025 a threat actor claimed to have taken six million records from Oracle Health, the former Cerner. Oracle said that Oracle Cloud Infrastructure had not been breached and that no customer data from it was exposed, while acknowledging unauthorised access to some legacy Cerner servers. The company did not confirm the scale, and its communications during the episode were criticised by security researchers as evasive.

Oracle's patching cadence, quarterly rather than continuous, and the very large number of vulnerabilities addressed in each critical patch update, have been a long-standing criticism, and the security researcher community's relationship with the company has been poor since a 2015 blog post by the then chief security officer instructing customers not to reverse-engineer Oracle products in search of vulnerabilities, which the company withdrew.

Licensing and audits

Oracle's licence terms, particularly for the database in virtualised environments, have been the subject of long disputes with customers who found that running Oracle software on a virtualised cluster could be interpreted as requiring licences for every processor in the cluster. The company's audit practice, under which licence compliance reviews are initiated and settled commercially, is widely regarded as a sales channel, and has driven both the growth of a specialist advisory industry and a steady migration away from the products.

Political and geopolitical positions

Following the beginning of the Gaza war in 2023, Oracle's senior executives including Safra Catz and Larry Ellison publicly aligned the company with Israel's military operations, issuing statements of solidarity, paying double salaries to Israeli employees and donating to organisations connected to the war effort. The positions drew criticism from employees and from external campaigners and were defended by the executives as matters of personal and corporate conviction.

Ellison has been a substantial donor to Republican candidates and causes, and the company's prominence at the White House announcement of Stargate in January 2025, and the role of Ellison's son David, whose Skydance acquired Paramount and later pursued Warner Bros. Discovery, have made the family unusually visible in American political and media affairs.

Labour

The large reductions of 2025 and 2026, including approximately 12,000 positions in India, drew criticism over their conduct, with affected employees reporting minimal notice. Oracle has also faced litigation over pay discrimination, including a Department of Labor action concerning pay disparities by race and gender at its Redwood Shores headquarters, which was dismissed in 2022 after the department withdrew.

Oracle in popular and business culture

Oracle occupies an unusual place in the culture of the technology industry: it is enormously successful, almost entirely invisible to consumers, and among the least liked companies in its sector by the people who work with its products.

Larry Ellison has cultivated this. He has raced in the America's Cup, winning it in 2010 and 2013 with Oracle Team USA, the 2013 victory from an eight to one deficit being among the most remarkable comebacks in the sport's history. He has flown fighter aircraft, bought 98 per cent of the Hawaiian island of Lanai in 2012, and conducted public feuds with the leaders of Microsoft, SAP, Hewlett-Packard, Salesforce and Amazon. A 1997 biography by Mike Wilson was titled The Difference Between God and Larry Ellison, the joke being completed by its subtitle, God Doesn't Think He's Larry Ellison.

The company's culture has historically been described as sales-led, competitive to the point of internal antagonism, and driven by quarterly targets. Its compensation structures for salespeople have periodically been the subject of litigation and of internal complaint, most prominently over commission caps applied retroactively to unusually large deals.

Among software developers Oracle's reputation is poor and specific: the licensing terms, the audits, the quarterly patch cycle, the handling of the open-source projects acquired with Sun Microsystems, and the Java litigation are each cited regularly. The company has made little attempt to change this, on what appears to be the correct commercial judgement that developer sentiment does not determine enterprise purchasing.

The contrast with the products themselves is worth noting. The Oracle Database is, by the general assessment of people who administer databases for a living, among the most capable pieces of software ever written, with a query optimiser, recovery mechanism and concurrency model that competitors have taken decades to approach. The resentment is directed at the commercial arrangements around it rather than at the engineering.

Products in brief

Database and data: Oracle Database and the AI Database 26ai, Exadata, Autonomous Database, MySQL and HeatWave, NoSQL Database, TimesTen, Berkeley DB, Essbase, Oracle Analytics.

Cloud infrastructure: compute including bare metal and NVIDIA accelerator instances, object and block storage, networking, Dedicated Region, Compute Cloud@Customer, Roving Edge, the generative artificial intelligence service, and the multicloud arrangements with Microsoft Azure, Google Cloud and Amazon Web Services.

Applications: Fusion Cloud ERP, EPM, SCM, HCM and CX; NetSuite; PeopleSoft, JD Edwards EnterpriseOne and Siebel, all still supported for existing customers; Taleo; Responsys and Eloqua.

Industry applications: Oracle Health, formerly Cerner; OPERA Cloud and Simphony under Oracle Hospitality; MICROS point of sale; Oracle Financial Services Analytical Applications and Flexcube for banking; Oracle Utilities; Primavera for construction and project management; Oracle Retail; and Oracle Communications.

Development and middleware: Java SE and EE, GraalVM, WebLogic Server, Coherence, APEX, SQL Developer, VirtualBox, Oracle Linux and Solaris.

Regulation and public policy

Oracle's regulatory exposure differs from that of the consumer technology companies. It is scrutinised less for competition and privacy and more for government contracting, export control and the specific conditions attached to its acquisitions.

Its government business is substantial, covering federal agencies, the Department of Defense, the Department of Veterans Affairs through Oracle Health, and a large number of state and local bodies, and the company has been a persistent litigant over procurement decisions, most visibly in its unsuccessful challenge to the award of the Pentagon's JEDI cloud contract to Microsoft in 2019, which Oracle pursued on the ground that the single-award structure was improper and that conflicts of interest had tainted the process. JEDI was cancelled in 2021 and replaced with a multiple-award arrangement in which Oracle participated.

The European Commission's conditions on the Sun acquisition, concerning MySQL's continued development, were the most significant competition remedy Oracle has accepted. Chinese approval of the Cerner acquisition and of other transactions has been granted without comparable conditions.

The company has lobbied consistently on intellectual property, arguing for strong software copyright during the Google litigation, on immigration and on tax, and it was among the companies that benefited substantially from the repatriation provisions of the 2017 American tax legislation.

Data localisation requirements in Europe, India, Saudi Arabia and elsewhere have been commercially favourable to Oracle, since Dedicated Region and Cloud@Customer allow a complete cloud to be placed inside a jurisdiction or inside a customer's own building in a way that the larger providers' architectures do not easily permit. The company has used this to win sovereign cloud business, including with governments that are uneasy about American hyperscale providers while remaining, itself, an American company subject to the same laws.

Assessment

Oracle has now had three distinct commercial lives, which is unusual for a company of its age.

The first was the database, where it took an idea IBM had published and declined to commercialise, built it portably, sold it aggressively, and held a dominant position for twenty-five years. The second was applications, assembled almost entirely by acquisition on the correct observation that enterprise software customers are captive and that buying the supplier is cheaper than winning the customer. The third is infrastructure, entered late, initially badly, and then successfully by building something technically different from its competitors and by being willing to sign contracts they would not.

The company's character has been consistent across all three. It competes on price and on lock-in rather than on the affection of its customers; it litigates; it acquires; and it is led, still, by a founder who has been in charge for forty-nine years and who owns a very large part of it. The resulting culture produces both the results and the resentment, and the two are not separable.

The present position carries a specific risk that is worth stating plainly. Oracle's valuation and its recent growth rest heavily on contracts to supply computing to a small number of artificial intelligence companies, principally OpenAI, financed in substantial part by debt raised against those contracts. If the demand persists, Oracle will have converted a mature software business into an infrastructure business at exactly the right moment. If it does not, the company will hold very large depreciating assets and the obligations that financed them. Ellison has made bets of this character before, on the network computer and on hardware, and has been wrong at least twice; he has also been right about the direction of the industry, if not its timing, more often than his critics allow.