/
.webp)
San Jose, California. 1973.
Inside IBM's Santa Teresa research lab, a mild-mannered British mathematician named Edgar Frank Codd is watching his life's work get slowly, quietly smothered.
Codd had published what would become one of the most important papers in the history of computing. "A Relational Model of Data for Large Shared Data Banks," 1970, ACM Communications. In eleven dense pages, Codd described a fundamentally new way to store and retrieve information — not in the rigid, hand-coded hierarchical trees that dominated computing at the time, but in flexible, mathematical tables that could be queried by anyone, in any order, without knowing how the data was physically stored.
It was, in retrospect, the blueprint for the entire modern technology industry.
IBM told him to slow down.
Not because the idea was wrong. It was clearly right. IBM's own internal researchers knew it was right. The problem was that IBM already had a product called IMS — a hierarchical database system that thousands of IBM customers were running and paying for. Commercializing Codd's relational model meant those customers might switch. It meant cannibalizing a revenue stream. It meant internal political battles with product teams who had built careers on IMS.
So IBM did what large, successful companies often do when confronted with a disruptive idea generated by their own people: they funded a research project to study it, kept it inside the lab, and quietly ensured it went nowhere fast.
IBM called the project System R. It began in 1973. It would not produce a commercial product until 1981.
Two years after a dropout from Chicago had already shipped one.
Born Without a Name
Lawrence Joseph Ellison was born on August 17, 1944, in New York City, to a nineteen-year-old single mother named Florence Spellman. He was nine months old when he caught pneumonia. His mother, unable to care for him alone, sent him to Chicago to be raised by her aunt and uncle, Lillian and Louis Ellison, who adopted the baby boy and gave him their name.
He did not learn he was adopted until he was twelve.
He did not meet his birth mother until he was forty-eight.
His adoptive father, Louis, was a self-made man — a Ukranian Jewish immigrant who had come to America with nothing and built himself into the solid middle class of Chicago's South Shore neighborhood. Louis Ellison believed in security, in practicality, in not dreaming too large. He found his adopted son baffling. Young Larry was bright — obvious to everyone who met him — but unfocused. Rebellious. Unwilling to work within structures he hadn't chosen.
"My father always told me I would never amount to anything," Ellison later recalled. He said it without self-pity. He said it like a man who had kept score.
The Two Dropouts
After high school, Ellison enrolled at the University of Illinois, Champaign. He was named science student of the year. Then, during his second year, his adoptive mother Lillian died. The woman who had been warm and loving and protective — the counterweight to Louis's cold ambition — was gone. Ellison dropped out.
The following fall, he enrolled at the University of Chicago. He lasted one semester. He had discovered computers. He was restless. He dropped out again.
He packed his things and drove to California.
California, Job to Job
For the better part of a decade, Ellison drifted through the technology industry on the strength of raw talent and personal magnetism, acquiring skills the way autodidacts do: by doing, by watching, by refusing to be intimidated by things he didn't yet know. He worked at Wells Fargo. He worked at Amdahl Corporation, the mainframe maker. He moved from company to company, learning programming on the job, never staying still long enough to be categorized.
In 1973, he landed at Ampex Corporation, a Silicon Valley electronics firm. There, he met two men who would change his life: Ed Oates, a programmer, and a gentle, brilliant senior developer named Bob Miner who would become Ellison's creative opposite — the man who would actually build the thing Ellison sold.
At Ampex, Ellison read Edgar Codd's paper.
Later, he would describe the experience as a revelation. Not because it was difficult to understand — it wasn't. It was because it was so obviously correct, and so obviously unbuilt. The concept was sitting there, published in a major academic journal, publicly available to anyone who cared to read it. IBM had invented it and wasn't building it. The commercial opportunity was lying in the open like money on the sidewalk.
Ellison reached down and picked it up.
The CIA Contract That Named a Company
In 1977, Ellison, Miner, and Oates left their jobs and founded a company they called Software Development Laboratories, incorporated in the state of California with $2,000 in capital — $1,200 of it Ellison's.
Their first contract was from the United States government. Specifically, from the Central Intelligence Agency.
The CIA needed a database system. They had a project for it. They gave that project a codename, the way intelligence agencies name everything: they called it Oracle. The system was meant to answer any question, to draw intelligence from disparate data streams and surface it to analysts — the modern classical oracle, a machine that could see everything and tell you what you needed to know.
Ellison and Miner built it. The CIA paid them. And when the founders needed a name for their flagship product — the relational database they were building for commercial sale — they asked the CIA for permission to use the codename.
The CIA agreed.
The company would eventually name itself after that first contract. Oracle. A company built, at its origin, to answer questions for the intelligence community, that would go on to store the questions and answers of the entire enterprise world.
The Version 2 Gambit
By 1978, Oracle Version 1 existed — barely. It was an internal prototype, written in assembly language, running on a PDP-11 minicomputer with 128 kilobytes of memory. It was functional enough to prove the concept. It was not ready to sell.
So they built Version 2.
When Relational Software Inc. — the new name for the company, adopted in 1979 — shipped Oracle to its first commercial customers, the product was labeled Oracle Version 2. There had been no Version 1 in any customer's hands. Version 1 had never been released. It was a prototype that lived and died inside their own offices.
The decision to skip Version 1 was pure psychology. Enterprise software customers in 1979 were cautious men in cautious organizations. They did not buy Version 1 of anything. Version 1 implied an experiment. Version 1 implied bugs. Version 1 implied that you were a test subject.
Version 2 implied maturity. Version 2 implied that someone else had already bled on your behalf.
They shipped Oracle Version 2 in 1979. IBM shipped its first commercial relational database, SQL/DS, in 1981. Oracle had a two-year head start, gained not by luck but by reading a paper, believing it, and moving while the company that wrote it was still in committee.
Selling Faster Than the Product Existed
Larry Ellison was not the engineer. That was Bob Miner — quiet, methodical, technically deep, the kind of programmer who could hold a whole architecture in his head and build it cleanly. Ellison was the one who went out and sold it.
He was extraordinary at it. Not because he was dishonest — he was, by most accounts, entirely willing to shade the truth about what was built versus what was coming — but because he had a gift for making people believe in the future he was describing. He recruited aggressively. He pushed his sales team hard. He put targets on whiteboards and held people to them with an intensity that Silicon Valley had not quite seen before.
Oracle sales culture became infamous. The company hired young, hungry, competitive reps and ran them through a training program so thorough and so demanding that Oracle alumni would eventually spread across the enterprise technology industry and reshape its entire sales profession. Salesforce was founded by a former Oracle executive. Marc Benioff was Oracle's youngest-ever VP. The Oracle sales machine was, in a sense, the business school that taught Silicon Valley how to sell to the enterprise.
The cost of that culture was real. By the late 1980s, Oracle reps were selling software that didn't fully exist, booking revenue on contracts that hadn't been delivered, and the company's 1990 earnings restatement nearly destroyed it. Oracle had to lay off hundreds of people. Its stock collapsed. Ellison survived — barely — and rebuilt on more honest foundations. But the near-death experience left a mark on how the industry understood the difference between aggressive and reckless.
Oracle Version 2 was not just first. It was technically superior to what IBM eventually shipped, because Oracle had built for SQL compatibility from scratch — deliberately modeling on IBM's own System R research, anticipating IBM's eventual direction, ensuring that when enterprises did migrate, Oracle was the natural landing point.
The strategy worked. Through the 1980s, Oracle became the de facto standard for enterprise relational databases. IBM finally shipped DB2 in 1983, four years after Oracle had already been in the market. By then, Oracle had deep relationships with Fortune 500 companies, government agencies, and the growing minicomputer market. The window that IBM had left open by protecting IMS had been permanently occupied.
By 1990, Oracle was a billion-dollar company. By the mid-1990s, it was the second-largest software company in the world, behind only Microsoft.
Edgar Codd watched all of this happen. He received the Turing Award — computing's Nobel Prize — in 1981, recognition from the academic community that his paper had mattered. But as Oracle and other vendors commercialized the relational model, Codd grew increasingly frustrated. The implementations deviated from his theoretical ideal. SQL, the query language that had become the industry standard, violated his relational model in dozens of small ways that he considered unacceptable.
In 1985, he published his "12 Rules" — a litmus test for what could legitimately call itself a relational database. Then in 1988, he published "Fatal Flaws in SQL," a direct critique of the language that had made Oracle rich.
No major database, including Oracle, fully satisfied Codd's 12 Rules. The father of the relational database watched the commercial world take his invention, sell it to everyone on earth, and get almost everything slightly wrong.
He died in 2003. Oracle's revenues that year were $9.5 billion.
The Predator Decade
By the early 2000s, Ellison had decided that organic growth was too slow for his ambitions. The enterprise software market was fragmenting — dozens of specialized vendors had built valuable customer bases in HR, CRM, ERP, supply chain. Ellison wanted all of it.
He started with PeopleSoft.
The story of Oracle's acquisition of PeopleSoft is one of the most theatrical episodes in enterprise technology history. On June 6, 2003, Oracle launched an unsolicited hostile bid for PeopleSoft — a company in the middle of its own acquisition of J.D. Edwards, a company with a fiercely loyal customer base, and a CEO named Craig Conway who had once been an Oracle executive himself and despised Ellison with a personal intensity that made the whole affair electric.
Conway called Oracle's bid "atrociously bad behavior from a company with a history of atrociously bad behavior." He rejected it four times. He likened it to being asked to sell your dog so someone could shoot it in the backyard. The U.S. Department of Justice filed suit to block the deal. The European Commission objected.
Then PeopleSoft's board fired Conway. They said he had misled analysts. And with their own CEO gone, the resistance collapsed.
Oracle closed the deal in January 2005 for $10.3 billion. It was 18 months of public combat that ended not with a negotiation but with an organizational decapitation.
The pattern repeated, methodically, for the next decade. Siebel Systems, the CRM giant, for $5.85 billion in 2005. BEA Systems, the Java middleware leader, for $8.5 billion in 2008. And then, in 2009, the acquisition that changed everything: Sun Microsystems for $7.4 billion.
The Sun Deal: Buying the Language
Larry Ellison said publicly that Oracle bought Sun for two things: Java and Solaris. That was true, but incomplete. By acquiring Sun, Oracle absorbed the company that had invented Java — the programming language running inside hundreds of millions of enterprise applications and virtually every Android device on the planet. Oracle also acquired MySQL, the world's most widely deployed open-source database, which now sat awkwardly inside a company whose flagship product competed with it.
The Sun acquisition made Oracle something it had never quite been before: a hardware company, a platform company, an operating system company, a programming language company, a database company, and an applications company simultaneously. Ellison described it as vertical integration. Critics described it as a diversification strategy executed through acquisition rather than invention.
Both were true.
Oracle vs. SAP: The Forty-Year War
The rivalry between Oracle and SAP is the longest-running cold war in enterprise technology. SAP, the German software giant founded in 1972, had built its empire on ERP — the integrated back-office systems that ran payroll, financials, manufacturing, and supply chain for the largest organizations on earth. Oracle had built its empire on databases and then, through acquisition, muscled into SAP's territory.
For four decades, they traded the same customers. SAP customers ran Oracle databases. Oracle customers ran SAP applications. The two companies were simultaneously competitors, partners, and mutually dependent in ways that neither ever fully acknowledged.
Oracle mounted a sustained campaign to displace SAP through acquisitions — PeopleSoft, JD Edwards, Siebel, and later NetSuite in 2016 for $9.3 billion — each designed to fill a gap in Oracle's application stack and press against SAP's installed base.
In 2024, for the first time in the history of enterprise software, Oracle surpassed SAP as the number-one ERP applications vendor. Oracle's ERP revenue reached $8.7 billion against SAP's $8.6 billion. A forty-year rivalry resolved by a margin that amounted, in a $1 trillion industry, to a rounding error.
Ellison was 79 years old. He had been building toward that number since he read a paper in a Silicon Valley office and decided the man who had written it deserved better than to be ignored.
1. The company was literally named after the CIA.
When most people hear that Oracle got its name from a CIA project, they imagine a loose inspiration — a poetic metaphor adopted in the spirit of intelligence gathering. The actual story is more literal. Ellison's team was directly hired by the CIA to build a database system the agency had already codenamed "Oracle." They built it. The CIA paid them. They liked the name so much that when they built their commercial product, they asked the CIA for permission to use it as their product name. The CIA said yes. Oracle the company exists because of a contract with the United States intelligence community. The name in the logo was not chosen in a conference room. It was inherited from a spy agency's filing system.
2. IBM invented the relational database and then chose not to build it.
This is not a story of a company failing to see an opportunity. It is a story of a company actively choosing not to take one. IBM's own researchers invented the relational model. IBM's own labs built System R. IBM's own management understood exactly what the technology could do. They did not commercialize it for nearly a decade because doing so would have threatened IMS, an existing product with existing revenue and existing customer relationships. The company that should have owned the relational database era sat on the invention while a dropout with $2,000 built a company worth hundreds of billions. It is the canonical case study in the danger of protecting what you have over building what's next.
3. There was never an Oracle Version 1.
Oracle's first commercial product was called Version 2. Version 1 existed only as an internal prototype that never shipped to a single customer. The founders made a deliberate decision to skip the number because enterprise customers in 1979 did not buy Version 1 software from companies nobody had heard of. Version 1 meant risk. Version 2 meant someone else had already taken the risk for you. It was a pure psychological play, and it worked. The entire Oracle empire was built on a product that announced itself as the second iteration of something that had never been the first.
4. Edgar Codd — the man whose paper made Oracle possible — died believing Oracle had betrayed his work.
Codd spent the last two decades of his life watching vendors commercialize his relational model in ways that violated its mathematical precision. He published a list of 12 rules defining what a true relational database must do. No major commercial database — including Oracle — has ever fully satisfied all 12. He published "Fatal Flaws in SQL" as a direct indictment of the language that had made Oracle a multi-billion-dollar company. He received the Turing Award. He received the recognition. And he watched the invention that bore his theoretical fingerprints get sold to the world in a form he considered incomplete. The man whose idea Oracle was built on spent his career arguing that Oracle had not actually built what he had imagined.
5. Larry Ellison never formally learned to code — he taught himself on the job at other people's companies.
The founder of the world's most powerful database company had no computer science degree. He dropped out of two universities before computer science was even a formal discipline in the way we know it today. He learned to program at Wells Fargo, at Amdahl, at Ampex — picking up skills at each stop, self-teaching in the way that the best technical autodidacts do, by building real things under real pressure. The man who read Codd's paper and decided to commercialize it did not have a credential that qualified him to understand it. He just read it, believed it, and found the person — Bob Miner — who could build it. The technical genius of Oracle's founding was not Ellison's. The audacity of it was entirely his.
Sources: Oracle Corporation founding records; Larry Ellison biographical accounts; Edgar F. Codd, "A Relational Model of Data for Large Shared Data Banks," ACM Communications, 1970; IBM System R project history; Oracle acquisition announcements 2003–2016; Apps Run The World ERP market share analysis, 2024.
