When Konrad Zuse switched on the Z3 in his Berlin workshop in May 1941, he had just built the first programmable computer in history out of roughly 2,000 surplus telephone relays, and almost no one in the world knew it existed because the war buried the story for two decades.

Individual programming in a dimly lit room with dual monitors.

On May 12, 1941, a 30-year-old civil engineer named Konrad Zuse stood in a cramped workshop at Methfesselstraße 7 in Berlin-Kreuzberg and demonstrated the Z3, a refrigerator-sized cabinet of clicking telephone relays that could be programmed to perform any calculation its operator could describe in binary. His audience that day was a small group of scientists from the Deutsche Versuchsanstalt für Luftfahrt, the German Laboratory for Aviation, including the professors Alfred Teichmann and Curt Schmieden. The machine they watched run through its paces was the first fully functional, programmable, automatic digital computer ever built. Almost no one outside that workshop would learn about it for two decades. Within two years, an Allied bombing raid would reduce the only working Z3 to splinters and scorched wire, and the man who invented programmable computing would soon be hauling a successor machine across the Bavarian countryside in the chaos of the last weeks of the war.

The story of the Z3 is the story of an invention that arrived in the wrong country at the wrong moment, then got buried by a war it had nothing to do with winning.

The workshop on Methfesselstraße

Zuse had quit his job as a stress analyst at the Henschel aircraft factory in 1935 because he was tired of doing the same arithmetic by hand, over and over, on tables of numbers that filled entire notebooks. Aircraft stress calculations in the 1930s meant pages of multiplication and division using slide rules and mechanical adding machines, and Zuse had decided there had to be a better way. He moved back into his parents’ apartment in Berlin and started building.

The first machine, the Z1, occupied most of the family living room. It was a mechanical computer made of thin metal sheets cut by hand, with sliding pins that represented binary digits. According to Hackaday’s detailed reconstruction of his early work, Zuse and a small circle of friends financed the project themselves, cutting the parts with a jigsaw in the apartment. His mother kept feeding them. His father was retired and patient. The Z1 worked, sort of, but the mechanical linkages jammed constantly.

By the time he started the Z3, Zuse had outgrown the family living room. He had rented his own workshop at Methfesselstraße 7 in the Kreuzberg district of Berlin, and it was there, not at his parents’ place, that the breakthrough machine took shape. The Z2 in between had replaced the arithmetic unit with electromechanical telephone relays salvaged from the German postal service. Those relays were the breakthrough. They were cheap, they were everywhere, and they switched cleanly between two states, which was exactly what binary arithmetic needed.

What 2,000 relays could do

The Z3, finished in 1941, used roughly 2,000 of those surplus relays. About 600 of them formed the arithmetic unit. The remaining 1,400 or so handled memory and control. (Some later sources put the number as high as 2,600. The figure varies depending on which parts of the machine are counted.) The Z3 ran on a 5-to-10 Hz clock, which means it could perform between five and ten operations per second. A multiplication took about three seconds. Modern processors do billions of operations in that same second.

But the Z3 had something none of its contemporaries had. It was fully programmable, using punched film tape, and it worked in binary floating-point arithmetic, a numerical format Zuse had designed himself years before anyone else implemented it in hardware. It could handle very large and very small numbers with the same machinery, a problem that bedeviled mechanical calculators of the era.

It was, by the strict definition computer scientists later applied, Turing-complete, though Zuse himself never used that language and the proof came decades later. Wired notes that the Z3 is now generally recognized as the first fully functional, programmable computer, predating the American ENIAC by about five years and the British Colossus by two.

Why the war buried it

The reason most people have never heard of Konrad Zuse comes down to geography and timing. He built his machine inside Nazi Germany during a war the regime was losing. The Reich’s research priorities ran toward V-2 rockets, jet engines, and submarine warfare. A young civilian engineer with a calculating machine in a Berlin workshop was not on anyone’s list.

Zuse did receive modest funding from the DVL after the partial success of the Z2 the year before, and a government supervisor from the Reich Air Ministry, Joseph Jennissen, oversaw the project as a Reichsmark contract to Zuse’s small company, Zuse Apparatebau. But after the May 1941 demonstration, when Zuse asked for further funding to replace the relays with faster electronic switches, the request was refused. The authorities deemed the project not war-important. There was no press. There was no public announcement. Germany was at war with the Soviet Union within six weeks.

The Allies, building their own early computers in secret on the other side of the front lines, had no way of knowing the Z3 existed. The American team behind ENIAC at the University of Pennsylvania, the British codebreakers at Bletchley Park building Colossus to crack the Lorenz cipher, the Harvard team assembling the Mark I under Howard Aiken, all of them worked in parallel ignorance of what Zuse had already done.

The bombing raid

On the night of December 21, 1943, an Allied air raid over Berlin destroyed the building on Methfesselstraße where Zuse kept the Z3. The Z1 and Z2 prototypes had already been lost in earlier bombings. The Z3, the only working copy, was reduced to debris. There are no surviving photographs of the original machine in operation. Only one engineering drawing of the Z3 survived the war. Everything later generations know about its internal design comes from Zuse’s own notes, drawings, and patent applications, plus a working replica he supervised in the 1960s for the Deutsches Museum in Munich.

By that point Zuse had already started on the Z4, a more ambitious successor. As the Red Army closed on Berlin in early 1945, he packed the Z4 into crates and left the city. The machine was first moved by truck through Göttingen, where Zuse demonstrated it to scientists at the Aerodynamische Versuchsanstalt under Albert Betz, and then south to the small Alpine village of Hinterstein in the Allgäu, where the Wehrmacht truck arrangement was helped along by Wernher von Braun’s staff. There Zuse hid the Z4 in a shed behind the Hotel Steinadler and in the local gymnasium, where it sat covered in hay for the better part of two years. The machine was eventually installed at ETH Zurich in 1950, where it became the first commercially deployed computer in continental Europe.

The twenty-year silence

The English-language history of computing was written in the 1950s and 1960s, largely by Americans and Britons, and largely from American and British primary sources. ENIAC got the credit. Colossus stayed classified by the British government until the 1970s, but its existence at least lived in institutional memory. Zuse’s work lived only in German, in patents and trade journals that the postwar Anglo-American computing establishment rarely consulted.

The recognition came slowly. The German computing community knew. The Zuse Institute Berlin, a major research center for applied mathematics and computer science, carries his name today. By the 1960s, English-language histories began to acknowledge him. By the time he died in 1995, he had been awarded honorary doctorates, the Werner von Siemens Ring, the Harry H. Goode Memorial Award, and a place in the standard textbooks, though usually as a footnote next to ENIAC rather than as the headline.

The replica in Munich

Zuse and his company supervised a working replica of the Z3 in 1961, rebuilt from his preserved drawings and patent material. That replica sits today in the Deutsches Museum, clicking and humming when curators run it for visitors, the relays snapping shut in sequence as the machine grinds through a calculation. It is slow enough that you can hear each operation as a distinct sound, a percussive clatter that lasts a few seconds before the answer appears on the output display.

The original Z3 ran on about 4,000 watts and weighed roughly a ton. Anyone who has heard the replica understands the sonic impression. Two thousand telephone relays switching in concert sounds like a room full of typewriters being fired from a small cannon.

What the Z3 could not do, and what it could

The Z3 had no conditional branching in the modern sense. Its program tape ran straight through, instruction by instruction, without the ability to jump backward based on a calculated result. Zuse worked around this by physically splicing the tape into loops when he needed repetition. In 1998, the computer scientist Raúl Rojas demonstrated that the Z3 was nevertheless Turing-complete in principle, because its instruction set could be coaxed, with enough tape and enough patience, into simulating any computation.

In practice, the Z3 was used for what the DVL had paid Zuse to build. It crunched matrices of complex numbers for aerodynamic problems, including the wing-flutter analysis developed by the German aerodynamicist Hans Georg Küssner, the same researcher whose name is attached to the Küssner effect in unsteady aerodynamics. The work would have taken human computers, almost all of them women in that era, weeks to grind through by hand. The machine that would eventually evolve into the device in your pocket started its working life calculating how fast a wing could vibrate before it tore itself off an airplane.

The patent that almost was

Zuse filed a patent application in June 1941 covering the architecture of the Z3, registered with the Reich Patent Office under the file number Z26476 IXb/42m. The application worked its way through the German patent system across the war and the postwar decades, with the Triumph-Werke company in Nuremberg leading the opposition. The Federal Patent Court rejected it definitively in 1967, twenty-six years after the original filing, on the grounds of insufficient inventive step. By that point ENIAC’s patents had been filed, contested, and partially overturned in American courts. The history of who invented the computer had hardened into a story that did not have much room for a young German engineer working with surplus relays in a Kreuzberg workshop.

Zuse himself, by his own later account, was disappointed but not destroyed by the ruling. He kept building machines. His company, Zuse KG, produced commercial computers throughout the 1950s and early 1960s before financial trouble pulled it into the orbit of Siemens, which absorbed it in 1967. He painted in his spare time, mostly landscapes and portraits, some of which he had sold to farmers in Hinterstein in 1945 to keep the family fed. He wrote a memoir in which he described the war years in plain, unsentimental prose.

The thing about telephone relays

The components that made the Z3 possible were not built for computing. They were built for routing voice calls through the German telephone network in the 1920s and 1930s. Each relay was a small electromagnet that pulled a metal contact closed when current flowed through its coil, then released it when the current stopped. Stacked by the thousand and wired into a logical pattern, those simple switches became something none of their designers had imagined.

The same pattern repeats throughout the history of computing. Vacuum tubes were built for radios. Transistors were built to replace vacuum tubes. The integrated circuit was built to miniaturize the transistor. Each generation of computing hardware started as something else and got pressed into a new job by someone willing to look at a familiar object and see a switch.

Konrad Zuse looked at a pile of surplus telephone relays in the late 1930s and saw a calculating machine. The relays are gone, the original Z3 is dust, and the building at Methfesselstraße 7 in Kreuzberg was destroyed in the same raid that took the machine. What survived is the idea, encoded now in every smartphone and laptop and traffic light, that a machine can be told what to do in a language of ones and zeros and will obey, reliably, one click at a time.

Subscribe to our newsletter!

Our latest tutorials delivered straight to your inbox

Make Tech Easier Editorial Team Avatar

Read next

On 15 September 2017, NASA engineers deliberately steered the Cassini spacecraft into Saturn’s atmosphere at 70,000 miles per hour, not because the mission had failed but because they were terrified of where it might fall if they left it in orbit.
At Kyoto University’s Seto Marine Biological Laboratory on the Shirahama coast, biologist Shin Kubota has watched the same Turritopsis dohrnii jellyfish die and un-die so many times in his tank that he has started writing karaoke songs about an animal that, for all practical purposes, just refuses to stay dead.
In August 1993, Garry Trudeau spent a week turning Apple’s new Newton MessagePad into a Doonesbury joke, including a panel where ‘Catching on?’ became ‘Egg freckles,’ and the $699 handheld never fully outran the punchline
In 1945, Ruth Bourne was one of the young Wrens operating more than 100 Bombe machines at Eastcote, phoning in good stops to Bletchley Park without knowing that Hut 6 codebreakers sometimes cheered on the other end of the line
In 2022, NASA engineers in Pasadena watched a new image come down from Perseverance and saw a small heap of broken rocks resting on the Martian surface in a pattern that looked almost intentional, and the explanation they eventually offered involved a vanished river that has not flowed in three billion years.
In 2023 a single corrupted bit of code sent Voyager 1 into months of unintelligible static, and a team of engineers — many of them retired, called back specifically because no one else understood the system — managed to locate the fault and revive a spacecraft 24 billion kilometres away
Every ‘free’ app on your phone is, on average, sending data about you to between 5 and 10 different companies you have never heard of, and the legal disclosures that authorize this are buried in privacy policies that, for the worst offenders, run longer than entire novellas — Meta’s runs over 19,000 words, and some telecom policies exceed 100,000
What is Emulation? Benefits, Downsides and More