Part IThe Story
The Power Supply
Two men took apart a computer on a table in an old woolen mill in Maynard, Massachusetts. The computer was one of the first IBM Personal Computers, a beige box that had appeared in 1981 and was about to reorganize an industry. One of the men was Avram Miller. The other was Ken Olsen, founder and president of Digital Equipment Corporation, then the most important computer company in the world not named IBM. The office was plain, as his offices always were. By most accounts he drove an economy car to work, preferred flannel shirts to suits, and kept his desk in a nineteenth-century textile building because the rent had once been cheap and he saw no reason to leave.
The story comes secondhand, told by Miller years later to the columnist Robert X. Cringely, and it has the compressed shape that anecdotes acquire over time. The essentials, though, have never been disputed. Olsen looked inside the IBM machine and found the power supply. It was, in Miller's retelling, puny. Olsen was amazed at how poor it was. A company that would ship a power supply like that, he reasoned, was not a company Digital needed to worry about.
It was an engineer's judgment, and on its own terms it was correct. A robust power supply makes a computer reliable. A weak one fails. Olsen had spent his adolescence in a Connecticut basement stripping parts from discarded radios, and his early career at MIT building machines for the air defense of North America, where an unreliable component could, in theory, let a Soviet bomber through. He knew what good work looked like, and he knew what it didn't.
What he did not see, or refused to see, was that the people who would buy the next hundred million computers would never open the case.
That table in Maynard is where the story of Ken Olsen folds in on itself. The man who had taken computing out of the air-conditioned temple and put it on the laboratory bench, who had sold interactive machines at prices that let a single scientist own one, who had trained the young Bill Gates without ever meeting him, stood over the next democratization of computing and judged it by its wiring. He was looking for craftsmanship. The market was looking for something else entirely: good enough, cheap enough, compatible with everything else. Eleven years later his board would force him out of the company he had founded with $70,000. Six years after that, what remained of Digital would be sold to Compaq, a maker of the kind of machine he had once taken apart and found wanting.
None of that was visible from the table. What was visible was a small, cheaply built box, and a man who had spent his whole life believing that the quality of the thing was the thing.
By the Numbers
Digital Equipment Corporation, 1957–1998
$70,000AR&D's 1957 seed investment, for 70% of the company
$94,000Module sales in 1958, DEC's first year, when it turned a profit
$18,000List price of the PDP-8 on March 22, 1965
600,000PDP-11s sold across all models by 1996
$14BAnnual sales at the late-1980s peak
35 yearsOlsen's tenure as president, 1957–1992
$9.6BCompaq's 1998 purchase price
A Machine You Don't Need
Stratford, Connecticut, in the 1930s was machine-tool country, a place where boys were expected to learn shop practice the way boys elsewhere learned to farm. Kenneth Harry Olsen grew up there in a Norwegian working-class neighborhood, the son of Oswald and Elizabeth Olsen, second-generation Scandinavians whose own parents had come from Norway on his father's side and Sweden on his mother's. Oswald designed machine tools, held several patents, and later sold machines for the Baird Machine Company. He was, according to Olsen's biographers Glenn Rifkin and George Harrar, "a fundamentalist by religion and a disciplinarian by nature," a man who believed in Puritan ethics applied equally to life and work.
He was also known for a strange sales habit. He advised customers not to buy any machine from him that they didn't really need.
Hold on to that sentence. It will come back.
Oswald kept a shop in the basement, and Ken and his younger brother Stan spent long hours down there inventing gadgets and repairing the neighbors' broken radios. The Depression made parts scarce; nobody threw out a radio in those years, and when someone did, Ken stripped it. "Anything I could get hold of I seized and studied and read," he told the Smithsonian's David Allison in a 1988 oral history. The library books on radio, he remembered, stopped at the spark transmitter, around the time of the First World War. Everything after that he had to learn from Popular Mechanics, Popular Science, and the parts themselves. The brothers built crystal sets that could pull in only one or two local stations, then a one-tube radio powered by battery cells bought on sale for a penny apiece, stacked to twelve volts because the normal two hundred volts was beyond their budget. "We built it and rebuilt it and rebuilt it again," he said. "And learned a lot on the way."
There is a better-known story from those years, one that has the flavor of family legend: the brothers rigged a transmitter and cut into a local station's broadcast to play a jingle Stan had written, called "Murphy's Meatballs." It is the only recorded instance of Ken Olsen doing anything for the pure hell of it, and even then he did it with his own equipment.
He played center on his high school football team, a position that suits the temperament: the player who touches the ball first on every down and is never the one to score. He graduated in 1944 and was drafted into the Navy, which spent eleven months giving him what he later called "an excellent education in electronics": radar, sonar, navigation systems. The fighting ended before he finished training. He served on a cruiser, saw Korea and China, and came home in 1946 with a technician's certificate and a claim on the GI Bill.
The GI Bill took him to MIT. He earned a bachelor's degree in electrical engineering in 1950 and a master's in 1952; his thesis was titled "A Magnetic Matrix Switch and its Incorporation into a Coincident-Current Memory," which is the kind of title a person writes when he has no interest in being read by anyone who does not already understand it. That same year of 1950 he married Eeva-Liisa Aulikki, a Finnish woman, completing an almost perfectly Nordic circuit: a Norwegian-Swedish son of Connecticut marrying a daughter of Finland. They would have three children and be married for fifty-nine years.
In Boston he joined the Park Street Church, the evangelical congregation at the corner of Boston Common. Its minister, Harold Ockenga, asked the young engineer to take charge of the Sunday school. Olsen did so, successfully, and discovered in the process that he liked to manage. It was his first organization. He would later become a deacon there and attend a monthly Boston-area prayer breakfast of some forty business executives. Christianity was, for the rest of his life, his primary way of defining himself.
The biographical compression writes itself, almost too neatly: a disciplinarian father who told customers not to buy what they didn't need; a basement where parts were scarce and everything was rebuilt three times; a Navy school that treated electronics as a matter of reliability under fire; a church that taught him he could lead. Every one of these would shape the company he built. Every one of them would also, eventually, help destroy it.
The Line Outside the TX-0
The computers of the early 1950s were buildings. Whirlwind, the machine Jay Forrester's team built at MIT beginning in 1944 (originally for a Navy flight simulator that was never finished), had racks that Olsen remembered as twenty-six inches wide and eleven feet high. It was the first computer to show real-time data on a display, and it evolved into the prototype for SAGE, the Semi-Automatic Ground Environment, the Air Force system that tied radar stations across North America to computers operated by men with light guns aimed at glowing screens.
Olsen came to it as a graduate student and stayed as one of roughly four hundred engineers MIT hired under its government contract at what became Lincoln Laboratory. He was put in charge of a fifteen-person team to design a small computer whose only job was to test the magnetic-core memory Forrester had invented. The Memory Test Computer was finished ahead of schedule. Olsen acquired a reputation as that rarest of things in a research lab: an engineer who could manage.
Years later he summarized what Whirlwind had taught him with a joke that was not entirely a joke.
— Ken Olsen, recorded remarks aired by WBUR, 2011It was a big one, the racks were 26 inches wide and 11 feet high. One of the things I learned from that was a computer should always be higher than a man. It never seems worth the money unless it's high.
Keep that one too. A computer should be taller than a man. Olsen spent his career making them shorter, and never quite made peace with how short they got.
Lincoln sent him to IBM, the prime contractor for building the SAGE computers, to serve as the lab's liaison. He worked out of an office inside IBM and found the place suffocating. The accounts agree on the substance: the bureaucracy, the layers, the people whose job it was to say no. IBM, he concluded, was building enormous, expensive machines through an enormous, expensive process. Robert Everett, his old supervisor at MIT, would write in Olsen's National Academy of Engineering memorial tribute that it was while stationed at IBM that Olsen "saw an opportunity to start a company to design and market computers that were smaller and less expensive than the large mainframes then being produced by IBM."
Back at Lincoln, the lab turned to transistors. Olsen supervised construction of the TX-0, an 18-bit machine that first ran in 1956 to prove that transistorized circuits could replace vacuum tubes, and then the much larger 36-bit TX-2, with a then-staggering 64,000 words of core memory. Core was so expensive that the lab stripped memory out of the TX-0 to feed the TX-2, and gave the hollowed-out remains of the smaller machine to MIT on permanent loan.
Then something odd happened, and Olsen noticed it.
Students lined up for hours to use the stripped-down TX-0. Nearby sat a faster IBM machine that was largely ignored. The difference was not speed. The TX-0 let you sit at it, type, watch it respond, change your mind, try again. The IBM machine made you hand over a deck of punched cards and come back later. One was a conversation. The other was a correspondence course.
Harlan Anderson, who watched the same lines, put it in the plainest terms: "We had a whole generation of MIT engineering graduates in the electrical engineering department who cut their teeth on our computer 'cause they couldn't get near the big IBM computers that were at MIT at the time."
Anderson was twenty-two when he arrived at Lincoln in 1952, and Olsen was his first boss. He had grown up in Freeport, Illinois, the kind of high school leader who married his junior-high sweetheart, Lois Kahl, the valedictorian of their class, in September 1950. At the University of Illinois he had taken programming courses on the ILLIAC I while it was still under construction, taught by the Cambridge computer scientist David Wheeler. He drove to Massachusetts in a very old car crammed with their belongings and Lois's younger brother; hours before he was due to report for the Korean War draft, word came from the head of the Selective Service postponing it indefinitely, because MIT needed him. He was a physicist, not an electrical engineer, a Midwesterner among New Englanders, a man who would later title his memoir Learn, Earn & Return and mean all three verbs equally.
The two men drew the same conclusion from the line outside the TX-0. People did not want the fastest computer. They wanted a computer they could touch. There was a market, they thought, for a small machine dedicated to interactive work, essentially a commercialized TX-0, cheap enough to be devoted to a single task in a single lab.
— Ken Olsen, quoted by the Computer History Museum, February 2011We had a dream of interactive computing. Normal computing was considered big, expensive, awesome, beyond ordinary people. Interactive computing was exciting and fun, and people could interact directly with the computer.
It is worth pausing on the phrase "beyond ordinary people." Olsen's instinct, formed in a Depression basement and a Navy radio shack, was populist in the specific way of a craftsman: he believed that tools should be in the hands of the people doing the work, not locked behind glass and administered by a priesthood. He was, in 1957, making exactly the argument that Steve Jobs would make twenty years later. He simply drew the circle of "ordinary people" to include engineers and scientists and stopped there.
Seventy Thousand Dollars and a Word You Could Not Say
In 1957 there was no Sand Hill Road. There was barely such a thing as venture capital. There was General Georges Doriot, a French-born Harvard Business School professor and Army quartermaster general, and his American Research and Development Corporation, a publicly traded fund that had been trying since 1946 to prove that you could make money backing technical companies before they had proven anything.
Olsen, Anderson, and Ken's brother Stan went looking for money and found the American business community actively hostile to computer companies. The logic was not stupid. Small computer firms had come and gone throughout the 1950s, each wiped out when some new technical development made its platform obsolete. Even RCA and General Electric, with all their resources, were losing money in the market. Why would anyone back three engineers from a government lab?
Doriot would, on conditions. He was worried that a company calling itself a computer company would never get follow-on financing. So he asked them to change their business plan to de-emphasize computers, and to change their name. The proposal they had brought him was for the "Digital Computer Corporation." The one they returned with, dated May 27, 1957, described a company that would begin by selling logic modules, circuit boards engineers could wire together to build test equipment, and only then, if that business sustained itself, move to building a complete computer in "Phase II."
The Digital Computer Corporation became the Digital Equipment Corporation. AR&D put in $70,000 for 70 percent. Dorothy Rowe, AR&D's senior vice president, joined the board and served as treasurer.
There is a second version of Doriot's reasoning, one Olsen told years later to Howard Anderson (no relation to Harlan), then a Harvard Business School lecturer. Doriot, in this telling, worried that DEC's directors were too old, and that it would take the company too long to make it in computers. A shorter horizon, a module business with revenue in year one, would make the investment look less risky and let the directors see results before they died. It is a very funny story and entirely in character for both men: the Frenchman's dry mortality and the Norwegian's willingness to repeat it without apparent embarrassment.
Doriot extended the disguise to his own shareholders. AR&D's 1957 annual report said only that "DEC manufactures and sells transistorized digital 'building block' units, improved versions of research, development, and testing devices." The word computer did not appear.
The company set up in a corner of a Civil War–era woolen mill in Maynard, twenty-odd miles west of Boston, where inexpensive manufacturing space was plentiful. Three employees; 8,500 square feet. Anderson was employee number two. The Assabet Woolen Mill had been built to make cloth for Union uniforms; now it would make circuit boards in extruded aluminum housings, connected with banana-plug patch cords, sold to other computer companies who needed to test their own machines. The Digital Laboratory Modules shipped in early 1958. The recession of the late 1950s did nothing to slow them. DEC sold $94,000 worth that first year and ended it in the black.
A company that is profitable in year one with a product it did not want to make is a company that has learned a lesson about patience. Digital would sell modules into the 1970s. It published extensive data about them in free catalogs that became enormously popular among engineers, a kind of textbook of practical digital design disguised as marketing.
By August 1959, the module business was stable enough for Phase II. Ben Gurley, a Lincoln Lab veteran, began designing the company's first computer. In keeping with Doriot's instructions, it would not be called a computer. It would be called a Programmed Data Processor: the PDP-1. "We aren't building computers," Gurley said, "we're building 'Programmable Data Processors.'"
The prototype appeared publicly at the Eastern Joint Computer Conference in Boston in December 1959. The first PDP-1 went to Bolt, Beranek and Newman in November 1960 and was formally accepted the following April. It sold for $120,000 in basic form, with 4,096 18-bit words of core memory and a speed of 100,000 operations per second, plus a Type 30 vector display and a modified IBM typewriter that was notoriously unreliable. Fifty-three were built before production ended in 1969. One of them, at MIT, became the machine on which Steve Russell and his friends wrote Spacewar!, which is to say that the first computer Digital ever sold helped invent the video game, an application Olsen would spend the next thirty years dismissing.
The disguise worked. It worked so well that the habit of disguise, of not calling a thing what it was, of building something important while describing it modestly, became part of Digital's character. Olsen would later say, in his quiet way, that he was in the business for the long run, not to build something quickly and sell to a bigger company, which was fashionable even then. The company he built on a word it could not say would eventually outgrow almost every company that could.
How Short a Computer Can Be
If the PDP-1 proved DEC could build a computer, the PDP-8 proved it could invent a category.
The path ran through Lincoln Lab again. In 1962 the lab used DEC's own System Building Blocks to build the LINC, a small 12-bit machine loaded with analog-to-digital converters for laboratory instruments. Scientists loved it. It has been called the first real minicomputer. DEC took the logic design, stripped out the expensive A-to-D hardware, and in 1963 shipped the PDP-5, introduced at WESCON on August 11 with an advertising line that was really a manifesto: "Now you can own the PDP-5 computer for what a core memory alone used to cost: $27,000."
Own. Not rent, not lease, not submit jobs to. Own.
Then, on March 22, 1965, came the PDP-8. It replaced the PDP-5's modules with DEC's new Flip-Chip boards and packed the whole machine into a tabletop case, the CPU visible through smoked plastic so you could watch the logic modules plugged into the wire-wrapped backplane. It came with 4,096 words of memory and a Teletype Model 33 for input and output. It listed for $18,000.
The sub-$25,000 price is what historians usually cite when they call the PDP-8 the first real minicomputer. Several competitors had just entered the market aimed squarely at the PDP-5; the PDP-8 trounced them and gave DEC two years of unrestricted leadership. Some 1,450 of the original "Straight 8" machines were built before newer implementations replaced it. The PDP-8/S, with a slow serial arithmetic unit, dropped the price below $10,000. By 1968, the PDP-9, a successor in the 18-bit line, sold for $19,900 and outsold all its predecessors combined, 445 machines.
Before DEC, as the Harvard lecturer Howard Anderson put it, computing was scarce and precious. After DEC, computing costs dropped like a rock. The minis were bought not by the IT department but by the factory manager or the lab director, who could justify each one by the specific job it did. They went into paper mills, subway systems, oil refineries, physics labs. They became components inside other companies' products. That last detail matters: Digital was not just selling computers, it was selling computing as a raw material.
Revenue told the story in a way that Olsen, who distrusted flair, could appreciate. In 1962, net sales were $6,535,502. In 1967, $38,895,782. The company went public in 1966. By 1972 it had 7,800 employees and $166 million in sales.
Not everything worked. In 1963 DEC also tried to go the other direction, up-market into mainframes, with the 36-bit PDP-6. It was priced around $300,000 and offered customers few obvious advantages over machines from better-established IBM or Honeywell. Somewhere between 23 and 26 were sold. It introduced an early time-sharing monitor that would evolve into TOPS-10, and its redesigned successor, the PDP-10, would later sell about 700 units and become the machine on which a generation of university computer scientists, including the teenage Bill Gates and Paul Allen, learned to program. But in the mid-1960s the PDP-6 was a failure, and failures need owners.
Harlan Anderson left Digital in 1966, after a dispute with Olsen about the management structure of the company. Rifkin and Harrar describe him as effectively squeezed out after the PDP-6 stumbled. The sources are thin on the texture of it, and Anderson himself was gracious about it for the rest of his life. He went to Time Inc. as its director of technology, where he led the magazine company's evaluation of the future of the printed word during the explosion of television; founded a venture firm in 1969; backed more than twenty small technology companies with his own money; endowed centers at Rensselaer and Illinois; served as a trustee of the Boston Symphony Orchestra; took up a late interest in the codebreakers of Bletchley Park. He died on January 30, 2019, at eighty-nine, eight years after Olsen.
The departure established a pattern. Ken Olsen would be Digital, and Digital would be Ken Olsen, and people who disagreed with him about how the company should be structured would, sooner or later, find themselves elsewhere. Some of them would find themselves across the street.
The Engineers Who Walked Out the Door
Edson de Castro was one of Digital's best engineers, and in the late 1960s he was leading a study of a machine called the PDP-X.
The industry was shifting under everyone's feet. Earlier machines had used word lengths built on 6-bit characters: 12 bits, 18, 36. The coming standard, ASCII, favored 8-bit units, which meant 16-bit machines. De Castro's group wanted DEC to build one. Olsen did not support it. He could not see what it offered that the existing 12-bit and 18-bit lines didn't. It is the same sentence his father would have said to a customer: you don't need this machine.
So de Castro and the leaders of the PDP-X effort left and founded Data General. Its 16-bit Nova shipped in 1969 and was a huge success. Tracy Kidder would later make the rivalry between the two companies the backdrop of The Soul of a New Machine, his Pulitzer Prize–winning account of a Data General team racing to build a 32-bit computer, a book that became something like scripture in Route 128 engineering circles and a quiet indictment of the company whose indifference had created its subject.
The Nova's success jolted DEC into a crash program. The machine that resulted, designed by Harold McFarland, Gordon Bell, Roger Cady, and others, nearly died too: when management first saw the proposal, drawn from one of McFarland's simpler designs from his research at Carnegie Mellon, it was not impressed and almost cancelled it. It survived. It was the PDP-11, released in 1970, and it became arguably the most influential minicomputer ever built.
Its central innovation was the Unibus, which treated every peripheral as an address in memory. Adding a new device meant plugging a board into the backplane and writing software that read and wrote to the right addresses. The ease of it spawned an enormous third-party market of add-ons. Ken Thompson and Dennis Ritchie at Bell Labs moved their new operating system, Unix, onto it; the C programming language absorbed some of its idioms. The PDP-11's operating systems, especially RT-11, became the models on which the first microcomputer operating systems were built, by programmers who had cut their teeth on the machine; CP/M borrowed its command syntax, and DEC's use of "/" for command-line switches is why MS-DOS and Windows ended up using "" for file paths. By 1996, when the last single-chip PDP-11s were produced, around 600,000 had been sold. AT&T became Digital's largest customer.
Gordon Bell, who oversaw so much of this, is the other essential character in the Digital story and the man whom Rifkin and Harrar called the Frank Lloyd Wright of the computer industry. He joined DEC in 1960, left, returned, and served as vice president of engineering from 1972 to 1983; he would later move to Microsoft Research. Where Olsen was the craftsman of the organization, Bell was its architect, the man who could hold an entire product line in his head and see where it should go next. The PDP-11 was his vindication, and the VAX, Digital's 32-bit extension of the PDP-11 announced in October 1977, was his masterpiece.
The pattern of the PDP-X (Olsen resists, engineers leave, a competitor wins, DEC scrambles, DEC triumphs anyway) would repeat in a dozen smaller ways. Olsen's company became, almost despite him, one of the great training grounds of the industry. The list of people who passed through it reads like an index to modern computing: Dave Cutler, who led the VMS operating system and then took his team to Microsoft to build Windows NT; Leonard Bosack, co-founder of Cisco; Radia Perlman, author of the Spanning Tree Protocol; Butler Lampson and Leslie Lamport, both Turing Award winners; Jeff Dean and Sanjay Ghemawat, who went on to build much of Google's infrastructure; Jim Keller, who worked on Alpha and StrongARM. Bill Gates, writing to Gordon College in 2006, noted that Olsen's influence was "still important at Microsoft through all the engineers who trained at Digital and have come here to make great software products."
That is a compliment. It is also a list of things Digital did not keep.
Do the Right Thing
On weekends, the mill in Maynard was loud.
Glenn Rifkin, who covered the company for Computerworld before writing its biography, remembered it as "a buzzing factory over the weekend," full of engineers who had come in because they wanted to. Olsen believed that you could find fulfillment in your work, and that the job of a company was to create an atmosphere in which people would thrive, would want to be there long hours. Digital offered day care and other benefits considered progressive at the time. For most of its first thirty-five years it boasted that it had never had a general layoff. At its height, Rifkin estimated, the company accounted for something like one in every five jobs in Massachusetts; in 1987 it was the largest corporate employer in both Massachusetts and New Hampshire, its buildings strung out along Route 2 and Route 128, the billboard for the "Massachusetts Miracle."
The organizing principle of this enterprise, repeated by nearly everyone who worked there, was three words long. Do the right thing. The Digital Equipment Philosophy, a document Olsen wrote with others, put it slightly more fully: treat customers, vendors, and employees with great respect, because they are the people who make the company's life possible; and if you are unsure what to do in a difficult situation, do what you think is the right thing. Win Hindle, who joined Olsen's management team in 1962 and rose to senior vice president of corporate operations, said flatly that "a lot of his deep religious values came out in that." Rifkin agreed that the motto came directly from Ken's faith, and noted the family resemblance to Google's later "Don't be evil."
The structure that carried this philosophy was as unusual as the philosophy itself. Olsen was frustrated, Hindle said, by "too many rules, too many people who could say no to something." He was frustrated by layers of management. "He did not like hierarchy." He communicated directly with whomever he needed to. What emerged was a matrix: engineers had both a line boss and a functional boss, organized horizontally by project and vertically by expertise. It gave people two bosses and, in theory, the freedom to route around both. Digital's version of matrix management spread through the industry and eventually well beyond it.
Then there were the salesmen, who were paid salaries, not commissions.
This was heresy in the 1960s and would still be heresy in most technology companies today. But it follows directly from the basement in Stratford. Oswald Olsen had told customers not to buy machines they didn't need. His son, running a company tens of thousands strong, insisted that Digital's sales force not sell customers products they didn't really need, and structured their pay so they had no incentive to. The ethic of a machine salesman in Depression-era Connecticut had become corporate policy at the second-largest computer company on earth.
It is easy to romanticize all this, and former employees do. At the June 17, 2006, groundbreaking for the Ken Olsen Science Center at Gordon College, roughly a thousand former "Digits" came to Wenham, and one of them, Debra Svenson, wrote afterward of the company that "honed our ethics, shaped our management styles and helped make us resilient," calling Olsen "the father and founder of Digital Equipment Corporation...the head of our DEC family." The word family recurs in these remembrances with a regularity that is almost liturgical.
Rifkin, who knew Olsen from outside the family, offered a less sentimental portrait. "Ken was an enigma," he said after Olsen's death. "He could be a very kind man. The employees just loved him as a paternalistic figure. But he could be a bear of a guy, too. He could be angry, he could berate somebody in a public meeting and leave them just shredded on the floor. He had no qualms about that."
Both things were true, and they were the same thing. The father who teaches you everything in his basement is also the father who decides what you may build. A company organized around one man's conscience is only as adaptable as that conscience. The matrix that gave engineers freedom also, as Digital grew, bred what the MIT organizational psychologist Edgar Schein would later describe as an environment that "felt more and more chaotic and out of control," a culture where any project could find a sponsor and no project could be killed, where consensus was required and therefore never quite arrived.
Olsen told a parable about this, in a 1987 interview with the Middlesex News, though he meant it to cut the other way. Two explorers raced to the South Pole. The Norwegian, Roald Amundsen, was sullen, not given to talking about himself. "He always worried about the details." The Englishman, Robert Scott, chatted endlessly about his plans. "He was show. Flair. Parties. Announcements. Bragging." Amundsen reached the pole first and came home. Scott reached it a month later and never made it back.
Olsen was, of course, the Norwegian. It is a lovely story and a revealing one, and it has one flaw as a guide to the computer business in 1987: Amundsen's expedition succeeded because it went light and took the shortest route. Digital, by then, was carrying everything.
The Sentence That Followed Him
In 1977, the year Digital's sales first exceeded $1 billion and its headcount reached 36,000, the year it announced the VAX-11/780, Ken Olsen gave a talk to the World Future Society in Boston. He said, according to the report that would follow him for the rest of his life: "There is no reason for any individual to have a computer in his home."
Olsen later admitted saying it. He insisted he had been misunderstood, that he had been talking about computers set up to control houses, a common futurist fantasy of the day, not about personal computers as they came to be. His own explanation, preserved in the files of Snopes, is characteristically literal and faintly exasperated: "A long time ago when the common knowledge was that PCs would run our lives in every detail, I said that if you stole something from the refrigerator at night you didn't want to enter this into the computer so that it would ... mess up the computer plans for coming meals."
Snopes rates the popular version a legend and offers a dry verdict: "the out-of-context misinterpretation of Olsen's comments is considered much more amusing and entertaining than what he really meant, so that is the version that has been promulgated for decades now."
Fine. The quote was taken out of context. The trouble is that the context does not help as much as Olsen's defenders would like.
In 1974, before the MITS Altair had debuted and set off the hobbyist boom, a Digital research group demonstrated two prototype microcomputers. Olsen chose not to proceed. In 1977, the year of the famous talk, the company rejected another personal computer proposal. At the beginning of the 1980s, Digital built the VT180, codenamed "Robin," a VT100 terminal with a Z80 microcomputer inside running CP/M, and made it available only to its own employees. The home-automation explanation may be accurate about the sentence. It is not accurate about the company.
There was a second remark, less famous, that nobody has seriously tried to recontextualize: "The personal computer will fall flat on its face in business." And there was the power supply on the table.
When IBM shipped its PC in 1981, Digital responded in 1982 not with one machine but with three, each incompatible with the others and with IBM, each tied to a different proprietary architecture. The Professional was built around a PDP-11 processor running a menu-driven operating system called P/OS; it outperformed the IBM PC and cost more, and every copy of every program had to be keyed to a specific machine. The emerging third-party software industry ignored it. Inside Digital, better software for it was not a priority, perhaps for fear of cannibalizing the PDP-11 line. The result, in the verdict of one contemporary user report, was "a superior machine, running inferior software." The DECmate II was a word processor descended from the PDP-8, not really suited for general computing. The Rainbow 100, the most popular of the three, had both a Z80 and an Intel 8088 and could run CP/M, CP/M-86, and eventually MS-DOS, but it was hardware-incompatible with the IBM PC, expensive, and thinly marketed.
Even the floppy disks were a statement. Digital's RX50 drive was mechanically almost identical to competitors' 5¼-inch drives, but Digital used a proprietary format, held 400 kilobytes rather than the industry-standard 360, and required buyers to purchase specially formatted media. Digital copyrighted the format and demanded license fees from anyone who sold compatible disks. Hobbyists eventually reverse-engineered it. By then it did not matter.
By late 1983, IBM was outselling Digital's personal computers by more than ten to one. That same year, according to the jacket copy of the Rifkin and Harrar biography, Wall Street called for Olsen's ouster. He survived. It is worth noticing that he survived; most founders would not have. The VAX was selling too well for anyone to argue with him.
Clayton Christensen, the Harvard professor whose theory of disruptive innovation was partly inspired by watching the minicomputer companies fail, would later reduce the whole episode to a single structural sentence. "Digital Equipment Corp. had microprocessor technology," he told MIT Sloan Management Review in 2009, "but its business model could not profitably sell a computer for less than $50,000." IBM, with the same processors, set up a separate operation in Florida that could make money at $2,000 and 20 percent gross margins. "The technology trapped in a high-cost business model had no impact on the world, and in fact, the world ultimately killed Digital."
Christensen's analysis is correct and incomplete. The business model was not an accident; it was the embodiment of a set of beliefs. Salaried salesmen who don't push product. Engineers free to build what they think is best. Machines built so robustly that customers keep them for decades; some PDP-11s more than twenty-five years old were still controlling factories, transit systems, and nuclear plants well into the twenty-first century. A refusal to sell people machines they don't need. These were the virtues that built Digital. Each of them, applied to a $2,000 beige box sold at a department store to people who would never open the case, became a liability.
The man who had once pointed out, with evident pride, that when the Soviet Union stole computer designs it stole DEC's rather than IBM's could not imagine being the one who was copied, cheapened, and beaten.
The Summit and the Cliff
For a while, it did not look like a mistake at all.
The VAX, with its 32-bit architecture, virtual memory, and backward compatibility with the PDP-11, propelled Digital into leadership almost immediately after its 1978 release. It was so successful that in 1983 the company cancelled its Jupiter project, the planned successor to the PDP-10 mainframe line, and bet everything on a single architecture. In 1984 Digital launched its first 10-megabit Ethernet products, the fruit of its collaboration with Xerox and Intel on the DIX standard; it was Digital, more than anyone, that made Ethernet commercially successful. Combined with DECnet and VAXcluster technology, which let customers tie multiple VAXes together into a single system and grow by adding machines rather than replacing them, it gave Digital a networked architecture that could attack IBM directly. In September 1985, dec.com became the fifth .com domain ever registered.
In 1986, while the rest of the computer industry was in a downturn, Digital's profits rose 38 percent. It took an estimated $2 billion in business away from IBM in the mid-1980s and by 1987 was threatening IBM's position as number one. In the mid-range market it had twice IBM's sales.
On October 27, 1986, Fortune put Ken Olsen on its cover. "Though he has not yet become a household name," Peter Petre wrote, "Kenneth Harry Olsen is arguably the most successful entrepreneur in the history of American business." In twenty-nine years he had taken Digital "from nothing to $7.6 billion in annual revenues." The magazine reached for the largest comparisons in the American industrial canon: Digital was bigger, even adjusting for inflation, than Ford Motor Co. when death claimed Henry Ford, than U.S. Steel when Andrew Carnegie sold out, than Standard Oil when John D. Rockefeller stepped aside. Olsen was sixty, "of the lucky generation of entrepreneurs who experienced hard knocks as kids during the Depression and benefited as businessmen from the surge in prosperity that followed World War II."
Two years later came the biography, The Ultimate Entrepreneur. Its title would age about as well as the Fortune cover. Sales peaked in the late 1980s at around $14 billion; headcount reached somewhere between 120,000 and 140,000, depending on who was counting, in operations across more than ninety-five countries.
The summit, in retrospect, was also the edge.
Olsen had begun describing Unix in public appearances, starting in 1987, as "snake oil." He approved an internal BSD-based Unix for the VAX called Ultrix but, by most accounts, it never got enthusiastic support. Meanwhile the Berkeley RISC and Stanford MIPS designs were promising 32-bit processors that would outperform the fastest VAXes, and Sun Microsystems and Silicon Graphics were selling Unix workstations to exactly the engineers and scientists who had once been Digital's natural customers. Digital recognized that RISC offered at least twice as many MIPS per dollar as VAX. Its response was not one strategy but several, all funded at once.
One faction wanted to push the VAX's aging CISC design as far as it would go, protecting the high-margin top of the market. That became the VAX 9000, introduced in October 1989, two years late, plagued with problems, and priced so high that it never succeeded. Another faction believed Digital needed its own RISC architecture; no fewer than four separate small projects ran in parallel at labs around the country before being consolidated into PRISM, which delivered a credible design in April 1988 and was cancelled shortly afterward. A third group, in Palo Alto, fed up with both, started a skunkworks and simply bought MIPS processors off the shelf; the DECstation 3100 appeared on January 11, 1989. Meanwhile the company spun up software efforts in every hot niche, sometimes several for the same niche, including a database, Rdb, that competed with Oracle, which only a few years earlier had been a major partner. Hundreds of millions of dollars went into DEC-centric products that customers largely ignored, a problem made worse, as Digital's own later historians acknowledged, by Olsen's aversion to traditional advertising and his belief that well-engineered products would sell themselves.
The matrix could fund everything. It could not choose.
By 1991, Digital was still unable to sell PCs to its own customers, who bought them from Compaq and HP to hook up to their Digital minicomputers. The first layoffs came, ending the company's proudest boast. After two years of operating losses, in July 1992, the board forced Ken Olsen to resign.
His successor, Robert Palmer, had joined Digital in 1985 to run semiconductor engineering and manufacturing and had campaigned relentlessly for the job; his success with the Alpha, a 64-bit RISC processor that would launch that year as one of the fastest chips in the world, made him the obvious candidate. The board gave Palmer a title that had never been used in Digital's thirty-five-year existence: chief executive officer. Olsen had always been simply president. A new, more modern logo was commissioned.
On July 20, 1992, Olsen addressed Digital's engineers. The recording survives in the company archives at the Computer History Museum, on Betacam SP: tape 1 of 3, running one hour, twenty-four minutes, and thirty-eight seconds. That he needed more than four hours of tape to say goodbye to his engineers is, perhaps, all one needs to know.
The man passed over in the succession was John F. Smith, who had joined Digital in 1958 as its twelfth employee, turning down a job at Bell Laboratories in New Jersey to work in a woolen mill in Maynard. He had risen to senior vice president in 1987 and chief operating officer in 1991, filling in at every trouble spot Olsen pointed him toward. He stayed on for a time after Palmer took over, to help. He retired on January 5, 1993, after thirty-five years. His position was not filled.
Digital reported $2.8 billion in losses for fiscal 1992. On April 15, 1994, it announced a quarterly loss of $183 million, three to four times what Wall Street expected, and its stock fell 20 percent in a day. Palmer sold the pieces: the database to Oracle, the disk business to Quantum, the terminals that had made the VT100 an industry standard to Boundless Technologies, the printers, the networking group, much of the chipmaking operation to Intel as part of a patent settlement. On January 26, 1998, Compaq, a company founded in 1982 to build IBM-compatible PCs, announced it would buy what remained for $9.6 billion. Digital had 53,500 employees, still 65 percent more than Compaq, producing about half the revenue. Compaq's first priority was to cut Digital's selling, general, and administrative costs, which ran to 24 percent of revenue, twice Compaq's own ratio.
The Boston Globe's headline was "The war lost, Digital surrenders."
Whirlwind in the Scrapyard
Around 1972, when Digital was still climbing, Ken Olsen and his old MIT supervisor Bob Everett began to worry about Whirlwind.
The great machine, with its eleven-foot racks, was about to be declared surplus. So were the remains of the TX-0, the stripped-down little computer the students had lined up for. Olsen and Everett rescued Whirlwind and parts of the TX-0 from the scrapper. They became part of a 1975 Digital exhibit in Maynard on the history of computer logic. Four years later, with Olsen's backing, the collection became a public museum in Digital's MR-2 building in Marlboro, Massachusetts, directed by Gwen Bell, Gordon Bell's wife. On September 24, 1979, the British pioneer Maurice Wilkes opened the Digital Computer Museum with a talk on programming Cambridge's EDSAC. In 1984 it moved to Boston's Museum Wharf as The Computer Museum, with Intel co-founder Robert Noyce giving the inaugural lecture. In 1996 it moved west, to be near the new center of the industry, and became the Computer History Museum in Mountain View.
In 2004, Compaq's successor sent Digital's own corporate archives there: 1,239 linear feet, 611 record cartons, tens of thousands of photographs, including the papers of Olsen and Harlan Anderson and Olsen's master's thesis on the magnetic matrix switch. The company Olsen founded to make computers smaller ended up as a collection in the museum Olsen founded to keep the big ones from being thrown away.
After Digital, he became chairman of a startup called Advanced Modular Solutions. Its heart, by the account of the journalist Udayan Gupta, who met him in 2002, was never quite in it.
On January 4, 1997, at seventy, Olsen sat for an oral history at the Colonial Inn in Concord, Massachusetts. He was unrepentant, and in places he was right. He had recently told an audience at MIT's Sloan School, he said, that "we've lost something and we've forgotten the scientific method," and a professor had agreed with him afterward: We don't have time for it any more. The computer industry, he complained, now offered the same answer to every question: faster computers on every desk, faster networks, more software, one enormous database that everyone could access. "When you ask them what's the problem, they say we have the answer, we don't have to know the problem." Did the people taking catalogue orders, writing letters, working in the library, all need access to all the data of the corporation?
Then he told a story about himself.
"Now I ran a fairly large company and people were shocked that I didn't have a computer in my office. I use a computer all the time. I tell my secretary what to do with it. When I want to find something that would be very useful, I call my son and say 'I paid for your Ph.D., find it for me.'"
It is a joke, and it is an entire worldview. The computer is a tool for solving a defined problem, operated by the person whose job it is to operate it. You do not need one on your desk any more than you need a lathe on your desk. You do not buy the machine you don't need.
His wife, Eeva-Liisa, died in 2009. He had been in failing health for more than a decade. He moved to Indiana to live with a daughter and died in hospice care in Indianapolis on February 6, 2011, two weeks short of his eighty-fifth birthday. His family gave no details. Gordon College, the small Christian college in Wenham where he had been a trustee since 1961, announced the death.
The tributes came from the people he had made possible. Gordon Bell: "In the heady days of Bill Gates and Steve Jobs, it's too easy to forget that it was Ken Olsen's vision of interactivity that took computing away from the centralized mainframe and into the hands of the people." Dan Bricklin, co-creator of VisiCalc and a former Digital employee: Olsen "is in the elite club of tech founders w/Gates & Jobs, and set the stage for them." Gates himself had called Olsen "a major influence in my life." There is also a line attributed to Gates, perhaps apocryphal, that the Hall of Fame essayist William Hermann placed at the head of his account of Digital's rise and fall: "When I finally met Ken Olsen, he was no longer Ken Olsen."
The Ken Olsen Science Center at Gordon College, which broke ground on June 17, 2006, was the first building Olsen ever allowed to carry his name. The town of Maynard declared that day Ken Olsen Day. The building was dedicated on September 27, 2008. Its lobby holds a Digital Loggia of Technology, documenting the company's machines and history, and, at the end of it, a small interactive kiosk.
It is not taller than a man. Former employees walk up to it and type in their stories.
Part IIThe Playbook
What follows is an attempt to extract usable principles from a career that is equally instructive in its triumphs and its failures, often in the same decision. Olsen is a rare case: a founder whose operating philosophy was coherent, consistent, and explicitly stated for thirty-five years, which makes it possible to watch the same beliefs produce a company worth $14 billion a year and then dismantle it. Most of the principles below are double-edged. The tactics are written to keep the edge that cut in his favor.
The Olsen Playbook
- 1.Follow the line, not the benchmark.
- 2.Disguise the ambition until the business can carry it.
- 3.Use price to draw the borders of a new category.
- 4.Let the customer finish the product.
- 5.Govern with a sentence, not a manual.
- 6.Pay people to tell the truth.
- 7.Decentralize authority, but centralize the kill switch.
- 8.Your next competitor is sitting in your design review.
- 9.Distrust your own contempt for the cheap.
- 10.A proprietary moat eventually surrounds you.
- 11.Schedule your own obsolescence.
- 12.Build institutions that outlive the company.
Principle 1
Follow the line, not the benchmark
The founding insight of Digital was an observation about behavior that contradicted the obvious metric. The IBM machine at MIT was faster. The TX-0 had been cannibalized for parts. By every specification a procurement officer would care about, the IBM machine won. And the students lined up for hours to use the TX-0 anyway, because it let them interact with it directly.
Olsen and Anderson did not argue with the benchmark; they ignored it. They inferred that the unit of value was not operations per second but the feedback loop between a person and a machine, and they built a company around serving users "where the graphical output or real-time operation would be more important than outright performance." That inference, not any particular circuit, was the company's first asset.
The irony, fully visible only in retrospect, is that the same insight should have told Olsen everything he needed to know about the personal computer. People lined up for Apple IIs the way they had lined up for the TX-0. By then he had become the man who measured the power supply.
Tactic: When usage and specifications disagree, trust usage, and write down exactly what people are queuing for before you decide what to build.
Principle 2
Disguise the ambition until the business can carry it
Georges Doriot forced Digital to become something it did not want to be: a module company, named for "equipment," selling test gear to other computer makers. The founders complied, and the disguise turned out to be a gift. The modules turned a profit in the first year, $94,000 of sales in a recession, which meant Digital funded its move into computers from its own revenue rather than from investors who had already watched RCA and GE lose money in the market.
The habit persisted into naming. The PDP-1 was a "Programmed Data Processor" precisely so it wouldn't be called a computer. Ben Gurley's line, "We aren't building computers," was strategic modesty: it kept Digital off competitors' radar and out of the category that scared financiers.
The underlying principle is sequencing. The ambition (interactive computing for ordinary people) never changed. What changed was the order in which the world was told about it, and the order in which the company exposed itself to risk.
Tactic: Define a Phase I product that is profitable on its own and technically prepares Phase II, and don't publicly name the ambition until Phase I is paying for it.
Principle 3
Use price to draw the borders of a new category
The PDP-8 is called the first real minicomputer not because of its architecture but because of its price. At $18,000, it crossed a threshold (the sub-$25,000 mark) below which a computer stopped being a capital project and became a piece of lab equipment, justifiable by a single department head for a single task. The PDP-5's advertising had already made the argument explicit: you could own the whole computer for what core memory alone used to cost.
Every step down the price curve opened a new class of buyer. The PDP-8/S went below $10,000. The PDP-9 at $19,900 outsold all of its 18-bit predecessors combined. The buyers were no longer IT departments but factory managers and lab directors. Price was not a concession Digital made to the market; it was the tool Digital used to create a market that hadn't existed.
$
Price as Category Design
Each price threshold Digital crossed brought a new buyer into computing.
| Machine | Year | Price | What it unlocked |
|---|---|---|---|
| PDP-1 | 1960 | $120,000 | Interactive computing for research institutions |
| PDP-5 | 1963 | $27,000 | "Own the computer for what core memory used to cost" |
| PDP-8 | 1965 | $18,000 | The minicomputer as a category; single-task, single-department purchase |
| PDP-8/S | 1966 | Under $10,000 | Computers as components in other products |
| DEC business model | 1980s | Could not profit below ~$50,000 | Nothing: the threshold Digital could no longer cross |
The last row is the cautionary one. Christensen's observation that Digital's business model could not profitably sell a computer below $50,000 means the company that had used price to create a category eventually priced itself out of the next one.
Tactic: Identify the price below which your product changes who can buy it, and treat reaching that number as a design requirement rather than a cost-cutting exercise.
Principle 4
Let the customer finish the product
Digital's early customers were technically proficient companies and universities, and early DEC software was sparse and often buggy. Rather than hide this, Digital built a relationship in which customers wrote their own software, fixed Digital's, and sent corrections back. DECUS, the user society founded in 1961, became one of the largest single-manufacturer computer societies in the world. The PDP-11's Unibus, which let anyone add a peripheral by plugging a board into the backplane, spawned a huge third-party add-on market that made the machine more valuable to everyone.
Some of the most consequential things ever done on Digital hardware (Unix on the PDP-7 and PDP-11, Spacewar! on the PDP-1, Microsoft's first software on a PDP-10) were not Digital projects at all. As WBUR's Curt Nickisch put it, it was Digital's customers "that came up with new applications and new uses for computers."
The strategic point is that an open, well-documented platform turns customers into an R&D department you don't pay. The free module catalogs, full of practical design information, were an early version of the same instinct.
Tactic: Publish your interfaces, document them generously, and create a formal channel for customers to contribute improvements back to you.
Principle 5
Govern with a sentence, not a manual
Digital's operating culture was organized around three words: Do the right thing. The Digital Equipment Philosophy told employees to treat customers, vendors, and employees with great respect and, when unsure, to do what they thought was right. For a company that grew from three employees to well over 100,000, that is an astonishingly short constitution.
The sentence did real work. It let Olsen push authority down without writing rules, which suited a man frustrated by "too many rules, too many people who could say no." It produced loyalty that outlived the company by decades; a thousand former employees showed up to a groundbreaking in 2006. It became, in the view of several observers, a template for later technology companies' value statements.
Its limit is that a moral sentence tells you how to behave but not what to build. "Do the right thing" could justify a salaried sales force and a no-layoff tradition. It could not adjudicate between the VAX 9000, PRISM, and the DECstation.
Tactic: Reduce your culture to one sentence employees can apply to any situation, and separately maintain an explicit mechanism for strategic choices the sentence cannot make.
Principle 6
Pay people to tell the truth
Oswald Olsen told customers not to buy machines they didn't need. His son made that a structural feature of the company: Digital paid its salespeople salaries rather than commissions, and Olsen insisted they not sell customers products they didn't really need.
The benefits were real. Customers trusted Digital's sales engineers as advisers rather than vendors, which suited a customer base of scientists and engineers who could tell when they were being sold. The sales force had no reason to push the wrong configuration or oversell capacity. Trust compounds; Digital's customers became famous for buying each new generation as soon as it shipped.
The cost is that a salaried, consultative sales force is expensive and slow, and it is badly matched to high-volume, low-margin products. By 1997, Digital's selling, general, and administrative costs ran to 24 percent of revenue, double Compaq's. The honesty was worth paying for; the question is whether you can afford the delivery mechanism in every market.
Tactic: Align compensation so that no one earns more by selling a customer something they don't need, and revisit the structure whenever you enter a market with fundamentally different margins.
Olsen's matrix gave engineers two bosses and the freedom to route around both. It fostered extraordinary innovation (the PDP-11, the VAX, Ethernet commercialization, clustering, AltaVista) and it spread throughout industry as a management technique.
It also meant that nothing could be stopped. When RISC threatened the VAX in the late 1980s, Digital simultaneously funded an upgraded CISC line, four separate RISC projects that became PRISM, a Palo Alto skunkworks using outside MIPS chips, and a sprawl of DEC-only software products in every hot niche. Edgar Schein described the result as a culture that "felt more and more chaotic and out of control." The same infighting that seriously delayed Digital's response came from the same freedom that had made it great.
⚖
The Matrix Under Stress
How Digital's decentralized structure handled the RISC threat, 1985–1992.
| Initiative | Faction's thesis | Outcome |
|---|---|---|
| VAX 9000 | Protect high-margin CISC at the top end | Risk Two years late (Oct 1989), never profitable |
| PRISM | Build Digital's own RISC | Risk Four parallel projects; finalized April 1988, then cancelled |
| DECstation 3100 | Ship Unix workstations now on MIPS chips | Try Some success; later displaced by Alpha |
| Alpha | Leapfrog to 64-bit RISC | Scale Class-leading performance in 1992, too late to save the company |
Tactic: Let teams start projects freely, but reserve for one accountable person the authority to end them, and use it on a fixed calendar.
Principle 8
Your next competitor is sitting in your design review
Edson de Castro did not leave Digital because he wanted to compete with it. He left because Olsen would not let him build the 16-bit PDP-X, reasoning that it offered nothing the existing 12-bit and 18-bit machines didn't. De Castro founded Data General, and the Nova became a huge success that forced Digital into a crash program. The PDP-11 that resulted was a triumph, but Digital had handed a rival a head start and a founding team.
The same dynamic repeated in quieter forms. Dave Cutler took the team that built VMS to Microsoft and built Windows NT. Digital's alumni went on to found or build Cisco, much of Google's infrastructure, and more. Some of this is the unavoidable cost of hiring brilliant people. Some of it is the direct consequence of a founder's veto.
The lesson is not that every proposal should be approved. It is that a rejected proposal from a strong team is not a closed question; it is a question you have transferred to the market.
Tactic: When you kill a proposal from a strong internal team, ask explicitly what happens if they build it elsewhere, and if the answer frightens you, find a way to let them build it inside.
Principle 9
Distrust your own contempt for the cheap
Olsen and Avram Miller took apart an early IBM PC and Olsen concluded, from its puny power supply, that Digital had nothing to worry about. It was a correct engineering judgment and a catastrophic business one. The buyers of personal computers did not evaluate power supplies. They evaluated price, software, and compatibility.
Every successful incumbent develops standards of quality that were, once, the reason it won. Olsen's standards came from Navy radar and air-defense computers, where reliability was everything. When a cheaper, cruder product appears, those standards generate a feeling of contempt, and the contempt feels like expertise. It is exactly the feeling IBM's engineers presumably had about the PDP-8.
↺
The Same Move, Reversed
What Digital did to the mainframe, the PC did to Digital.
| Dimension | DEC vs. IBM mainframes (1960s) | PC vs. DEC (1980s) |
|---|---|---|
| Performance | Lower than the incumbent | Lower than the incumbent |
| Price | A fraction of the incumbent's | A fraction of the incumbent's |
| New buyer | Lab and factory managers | Individuals and small offices |
| Incumbent's reaction | Initially dismissive | "Will fall flat on its face in business" |
Tactic: When you find yourself dismissing a competitor for poor quality, list who is buying it anyway and what they are getting that you don't offer.
Principle 10
A proprietary moat eventually surrounds you
Digital's proprietary architecture was, for a long time, a source of enormous strength. VAX and VMS locked in customers; the VT100 became a standard; DECnet created one of the most important networks of its era. But when Digital entered the PC market in 1982, it brought the proprietary instinct with it: three mutually incompatible machines, none compatible with IBM; per-machine software keys; a copyrighted floppy format that required licensed media. The third-party software industry simply went elsewhere.
The same instinct governed the Unix question. Olsen called Unix "snake oil" beginning in 1987, while corporate customers were building client-server networks on it. As WBUR's Curt Nickisch put it, Digital "got on the wrong side of a proprietary battle," like a railroad using a different track gauge after the rest of the country standardized.
The irony is that Digital had championed standards when it suited: ASCII, Ethernet, the ANSI terminal standard the VT100 implemented. The failure was not ideological but situational: an inability to recognize when a market had tipped toward a standard Digital didn't own.
Tactic: For every proprietary element of your product, name the date by which you will reassess whether a de facto standard has emerged, and be prepared to adopt it.
Principle 11
Schedule your own obsolescence
Olsen survived Wall Street's call for his ouster in 1983 because the VAX was carrying the company. He survived nine more years. When the board finally forced him out in July 1992, after two years of operating losses, the company that had never had a general layoff was already laying people off, and its succession was chaotic: the loyal operator John F. Smith passed over for Robert Palmer, who had relentlessly campaigned for the job and was given a CEO title Digital had never used.
The American National Business Hall of Fame's essay on Olsen reduces the lesson to a sentence: the top manager with unprecedented success may eventually need to be replaced at exactly the moment he is most reluctant to give up control. The line attributed to Bill Gates, "When I finally met Ken Olsen, he was no longer Ken Olsen," gets at the same thing from the other direction. The founder the industry admired had been replaced by a later version of himself, and no one inside the company had the standing to say so.
⏱
Decision Points in the Olsen Era
Moments where the founder's judgment shaped the company's trajectory.
1957
Accepts Doriot's terms: $70,000 for 70%, name changed from "Digital Computer Corporation."
1965
PDP-8 launched at $18,000, creating the minicomputer category.
1966
Harlan Anderson leaves after a dispute over management structure; DEC goes public.
1968
Declines to back the 16-bit PDP-X; its leaders found Data General.
1974
Declines to pursue two prototype microcomputers.
1977
VAX-11/780 announced; World Future Society remark; another PC proposal rejected.
1982
Three incompatible personal computers launched.
1983
Jupiter cancelled; company consolidates on VAX. Wall Street calls for his ouster.
1986
Named "America's Most Successful Entrepreneur" by Fortune.
1987
Begins calling Unix "snake oil."
1992
Forced to resign in July after two years of operating losses.
Tactic: Agree with your board, in writing and while things are going well, on the specific conditions under which you will step aside.
Principle 12
Build institutions that outlive the company
Around 1972, while Digital was still ascending, Olsen and Bob Everett rescued Whirlwind and parts of the TX-0 from the scrapper. That act of preservation became a 1975 exhibit, then the Digital Computer Museum opened by Maurice Wilkes in 1979, then The Computer Museum in Boston, and finally the Computer History Museum in Mountain View, which now houses Digital's own archives.
The company is gone. The museum remains, as does the Digital Federal Credit Union, chartered in 1979 for employees; as do the Loggia and kiosk at Gordon College; as do the practices and people Digital sent into the rest of the industry. Olsen's most durable creations were not his machines, most of which are now emulated rather than run, but the institutions he built around the work.
The principle for operators is that the company is one vehicle among several for what you care about. Olsen cared about interactive computing and about the people who did the work. The museum and the diaspora carried both forward after Digital could not.
Tactic: While your company is strong, invest in at least one institution (a user community, a foundation, an archive) designed to survive the company itself.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Ken Olsen, quoted by Gordon CollegeEven though I have been an entrepreneur, I have always been a scientist first and foremost.
— Ken Olsen, interview published by Daniel ScrivnerThe best assumption to have is that any commonly held belief is wrong.
— Ken Olsen, interview with the Middlesex News, 1987No one is productive, no one is creative without extreme discipline. If you don't have a goal, it's hard to create.
— Ken Olsen, Concord Oral History Program, January 4, 1997Now I ran a fairly large company and people were shocked that I didn't have a computer in my office. I use a computer all the time. I tell my secretary what to do with it.
— Gordon Bell, Digital VP of Engineering 1972–1983, on Olsen's death, 2011In the heady days of Bill Gates and Steve Jobs, it's too easy to forget that it was Ken Olsen's vision of interactivity that took computing away from the centralized mainframe and into the hands of the people.
Maxims
- Watch the line. What people wait for tells you more than what the benchmark says they should want.
- Earn the right to your ambition. A profitable first product that prepares the second buys you independence from nervous capital.
- Price draws the map. The number on the tag decides who is allowed into your category.
- Open platforms recruit free engineers. Document everything and let customers finish what you started.
- One sentence can run a company. But it cannot choose its strategy for it.
- Don't sell the machine they don't need. Honesty in sales compounds, so build compensation that rewards it.
- Freedom needs a stop button. Decentralized creation without centralized termination becomes paralysis.
- A rejected idea is not a dead idea. It is an idea you have handed to someone else.
- Contempt is not analysis. The crude, cheap competitor is usually doing to you what you once did to someone else.
- Know when to leave. The founder's last great decision is recognizing when he has stopped being the founder.
