In the autumn of 1957, two men climbed a wooden staircase in an old textile mill beside the Assabet River in Maynard, Massachusetts. The mill had been built in the previous century to make woolen blankets, some of which had been sent to soldiers during the Civil War. The looms were long gone by the time the two men arrived; the building smelled of dust, machine oil, and a century of labor. They rented just under nine thousand square feet for three hundred dollars a month and furnished it with borrowed, secondhand furniture.
Behind them stood seventy thousand dollars in investment, a large measure of confidence, and a company name chosen with deliberate care, because the man funding them was uneasy about one word in particular: computer. That company would become the second-largest computer manufacturer in the world. Within four decades, it would vanish completely. The two men were Ken Olsen and Harlan Anderson.
Olsen was born in Bridgeport, Connecticut, in 1926, the son of a machine-tool designer of Norwegian and Swedish descent. As a boy he was quiet and methodical, the kind of child neighbors brought their broken radios to because he had a patience for taking things apart that others lacked. He served in the Navy near the end of World War II, then enrolled at MIT, earning a bachelor’s degree in 1950 and a master’s in electrical engineering two years later. At MIT’s Lincoln Laboratory, Olsen worked on SAGE, a Cold War air-defense network designed to track Soviet bombers approaching North American airspace in real time.
It was important, well-funded work. But the machine that changed how Olsen thought about computing was a smaller device kept nearby: an experimental computer called the TX-0. Olsen noticed something that the budget planners around him did not. Students would wait in line for hours, sometimes all night, for a chance to use the TX-0.
Meanwhile, a fast and expensive IBM machine in the same building sat idle most of the time. The difference was not processing speed. The TX-0 let a person sit directly in front of it, type, see the reaction, adjust, and try again. With the IBM machine, you handed your punch cards to an operator and waited until the next day for an answer.
Olsen became convinced that the future of computing was interactive: sitting directly at the machine and working in real time. His colleague Harlan Anderson shared that belief. Together they imagined something that seemed almost laughable at the time. What if a computer did not have to be kept behind glass in an air-conditioned room?
What if it could sit on a workbench on a factory floor or in a laboratory? They needed money to build it, and in the late 1950s American venture capital firms would not go near computers. Most computer startups had failed or were failing, and the word itself carried a kind of financial stigma. One man was willing to take the risk: Georges Doriot, a French-born Harvard Business School professor who ran American Research and Development Corporation, one of the country’s first venture capital firms.
Doriot invested seventy thousand dollars for a majority share in Olsen and Anderson’s new company. He advised them not to use the word computer in their name, because a recent magazine article had claimed that almost no one in the computer business was making a profit. They called themselves Digital Equipment Corporation, or DEC. For the first three years, DEC sold no computers at all.
The company sold small electronic logic modules that other engineers used to design their own circuits. It was safe, profitable, ordinary work, and it gave Olsen and Anderson a reputation for reliability before they could sell their bigger idea. That idea arrived in 1960 with a machine called the PDP-1, designed by an engineer named Ben Gurley. They called it a programmed data processor rather than a computer, again avoiding the word.
It sold for about one hundred twenty thousand dollars, a fraction of the cost of an equivalent IBM machine, and it brought something unusual for the time: a screen and a keyboard. A person could sit in front of it and interact directly, just as Olsen had watched students do with the TX-0. About fifty PDP-1s were built. One of those machines, installed at MIT, found a use its makers had never imagined.
In 1962, a programmer named Steve Russell and a group of students spent about two hundred hours building a two-player space combat game called Spacewar. It spread from machine to machine, from installation to installation, until DEC itself began shipping new PDP-1s with the game preloaded as a demonstration program. A company created to sell serious scientific instruments had, more or less by accident, produced the first widely shared video game. DEC’s true turning point came on March 22, 1965, with a machine called the PDP-8.
It sold for eighteen thousand dollars, a fraction of the cost of an IBM mainframe, and it is generally considered the first commercially successful minicomputer. More than fifty thousand units were eventually sold. Its lead designer was a young engineer named Edson de Castro, and a year after its release DEC introduced a smaller version, the PDP-8S, priced below ten thousand dollars, the first computer to cross that threshold. These machines reached places computers had never gone before: newspaper composing rooms, nuclear research labs, factory floors, university classrooms.
For the first time, a single department, rather than an entire institution, could afford its own computer. But success carried a quiet problem. De Castro wanted to build a 16-bit successor to the PDP-8, which he called the PDPX. Olsen reviewed the proposal and saw no need for it; he could not understand what a 16-bit machine would offer that DEC’s 12-bit line did not already provide.
He killed the project. De Castro left DEC in 1968 at the age of twenty-nine, taking a group of engineers with him, and founded a company called Data General. Their first product, a 16-bit machine called the Nova, sold for less than the PDP-8 and outsold it. DEC sued for trade-secret infringement.
The lawsuit went nowhere. The pattern would repeat itself almost exactly several times in DEC’s history. An engineer saw the next step clearly, proposed it, and was refused by founders comfortable with what already existed. The engineer left and built exactly what DEC had refused to build.
The company that saved DEC from its first mistake was, in a sense, DEC itself, through an engineer named Gordon Bell. Bell had joined in 1960 as employee number eighty, straight from MIT. He designed some of DEC’s early architectures, then left in 1966 to teach at Carnegie Mellon. When he returned in 1972 as vice president of engineering, he found the company falling behind.
Data General’s Nova was taking over the minicomputer market with a 16-bit design, and DEC had no answer. Bell oversaw a crash effort to build the machine that became the PDP-11, with the core design by Harold McFarland, who, along with another engineer, Roger Cady, brought research from Carnegie Mellon. When the design was presented internally, management nearly canceled it. It reached the market in 1970 at about twenty thousand dollars.
Eventually roughly six hundred thousand units were sold, making it one of the most successful computer lines in history. Almost by accident, it also became one of the most important machines in the history of software. Ken Thompson and Dennis Ritchie wrote Unix on a PDP-11. Ritchie created the C programming language on a PDP-11.
A large part of the software infrastructure the modern world runs on originated on a machine DEC’s leadership had nearly refused to build. By 1975 Bell was thinking about what came after the PDP-11. On April 1 of that year, he gathered six engineers on the third floor of Building 12 in the Maynard mill complex to begin work on a 32-bit successor. Bill Strecker was made chief architect.
They called the project VAX, for Virtual Address Extension. The logic beneath it carried a quiet irony. The founding argument of DEC, from the PDP-1 onward, was that small, cheap, accessible machines would one day replace room-sized mainframes. The PDP-8 had proved it.
The PDP-11 had proved it again. When VAX arrived, it proved the argument so completely that IBM lost billions responding to it. But buried in that same argument was a question no one in Maynard was eager to ask aloud: if small machines always eventually won, what would happen when something even smaller than VAX came along? The VAX 11/780 was announced on October 25, 1977, with the first units shipped the following year; Carnegie Mellon University received the first machine.
It was a 32-bit superminicomputer that could compete directly with IBM mainframes at a fraction of the cost. Strecker’s hardware architecture gave it that power. The operating system, called VMS, built by an engineer named Dave Cutler, gave it a reputation for reliability. The numbers that followed were extraordinary.
DEC’s revenue grew from about seven hundred million dollars to over fourteen billion. More than four hundred thousand VAX systems were eventually sold. In 1986 alone, while the rest of the computer industry struggled through a downturn, DEC’s profits rose by about forty percent. At its peak, the company employed over one hundred twenty thousand people in eighty countries and reached a market capitalization of about twenty-four billion dollars.
In Massachusetts, DEC became the second-largest employer after state government. The Maynard mill complex, purchased in 1974 for just over two million dollars, eventually expanded to over one million square feet. Beneath the growth was a work culture that former employees remembered with deep respect for decades. Olsen never moved to a bigger house even after DEC made him wealthy.
He drove a Ford. Engineers had extraordinary freedom: if you had an idea and could convince a few colleagues it would work, you could often build a prototype without any formal proposal. Product divisions competed internally for resources and attention. In good times this structure produced remarkable creativity; in later years it created something like civil war, with salespeople from different divisions recommending conflicting products to the same customers.
The culture had its cruel edges. Harlan Anderson, DEC’s co-founder, who had shared Olsen’s vision of interactive computing at Lincoln Lab, was forced out in 1966 after a disagreement with Olsen. He had been executive vice president. Olsen moved on without comment.
It was a small, harsh moment in a much larger story, but it revealed what kind of leader Olsen was: completely certain of his own vision, and uninterested in being challenged on it, even by the man who had helped build the company. In 1977, the same year VAX was announced, Olsen said something that would follow him for the rest of his life and beyond. At a meeting of the World Future Society, he said, in his words, that there was no reason for any individual to have a computer in his home. Those eleven words became one of the most quoted and ridiculed statements in American business history, cited as proof that even one of the inventors of modern interactive computing had failed to understand the future of the personal computer.
The record complicates that story. Olsen insisted for the rest of his life that he had been talking about a specific and narrow concept then circulating in the mid-1970s: a computer that would control household lights, appliances, and heating systems, as imagined by science fiction. By his own account and that of people who were present, he was not dismissing the idea of a personal computer on a desk. The quote drifted from its context over the following years and became a legend repeated so often that the correct explanation no longer reached people.
Both things were true: the quotation was unfair to him, but the hesitation it pointed to was real. DEC tried to build its own personal computer, the Rainbow 100, in 1982. It was not compatible with the new IBM PC standard, and it failed in the market. By then it hardly mattered, because throughout the 1980s VAX made DEC richer than ever before.
But inside the mill, the next version of the old pattern was already forming. Dave Cutler, the architect of VMS, began work on a next-generation processor called Prism and a new operating system called Mica. Engineers around him argued with growing concern that the future lay with Unix and microprocessors, not with DEC’s own VAX architecture. DEC’s operations committee, protective of VAX and its enormous profits, did not want to hear it.
In 1988, DEC canceled the Prism project. On paper the logic was defensible: protect the existing profitable VAX line rather than fund an expensive new architecture with an uncertain future. Dave Cutler was furious. That same summer, a Microsoft executive named Nathan Myhrvold arranged a meeting between Cutler and Bill Gates.
On October 31, 1988, Cutler left DEC for Microsoft. Most of his team at DEC’s Western Research Facility followed within days. Over the next several months, Cutler’s group wrote the specifications for an entirely new operating system. The original specification is now held by the Smithsonian Institution.
The system that came out of it, Windows NT, was deeply influenced by the Mica operating system Cutler had designed for the canceled Prism project. DEC sued Microsoft. The companies eventually settled, with Microsoft paying royalties and agreeing to port Windows NT to DEC’s own Alpha processor. It was a kind of resolution, but the deeper loss, an entire generation of DEC’s most skilled operating-system engineers leaving to build someone else’s future, was not something any settlement could repair.
From the early 1990s, the ground under DEC shifted in ways that grew increasingly difficult to answer. Personal computers, which DEC had never fully embraced, were doubling in capability at a pace that showed no sign of stopping. Sun Microsystems was cutting into DEC’s workstation business with Unix-based machines. DEC’s own VAX 9000 project, meant as a direct competitor to IBM mainframes, became an expensive and largely unsuccessful endeavor.
In 1991, DEC announced its first full-year loss in company history. In 1992, Ken Olsen was forced to step down as chief executive. He was sixty-six years old. He never returned to the company.
His successor, Robert Palmer, took over a business that needed to shrink. Between 1990 and 1998, DEC’s workforce fell from one hundred twenty thousand to roughly fifty-three thousand, a reduction that cost the company an estimated 4. 8 billion dollars to carry out. The company Olsen had run for decades like a family had become one measured by the number of employees laid off each quarter.
Yet amid that decline came one genuinely remarkable moment. In 1995, a group of researchers at DEC’s Western Research Laboratory in Palo Alto, led by a computer scientist named Louis Monier, built software with a very simple purpose: to demonstrate how quickly DEC’s new Alpha processor could search and index vast amounts of text. What they built was effectively the first full-text searchable index of the entire World Wide Web. They called it AltaVista.
On its first day, December 15, 1995, it handled about three hundred thousand search queries. By 1997, it was handling around eighty million searches per day. By most accounts AltaVista was the first modern search engine, built inside a company that, at the same time, was laying off thousands of workers and had no clear strategy for the innovation it had created. In 1997, two Stanford graduate students, Larry Page and Sergey Brin, reportedly presented their new PageRank system to DEC.
DEC did not act on it. Many companies reject ideas that later prove important, and past mistakes always look clearer than present ones. But it was a documented moment worth considering quietly: the company that had invented the search engine almost by accident had been shown an early version of Google and let it slip away. By the mid-1990s, DEC’s Alpha processor was genuinely the fastest microprocessor in the world.
But DEC could not sustain the enormous capital expense of competing as a semiconductor manufacturer against companies like Intel. In May 1997, DEC sued Intel for patent infringement on the Alpha design. The lawsuit was settled, and as part of that settlement, DEC sold its semiconductor manufacturing plants to Intel. The company that had once built its own chips from scratch was effectively selling the machinery that had been the foundation of its independent existence.
On January 26, 1998, Compaq Computer Corporation announced it would acquire Digital Equipment Corporation for 9. 6 billion dollars. By then DEC’s workforce had fallen from its peak of one hundred twenty thousand to roughly fifty-three thousand. A company once valued at about twenty-four billion dollars sold for less than half that.
Compaq struggled to absorb what it had bought. Four years later, in 2002, Hewlett-Packard acquired Compaq for approximately twenty-five billion dollars. The name, architecture, and remaining products of Digital Equipment Corporation disappeared completely into a company with origins entirely elsewhere. Ken Olsen died on February 6, 2011, at the age of eighty-four, in hospice care in Indianapolis.
His family released no cause of death. Gordon Bell died on May 17, 2024, at the age of eighty-nine. Edson de Castro died on September 6, 2024, at the age of eighty-five. Within just over a decade, the men who had created, argued over, left, and in some cases returned to the minicomputer industry were all gone.
The machines did not entirely disappear. PDP-11s still survive in some industrial and military systems decades after the last unit left the factory, not out of nostalgia but because they remain reliable. Working PDP-8s are being collected and restored, and a long-term software project called SIMH can run DEC’s original operating systems perfectly on modern hardware. OpenVMS, the direct descendant of the operating system Dave Cutler built for VAX, is still commercially operated and runs on modern processors where reliability matters more than novelty.
Cutler’s own path is perhaps the clearest link between DEC’s laboratories and the digital world we use daily. The operating system he built after leaving DEC, Windows NT, became the architectural foundation of every subsequent version of Windows. He later worked on the core operating system of Microsoft’s Azure cloud platform and the hypervisor for the Xbox One. The engineering principles conceived on the third floor of a converted wool mill in Maynard, Massachusetts, still work silently inside much of the world’s critical cloud infrastructure.
DEC’s Palo Alto research lab helped create the Ethernet networking standard alongside Intel and Xerox, which would eventually connect offices around the world. DEC’s VAXcluster technology, introduced in 1984, allowed multiple machines to share storage and automatically take over if one failed, an architecture that anticipated the concepts of cloud computing by nearly two decades. When the company laid off seventy thousand workers in the 1990s, those engineers did not vanish from the industry. Many spread across Route 128 in Massachusetts and into Silicon Valley, taking what they had learned to companies like Cisco and Juniper and building the networking infrastructure that silently underpins the internet today.
The mill in Maynard still stands. Its clock tower, built in 1892, must still be wound by hand once a week, ninety turns of a crank after climbing one hundred twenty steps. The building now has an entirely different name. Where engineers once debated processors and operating systems that would shape the next fifty years of computing, small businesses now rent offices.
On an ordinary afternoon, driving past it, nothing about the building tells you what happened inside: nothing except its size and a strange silence hard to put into words. In 1997, in a research lab DEC had created, a small group of engineers built the closest thing yet to a modern search engine, using a chip built to prove how fast DEC’s technology could go. That same year, two young researchers who would later create Google came in, showed DEC an early version of what the future of search might look like, and left without any response. What it means to have built the future twice, once with minicomputers and once with a search engine, and to have held onto neither, has no simple or clean explanation.
But whenever we talk about companies that could once see everything clearly, it is worth asking how far we ourselves can see, until the next new technology strikes us the same way.
