UNIVAC and the First Commercial Computers

By Victor Hoang · Updated Oct 23, 2026

The problem was counting. Governments and large firms had more records than clerks could process. The machines of the day read cards and printed tables.

The first commercial computers arrived within months of each other in 1951. The Ferranti Mark 1 reached Manchester University in February. UNIVAC I went to the US Census Bureau at the end of March. LEO I ran its first routine office job for J. Lyons & Co. in September. Which was "first" depends on whether you mean first delivered, first sold in America or first to do a business's daily work. Two engineers who left a university over a patent dispute built an industry, then watched a bigger company outsell them. For the wider field, see The History of Computers, the computer timeline and who invented the computer.

Some terms need a plain definition. A "commercial" computer is one built to be sold to customers, not built once for a lab or the military. A "stored-program" computer keeps its instructions in memory, which the stored-program idea made possible. "Magnetic tape" replaced the stack of punched cards as the way a machine held its data.

Timeline at a Glance

The Tabulator Years (1890-1945)

Governments buy machines to count people

The earliest documented case of selling calculating machinery to a government is the 1890 US census. IBM's history says the count used Herman Hollerith's punched cards and finished months ahead of schedule and under budget, and that punched cards later became a core product of what would become IBM. That is a precursor, not a computer: a tabulator added up holes in cards and could not run a program. The full story is in Hollerith, punched cards and the 1890 census, and the older tools are in the history of mechanical calculators.

The Census Bureau kept using this family of machines for decades. Its own history says it used updated versions of Hollerith's 1890 counting machine through the 1940 census. Then the war changed the tools. The Army's electronic ENIAC, told in the history of ENIAC, showed that vacuum tubes could calculate far faster than cards. The same Census page says that during the ENIAC project John Mauchly met with several Census Bureau officials to discuss peacetime uses.

This era set the buyer. Large statistical offices already paid for machines, so a faster one had a ready customer.

For teams today: the first market for a new technology is usually the buyer who already pays to solve the problem a slower way. Find that buyer before you design for a hypothetical one.

The Census Bargain (1946-1950)

Two engineers leave a university and run out of money

The turning point is the departure. The Computer History Museum says that in 1946 Mauchly and Eckert left academia and formed the Eckert-Mauchly Computer Corporation. Its calendar page adds that the founders left after a dispute with the University of Pennsylvania over ENIAC patent rights and that the company was incorporated on 8 December 1947. So 1946 is the break and 1947 the legal date.

The first customer was a government agency. The Census Bureau says that in 1946 Mauchly and Eckert secured a study contract from the National Bureau of Standards to begin work on a computer for the Census Bureau. The study was scheduled for six months and took about a year, and the result was the specification for UNIVAC. The Census page says the two began building UNIVAC I in 1948. The IEEE history wiki describes about five years of development under a Census Bureau contract.

Building took money the founders did not have. A small contract kept them going: on 9 October 1947 the company signed to build the BINAC, and the Computer History Museum gives the price as $278,000. The page calls BINAC the first stored-program electronic digital computer completed in the United States, with a 512-word memory and about 700 vacuum tubes where ENIAC used 18,000. It was the company's only other machine, built for Northrop Aircraft, per the Museum's company page.

The financing then broke. According to the Museum, American Totalisator, a racetrack betting machine maker, invested in 1948 and its support collapsed in 1949 when its founder died. The IEEE wiki names the chairman, Harry L. Straus, who died in a plane crash in October 1949. Talks with IBM and National Cash Register never produced a deal, and the founders sold to the first bidder. Remington Rand bought the company in 1950. The IEEE wiki adds that Remington Rand's calculating machine division was led by former Manhattan Project director Leslie Groves.

Meanwhile a key programmer arrived. The Museum says Grace Murray Hopper joined in 1949 as a senior mathematician. Her later work gave UNIVAC users a way to program in English-like words.

This era produced a design and a parent company. It also created a pattern: the technology was ready before the founders could pay for it.

For teams today: a product with one big customer and no cash runway is for sale, whether you say so or not. Plan the financing before the build is done.

Three Claimants (1951)

Manchester, Philadelphia and London all say first

The turning point is the calendar. Three machines went into service within about seven months, and each has a fair claim. We separate the claims below by what "first" means.

Take the claims one at a time.

Ferranti Mark 1. The Computer History Museum says the title of first commercially available general-purpose computer probably goes to Britain's Ferranti Mark I, with about nine units sold between 1951 and 1957. It descended from the Manchester University machines. The British government signed a contract with Ferranti in 1948, and the first machine arrived at Manchester in February 1951.

UNIVAC I. The date of delivery depends on the source. The IEEE history wiki gives 31 March 1951 for delivery to the Census Bureau. The Census Bureau's own history says a contract for the machine was signed on 31 March 1951 and the dedication was in June, with officials attending on 14 June 1951 at the Eckert-Mauchly Laboratory in Philadelphia. The Smithsonian says the first UNIVAC was delivered to the Bureau of the Census in early 1951, and the Computer History Museum says the Bureau received it on 30 March 1951, though the machine physically moved there months later. So 31 March 1951 is the date of record, but it is a signing in one source and a delivery in another. By the Smithsonian's count, some 46 copies had been installed by 1957, and the Computer History Museum gives the cost as about one million dollars each.

LEO I. J. Lyons, a British catering and food company, designed and built its own machine at Cadby Hall in London. The IEEE milestone proposal says LEO I was strongly based on Cambridge's EDSAC, that J.R.M. Simmons led the project and suggested the name Lyons Electronic Office, and that Dr. John Pinkerton was chief engineer. Bakery Valuations ran on 5 September 1951, and LEO took the calculation over completely on 29 and 30 November 1951. The same page describes a payroll system for 30,000 Lyons staff at 1.5 seconds per payslip against 8 minutes by hand. LEO Computers Ltd. was formed in 1954 to sell the machines, which the page counts as 11 LEO II and 61 LEO III.

A fair summary: Ferranti was first to deliver, UNIVAC was first sold in the US, and LEO was first to run a business's daily clerical work. Different sources pick different winners because they mean different things.

For teams today: "first" is a claim about a definition. When you say your product is first, name the definition, or a competitor will.

The Election and the Compiler (1951-1954)

The public meets the machine, and programmers meet a new tool

The turning point is a television broadcast. On the night of the November 1952 presidential election, CBS showed viewers a UNIVAC. The Computer History Museum says Charles Collingwood ran a mock-up console in CBS's New York studios while the real machine sat at the UNIVAC division in Philadelphia. An early printout gave Eisenhower's win 100-to-1 odds, and the third digit was truncated because the programmers had not expected either candidate to be favored so strongly. Eisenhower received 442 electoral votes and Stevenson 89. After midnight Remington Rand's Art Draper told viewers, "We were wrong. Next year we'll believe it." Collingwood's on-air question to the machine, "Have you got a prediction for us, UNIVAC?", produced dead air.

The scene shows a new problem for computing: people could now argue on live television about whether to believe a machine's answer.

The less visible change happened on the tape. Hopper's work on UNIVAC produced what the Computer History Museum calls A-0, completed in 1952, a program that lets a user write English-like words instead of numbers, written for UNIVAC I. The Museum notes it had several features of a modern compiler. Hopper recalled the effort in plain terms: "It was a selling job to get people to try it." Her successor systems, MATH-MATIC and FLOW-MATIC, shipped in 1957. The full line from A-0 to modern code is in the history of programming languages.

Customers were also learning what to trust. The Museum says early-1950s business computers spent roughly 10% of their time processing data and 90% handling cards and printing, and that UNIVAC's tape drives read and wrote about ten times faster than punched cards. Buyers were wary of tape because it held information they could not see. A Census Bureau official, per the same page, testified of UNIVAC: "We have never found it in error".

The first private buyer came late. The Museum says General Electric became UNIVAC's first commercial customer in 1954, using it for payroll, budget analysis and inventory at its appliance division. The machine was fully operational in 1956, and GE underestimated the integration work, which made other prospects nervous.

For teams today: a demo wins attention, but a first customer's rough integration sets the market's opinion. Staff the first deployments as if every reference call depends on them.

The IBM Reversal (1952-1960)

The company that sold punched cards sells computers

The turning point is IBM's decision to enter. IBM's history says the 701 began after the Korean War broke out in June 1950, that the design started in early 1951, and that Thomas Watson announced it to shareholders in April 1952. Ten government and defense customers had ordered by May 1952. Thomas Watson Jr. said the orders showed IBM "knew that we were in the electronics business". The 701 was a scientific machine. A business-oriented 702 followed.

The next move was a cheaper machine. IBM announced the 650 at Endicott, New York on 14 July 1953. Executives expected to sell about 50, and nearly 2,000 were installed before production ended in 1962. IBM's page calls it the first computer built for mass production and the "workhorse of commercial computing".

Meanwhile Remington Rand lost its lead. The Computer History Museum says the company's large management structure often frustrated engineers, many of whom left to found firms such as Control Data, and that many people, including J. Presper Eckert, believed it let IBM take the lead. The IEEE wiki adds that IBM won the SAGE military contract, which the page says gave it an advantage in computer production. Sperry and Remington Rand merged in 1955.

No single sales chart settles exactly when IBM pulled ahead, because installation counts differ by what is included. The documented parts are IBM's orders, IBM's volume on the 650, and the Museum's account of why the first mover stalled.

The era moved computers from a few dozen custom installations to a product line sold in volume. Its leftover problem was software: customers needed programs written, and the makers had to teach them how. For the wider path, see generations of computers.

For teams today: the first mover owns the proof, not the market. A larger rival with distribution and a cheaper offer can take a market you opened.

The Unisys Legacy (1955-present)

A name survives through mergers

The turning point is a series of mergers, not a product. The Computer History Museum says Remington Rand merged with Sperry in 1955 to form Sperry Rand, and that Sperry Rand later merged with Burroughs to form Unisys, which is still in business. The Smithsonian's UNIVAC I model was a gift of the Univac Division of Sperry Rand.

What lasted is the pattern. UNIVAC proved a computer could be a product sold to a statistics office, an insurer or a manufacturer. The Computer History Museum says Life and Casualty Insurance of Tennessee ran two UNIVAC I machines until they were shut down in 1970. The same ideas, tape for data and compilers for programs, carried forward into the mainframes that ran enterprise accounting for decades, machines later rebuilt around integrated circuits instead of tubes, and into the cloud services described in the history of cloud computing.

For teams today: products outlive the company that made them. Document why a system was built the way it was, because the maker may not be around to explain.

Key People

Key Facts

• The Ferranti Mark 1 was delivered on 12 February 1951, about six weeks before UNIVAC I's 31 March date.
• The Census Bureau dedicated UNIVAC I on 14 June 1951 in Philadelphia.
• By 1957 some 46 UNIVAC I machines had been installed, at about one million dollars each.
• LEO I ran Bakery Valuations on 5 September 1951 and took it over completely on 29 and 30 November.
• The IBM 650 was announced on 14 July 1953; IBM expected about 50 sales and installed nearly 2,000.
• On the 1952 election night UNIVAC's early printout gave Eisenhower 100-to-1 odds; he took 442 electoral votes.

The Pattern Behind the History

The same four moves show up in the first commercial computers. We call the sequence the Government Seed Cycle. It's our synthesis of the history, not a research finding.

1. A public buyer seeds the idea. The Census Bureau, the Army and the British government paid for the first designs.
2. The inventors run out of money. Eckert-Mauchly sold in 1950 before its first machine was delivered.
3. A larger firm buys the product. Remington Rand, then Sperry, held UNIVAC.
4. A rival with distribution wins the volume. IBM's 650 sold nearly 2,000 units against 46 UNIVAC Is.

What It Teaches Teams Today

• Name your buyer. The Census Bureau paid for a machine because its problem was already costly.
• Define "first" before you claim it. Delivered, sold, in service and running daily work gave three different winners.
• Protect your funding. The company that designed UNIVAC sold itself before it could ship.
• Plan for the follow-up buyer. IBM's cheaper, smaller 650 widened the market the first mover had opened.
• Know which date you mean. Contract, delivery, dedication and acceptance fell on different days.

UNIVAC shows that being first to market and being first to scale are different jobs. Next time you launch something new, ask who will fund version two and who is positioned to sell it at volume.