Tommy Flowers and Colossus

Abstract

Tommy Flowers was a Post Office engineer who built Colossus (the world’s first large-scale programmable electronic computer) for Bletchley Park in 1943 to break the Lorenz cipher used by the German high command. He used 1,500 vacuum tubes when everyone told him vacuum tubes were too unreliable to work at that scale. He was right and they were wrong. After the war, his work was classified for thirty years. While Flowers lived in relative obscurity, ENIAC was celebrated as the first electronic computer, and when Flowers finally told his story in the 1970s, computing history had to be substantially revised.

Colossus Mark 2 operated by Dorothy Du Boisson and Elsie Booker, 1943. Image: public domain, via Wikimedia Commons.

The Engineer from Poplar

Thomas Harold Flowers was born on December 22, 1905, in Poplar, east London, a working-class neighborhood of dockworkers and small tradesmen. He left school at fourteen to work as an apprentice in the mechanical engineering workshop of the Royal Arsenal at Woolwich, taking evening classes to improve his education. He eventually earned a degree in electrical engineering through correspondence courses while working, a pattern of self-improvement through determination that recalls George Boole’s earlier path.

In 1930, Flowers joined the Post Office Research Station at Dollis Hill in north London, where he worked on telephone exchange technology. The Post Office ran Britain’s telephone network, and its research laboratory was one of the world’s most advanced centers for electronics. By the late 1930s, Flowers was working on electronic telephone exchanges, systems that used vacuum tubes for switching, at a time when the engineering consensus held that vacuum tubes were inherently unreliable for large-scale applications.

Flowers’s experience contradicted this consensus. He discovered that vacuum tube failure was mostly caused by thermal stress during power cycling, turning tubes on and off. If tubes were left permanently powered, they ran for years without failure. This observation, counterintuitive to those who understood vacuum tubes primarily as laboratory devices, was the key insight that made Colossus possible.

Bletchley Park and the Lorenz Cipher

Bletchley Park was Britain’s wartime code-breaking center, staffed with mathematicians, linguists, chess players, and crossword solvers (anyone with evidence of puzzle-solving ability. The center’s best-known work was breaking Enigma) the cipher machine used by German Army and Navy forces, using electromechanical Bombe machines designed by Alan Turing.

But there was a second, harder problem: the Lorenz SZ40/42 cipher machine, used for communications between Hitler and his senior commanders. While Enigma produced output that looked like random letters, the Lorenz machine used a teletype-compatible bitstream encoding and a complex system of twelve rotors operating in concert. Breaking it required analyzing streams of bits (hundreds of thousands of bits per message) and the volume was beyond what the Bombes could address.

The mathematical framework for attacking Lorenz was developed by Bill Tutte (who reconstructed the machine’s logical structure from a single intercepted operator error in 1941, without ever seeing the physical device: on August 30, 1941 a German operator in Athens sent the same long message twice on the same key setting, abbreviating the second version, which let John Tiltman recover the key stream and Tutte deduce the twelve wheels behind it) and Max Newman, who recognized that the attack required high-speed comparison of bit streams and that a machine was needed.

The first attempt was Heath Robinson (in service from June 1943), a machine using paper tape loops to represent both the ciphertext and the pattern wheel settings. It worked, but the paper tapes were fragile, the synchronization unreliable, and the speed insufficient.

Building Colossus Against Expert Opinion

Max Newman approached Tommy Flowers in February 1943 with the Heath Robinson problem. Flowers proposed a radical solution: eliminate one of the paper tape loops by implementing the Lorenz wheel patterns electronically, in vacuum tubes. The machine would read ciphertext from one tape at high speed and compare it electronically against simulated wheel patterns stored in tube circuits.

The proposal required approximately 1,500 vacuum tubes, an order of magnitude more than any machine then in existence. The expert consensus at Bletchley was that this was technically impossible: 1,500 vacuum tubes would fail constantly, the machine would never work reliably, and the project would be a waste of scarce wartime resources.

Bletchley greeted the proposal with, in the standard account, great scepticism, and did not commission it. Flowers was working at Dollis Hill, not Bletchley, and had no champion there.

He built it anyway. W. Gordon Radley, director of the Post Office Research Station, backed him, and Flowers built Colossus at Dollis Hill with Post Office resources and staff. He invested £1,000 of his own money (approximately £50,000 in 2024 terms) when materials were not otherwise available.

The first machine ran at Dollis Hill in December 1943, was taken apart, and as Colossus Mark 1 arrived at Bletchley Park on 18 January 1944 and attacked its first message on 5 February. The machine used about 1,500 vacuum tubes (other accounts, including the HNF’s, give 1,600), read paper tape at 5,000 characters per second, and could perform Boolean logic operations at electronic speed on the comparison results. It did not break the Lorenz cipher directly (the code-breaking logic was implemented on its plugboard by human operators) but it performed the exhaustive statistical analysis that made breaking possible, in hours rather than weeks.

Colossus Mark 2, with 2,400 tubes and five parallel processing streams, was operational by June 1, 1944, three days before D-Day. By the war’s end, ten Colossus machines were running at Bletchley, processing intercepted German high-command communications. The intelligence produced (called Ultra) is credited by historians with shortening the war by at least two years.

Info

Colossus was not a general-purpose computer. It could not be programmed to perform arbitrary computations, its Boolean logic operations were fixed in hardware and its plugboard configuration determined what comparison it performed on the tape data. It had no stored program, no arithmetic unit for general calculation, and no means of branching on computed results in the modern sense. The question of whether Colossus counts as “the first computer” depends on definitions that historians continue to debate. What is not debatable: it was the first large-scale electronic digital device capable of conditional operations, operating at electronic speeds, that processed real-world data to produce actionable results.

The Thirty-Year Silence

When the war ended, eight of the ten machines were broken up and their parts returned to the Post Office; the surviving two went to GCHQ, where one was dismantled in 1959 and the other in the 1960s. Everyone who had built or operated them was sworn to secrecy under the Official Secrets Act. The reasons were strategic: Britain wanted potential adversaries to believe the Lorenz cipher had been broken by other means, and the ability to read ciphertext that others believed secure was a permanent intelligence advantage worth protecting.

Flowers was not permitted to speak about Colossus or claim credit for his work. He was paid nothing beyond his salary. He had invested £1,000 of his own money in the project and applied for reimbursement, and was refused. When he tried to apply his wartime experience to postwar projects, he could not describe what he had done.

During the 1950s, Flowers participated in the design of a National Data Processing Service for Britain (a network of shared mainframes for commercial use) which was rejected by the Treasury. He believed that if he could have described Colossus, the project might have been authorized; Britain might have built its digital infrastructure a decade earlier. He could not.

Meanwhile, ENIAC (completed at the University of Pennsylvania in 1945, a year after Colossus Mark 1) was publicly celebrated as the world’s first electronic digital computer. Flowers and his colleagues watched in silence.

The secrecy began to lift in the 1970s. The computer scientist Brian Randell pieced together what he was allowed to publish in a 1972 paper and a 1976 conference talk, and the BBC series The Secret War (1977) put Flowers on screen. F.H. Hinsley’s official history of British intelligence acknowledged Colossus’s existence in 1981. Flowers spoke publicly for the first time in the late 1970s. A replica of Colossus Mark 2, begun in 1993 by the engineer Tony Sale and built by volunteers at The National Museum of Computing at Bletchley Park, was completed in 2007 and is operational. It was inaugurated on November 15–16, 2007 with a “Cipher Event”: the Heinz Nixdorf MuseumsForum in Paderborn transmitted messages enciphered on a real Lorenz SZ42, and the rebuilt Colossus broke one in 3 hours 35 minutes. A radio amateur in Bonn, Joachim Schüth, was faster: his own program, written in Ada, decrypted the message on a laptop in under a minute. After Sale’s death in 2011 the Computer Conservation Society named a biennial award for computer restoration after him. In January 2024 GCHQ released photographs of a Colossus still in existence in 1963.

Legacy

Tommy Flowers died on October 28, 1998, at age ninety-two, having lived long enough to receive some recognition: an MBE (Member of the Order of the British Empire, 1943, secret), an honorary Doctor of Science from Newcastle University, a blue plaque (2023) at the former Post Office Research Station, and the recognition of historians who revised computing’s canonical narrative when the Colossus story emerged.

Max Newman, who had brought Flowers the problem and whose section ran the machines, moved to the University of Manchester in 1945, obtained the Royal Society grant that funded its Computing Machine Laboratory, and in 1948 recruited Alan Turing there, a direct line from Colossus to the Manchester Baby. Computing stayed in the Newman family: his son William Newman co-wrote Principles of Interactive Computer Graphics (1973) with Robert Sproull, the standard graphics textbook of its decade, and spent the late 1970s at Xerox PARC working on interactive systems.

The structural significance of Colossus is independent of the definitional debates about what counts as “the first computer.” Flowers demonstrated empirically that vacuum tubes could be made reliable at scale (not by improving them, but by keeping them powered on) which was the key enabling insight for every subsequent vacuum-tube computer, including ENIAC. His approach influenced the engineers who built ENIAC (some of whom had indirect knowledge of what was being done in Britain) and established the practical parameters within which first-generation electronic computing was possible.

The Colossus story is also a story about institutional resistance to innovation. Every expert Flowers consulted told him 1,500 vacuum tubes would not work. He built it, kept it powered on, and proved them wrong. The lesson is not that experts are unreliable, but that expertise in the behavior of a technology at previous scales may not extrapolate to new scales, a lesson computing has relearned repeatedly.