Soviet Silicon Valley

On the afternoon of May 4, 1962, Nikita Khrushchev climbed the steps of an unremarkable building in Leningrad and walked into a laboratory run by two men he had never met. Their files said they were Greek and South African by birth, a lie the KGB had stitched together with the casual confidence that bureaucracies reserve for their own products. In fact they were Americans. Philip Staros, the laboratory’s director, had been born Alfred Sarant in the Bronx. Joseph Berg, the chief engineer, had been Joel Barr of Brooklyn. They had spent the war stealing radar designs from the United States Army Signal Corps at Fort Monmouth, New Jersey, and had spent the dozen years since their flight from the FBI assembling, in modest stages, the components of what they hoped would become a Soviet electronics industry. They had about an hour to convince the First Secretary of the Communist Party of the Soviet Union to give them a city.

The performance had been carefully arranged. Staros, the showman, walked Khrushchev past a small computer that responded to spoken queries with biographical facts about the visitor himself. He pressed into the Premier’s hand a tiny radio, smaller than a thumbnail, that worked. He talked, in the Russian he had learned by force in the 1950s, about a coming revolution in which thinking machines would shrink to the size of postage stamps and reorder Soviet industry the way electricity had reordered tsarist Russia. Berg hovered, quiet, technical. By the end of the visit Khrushchev had embraced Staros and pulled Berg in beside him and announced that any apparatchik who got in their way could come and talk to him directly. Three months later, on August 8, 1962, Khrushchev signed the postanovlenie authorizing a closed scientific city dedicated to microelectronics, on a site north of Moscow near the railway station of Kryukovo. In January of the following year the place got a name. Zelenograd, Green City. The Soviet Silicon Valley had been ordered into existence.

The city had been there in some form already. Soviet planners had begun laying it out in 1958 as a satellite town for textile workers, a garden-city experiment that the architects Igor Rozhin and later Igor Pokrovsky had borrowed from British new-town planning, with wide boulevards and microdistricts of apartment blocks set among preserved pine forest. What changed in 1962 was not the masonry but the mission. Alexander Shokin, chairman of the State Committee for Electronic Technology and the man who would in 1965 become the founding minister of a full Ministry of Electronic Industry, had pushed for a dedicated microelectronics center for years. Staros and Berg’s audience with Khrushchev gave him the political cover to convert the half-built town from looms to lithography. Shokin would run Soviet electronics for the next quarter century. He would also outmaneuver Staros and Berg almost immediately.

To understand how a Soviet city ended up being founded by two men from the Bronx and Brooklyn requires going back to a Manhattan apartment in the late 1930s. Barr and Sarant had met as engineering students at the City College of New York, the polyglot, mostly Jewish, mostly radical campus that produced an outsized share of America’s left during the Depression. Both joined the Communist Party. Both took electrical engineering jobs that brought them into the orbit of military research during the war, Barr at the Signal Corps and Sarant briefly at Bell Labs and then at Fort Monmouth. They were recruited, along with their friend Julius Rosenberg, into a Soviet espionage network that handed Moscow several thousand pages of technical documentation on radar, fire-control, and proximity fuzes. The historian Steven Usdin, drawing on long interviews with Barr in his last years and on declassified Soviet archives for his book Engineering Communism, estimated that the two of them passed perhaps 9,000 pages covering more than a hundred weapons programs. They were among the most productive industrial spies the United States ever produced.

When David Greenglass was arrested in June 1950 and Julius Rosenberg shortly afterward, the network unraveled. Sarant slipped across the Mexican border in July with the woman who would be his second wife. Barr, already in Paris, vanished a few weeks later. Soviet intelligence reassembled the two of them in Prague in 1951 and put them to work designing an automated anti-aircraft system at a Czech defense plant. They were good at it. By 1956 the KGB had decided they were too useful to leave in a satellite, and moved them to Leningrad with new names, new biographies, and a research institute of their own, the Special Laboratory at the Leningrad Construction Bureau, which everyone called LKB. There they built the UM-1, a control computer that could fit on a desk, and the Uzel, a computer that went into Soviet submarines and, in updated form, was still running in foreign navies into the twenty-first century. They were not, in any technical sense, ahead of the Americans. By Soviet standards they were miraculously ahead of every other Soviet team.

This is what Khrushchev saw on the afternoon of May 4, 1962. Two of his own engineers, performing the kind of theater that Western executives staged for journalists every quarter, except that here the audience was the head of state. Integrated electronics, in some recognizable form, working in front of him. His briefers had recently warned him that the Americans were building computers smaller than refrigerators and intended to use them to guide missiles. The Soviet electronics industry, scattered across dozens of ministries and research institutes, was producing point-contact transistors and vacuum tubes at scales and prices that bore no relationship to what was happening at Fairchild and Texas Instruments. Khrushchev was a peasant by origin and a believer in physical demonstrations. The radio in his hand was the demonstration. He gave them their city.

Staros had hoped to run that city himself. He did not get the chance. Shokin saw to that, in the bureaucratic style that had carried him through Stalin’s purges and would carry him through Brezhnev’s stagnation. Zelenograd’s leadership filled with Shokin’s people. Staros was kept on as a scientific director from a Leningrad base, his influence steadily reduced. When Khrushchev was forced into retirement in October 1964 and replaced by Brezhnev and Kosygin, the protective umbrella over the two Americans collapsed. Staros tried, after a respectful interval, to ask the new General Secretary for the kind of audience Khrushchev had granted. He was refused. By the end of the 1960s Berg had been pushed out of administrative work entirely. Staros lived on as a scientist of declining importance in Leningrad until his death in 1979. Berg outlived the country he had spied for, dying in Moscow in 1998 in an apartment on Tverskaya, a few hundred meters from what had been, in the 1950s, the FBI’s recurrent nightmares.

The city they had imagined kept growing without them. By 1963 the institute that would become Angstrem had been chartered as NII-336. By 1964 the scientific research institute of molecular electronics, NIIME, had spun off the production unit that would be named Mikron. The Moscow Institute of Electronic Technology, MIET, opened its doors in 1965 and within five years occupied a striking concrete-and-glass campus designed by Felix Novikov and Grigory Saevich in a brutalist idiom that proclaimed, with the architectural confidence of the late Khrushchev years, that this was a city of the future. Within a decade more than twenty research institutes and a dozen production lines had been concentrated inside the Zelenograd cluster, ringed by the security perimeter that closed cities required. By the late 1970s tens of thousands of engineers worked there. On paper, it looked like exactly what the Americans had built around Stanford.

It was not. Zelenograd’s first integrated circuit, marketed under the family name Tropa and produced at Mikron beginning in 1965, was a recognizable knockoff of a Texas Instruments part. The next was a knockoff of a Fairchild part. By the early 1970s, Soviet engineers had been instructed, by a policy decision that even Soviet electronics historians would later call the most damaging single choice in their industry’s history, to stop designing original chips and to copy Western parts directly. The logic was straightforward. Reverse-engineering an American integrated circuit was difficult but tractable, particularly with KGB-supplied samples. Designing a competitive one from first principles, with Soviet equipment and Soviet supply chains, was harder. Each generation of American devices had been validated by a market that immediately demanded the next generation. Each Soviet generation had been validated by a ministerial sign-off. The American chip ecosystem ran on iteration. The Soviet one ran on permission.

What Zelenograd lacked was the loop. Fairchild sold transistors to a hobbyist who built a missile guidance set and complained about the tolerances; Fairchild used the complaint to redesign the next batch; the redesigned batch went into a commercial computer that found a customer in an oil company; the oil company demanded reliability data that drove a new test methodology; the new methodology produced yields that funded the next process. Each cycle in California ran in months. In Zelenograd the loop barely existed. The customer, almost always, was the Soviet military, mediated by a planning ministry that translated specifications into five-year quotas. The user of the chip rarely spoke to the designer of the chip, and when they did the conversation was filtered through Gosplan, where the unit of analysis was the plant’s quarterly output target rather than the device’s performance. A radar engineer in Saratov who needed a faster logic gate had no mechanism to drive Mikron’s process to make one. He ordered what was on the catalog, and if the catalog part failed, he filed paperwork. The paperwork, by the time it reached anyone with the authority to act on it, was years stale.

Soviet officials understood this. Shokin’s ministry tried, in the late 1960s and again in the 1970s, to set up internal mechanisms that would mimic the feedback loops of capitalist industry, including consumer-oriented production targets and a small line of pocket calculators and digital watches that Mikron and Angstrem turned out for the domestic market. The volumes were trivial. The civilian Soviet economy did not, except in narrow categories, demand chips. There was no Soviet Heathkit, no Soviet Apple, no garage culture inventing uses for unwanted silicon. The few Soviet attempts at a microcomputer industry, including the Elektronika line, ran on Western copies and never reached the volumes that would have driven Soviet fabs up the learning curve. A 1986 CIA Office of Soviet Analysis assessment, declassified in the 1990s, concluded that Soviet microelectronics ran two generations and roughly five years behind the United States and Japan, that the gap was not narrowing, and that even a successful follower strategy would condemn the USSR to that distance permanently because the leaders kept moving. By the early 1980s Mikron was producing at process geometries that American fabs had abandoned for civilian work. By the late 1980s, two generations had become three.

Foreign visitors to Zelenograd in the 1970s came away struck by the gap between the city’s appearance and its output. The boulevards were broad and well-paved. The cafeterias served better food than most Soviet workplaces. The apartment blocks had hot water, and the schools were excellent. Set in pine forest and laid out with the deliberate quietness of a planned capital, the city was as physically pleasant as anywhere in the Soviet Union. What was missing was the noise. Stanford in 1975 sounded like a hobbyist club having an argument that occasionally became a corporation. Zelenograd sounded like a research institute. Engineers worked, were paid, met production targets, and went home. They did not, with rare exceptions, found companies. They could not. There were no companies to found. The plants were ministerial. The institutes were ministerial. The buildings on the boulevards belonged to the same office, and the office answered to the Politburo.

By the time Mikhail Gorbachev took power in 1985 and began, with characteristic energy, to fret in public about the Soviet Union’s microelectronics deficit, the structure had set hard. Gorbachev visited Zelenograd in 1985 and again in 1991, demanded reform, demanded computerization of the economy, demanded chips. The chips Mikron and Angstrem could deliver were, by then, copies of late-1970s American designs, produced in volumes that the People’s Liberation Army of China would soon dwarf. Western export controls under the COCOM regime tightened the supply of stolen samples and fab equipment that Soviet copying had relied on, and the gap widened further. When the Soviet Union dissolved at the end of 1991, Zelenograd’s military customer evaporated overnight. The cluster lost what had been, in effect, its only customer. Mikron survived in reduced form, eventually finding a niche in transit cards. Angstrem limped along on Soviet-vintage processes for another twenty years before going bankrupt in 2017 without ever having produced a chip below the 250-nanometer node, a geometry that Intel had reached in 1997.

What had gone wrong was visible from the start to anyone willing to read the two systems as feedback loops rather than as factories. The Americans had built their industry by accident, on top of small decisions made by engineers who could leave their employers, raise capital from venture investors chasing the next thing, sell to customers who could buy elsewhere if the product was bad, and read each other’s papers in journals no one had to clear with a censor. The Soviets had built theirs on purpose, in a single coherent plan signed by the General Secretary, financed by the central budget, populated by handpicked specialists, and shielded from competition. The American industry generated chips. The Soviet industry generated five-year reports. Khrushchev had thought, on May 4, 1962, that the difference between the two systems was a matter of priority and resources, and that with the right city and the right men he could close it. The right city was built. The right men, two of whom were standing in front of him, had been delivered. The thing did not work, because the thing that mattered could not be ordered into existence by decree.

Staros wrote in his last years, in unpublished notes that Usdin would later draw on, that he had come to understand the limits of his project only after Khrushchev fell. The thing he had been trying to build, he realized, was not a city or an industry or a set of machines. It was a culture in which engineers and customers and capital argued with each other in real time. He had assumed that with enough talented people and enough state backing, the argument would start. It did not, because the argument required participants who could walk away, and in Zelenograd, no one could. The city had been planned. The argument could not be.

A different argument, conducted in different rooms several thousand kilometers to the east, would soon produce a different question for Soviet planners. If you could not invent the chips faster than the Americans, perhaps you could simply, and at scale, take them.