Part IThe Story
What Causes Lightning
Do you know what causes lightning?
He asked it of a visitor in Redmond, Washington, in the last week of 1993, and he expected an answer. Not the dictionary answer, the one about a discharge of atmospheric electricity, which describes what happens without explaining why it happens. Leonard Bosack wanted the mechanism. Well, do you? The reporter from the Seattle Times who recorded the exchange described a man of forty-one with "a habit of asking point-blank questions when he is talking and staring intently out the window when he is not." His wife, Sandra Lerner, thirty-eight, sat nearby and peppered her own brisk speech with phrases like "Well, duh!" The two of them had, as Lerner put it to the same reporter, "an unreasonable amount of money." They were giving most of it away.
Three years and four months earlier, Bosack had walked out of the company he and Lerner had founded in their living room. He walked out the day they fired her. On that day, August 28, 1990, Cisco Systems had been public for just over six months and had come to market at a valuation of $224 million. Not quite ten years later, in late March 2000, it would be the most valuable company on earth, worth more than $500 billion, briefly larger than Microsoft. The founders were not there for any of it. They had sold their stock for roughly $170 million and moved to the Pacific Northwest, where Bosack was building, in a new company called XKL Systems, a device to speed up the rate at which a computer could move data in and out of itself. A plumbing problem. He has spent his entire working life on plumbing problems.
That is one version of the story, and it is the one most often told: brilliant engineer, outmaneuvered by venture capital, leaves a fortune on the table out of loyalty to his wife. It has the shape of a parable. Like most parables about Cisco, it is true in its outline and misleading in its emphasis, and the man at the center of it, who has given remarkably few interviews over four decades, seems to prefer it that way. Or seems not to care. With Bosack it is often hard to tell the difference.
The question about lightning is a better entry point than the firing. It shows a mind that distrusts the summary and wants the physics, a mind for which the explanation everybody repeats is precisely the explanation most worth interrogating. It is also, as it happens, the right question to bring to the founding of Cisco itself, an event about which almost everyone knows the dictionary definition and almost no one knows the mechanism.
By the Numbers
The Bosack Ledger
12/10/1984Date cisco Systems was incorporated in Menlo Park
$109KTotal sales in the roughly eight months to July 31, 1985
$2.5MSequoia Capital's 1987 investment, led by Don Valentine
$224MCisco's market capitalization at its February 16, 1990 IPO
$170MWhat Bosack and Lerner sold their Cisco stock for after leaving
~70%Share of their Cisco wealth the couple said they had given away by 1993
6.071 msXKL's fiber-plus-equipment latency over 1,231 km, against a physical floor of 4.106 ms
The Hobbit Under the Quad
Begin with a gift. Around 1980 and 1981, the Xerox Palo Alto Research Center gave Stanford University a number of its Alto workstations along with the Ethernet boards that let them talk to one another. The Alto was years ahead of anything else on a desk, and its graphical interface is the part of its legacy that ended up in museums. But at Stanford the thing that caught was the cable. Ethernet suggested that the university's scattered machines, which belonged to different departments, ran different operating systems, and spoke different protocols, might be made to behave as a single system.
The work happened underground, more or less. One veteran of the effort described the warren of offices beneath Margaret Jacks Hall and the Stanford Quad as "straight out of the Hobbit." Staff members and graduate students there began linking the computer systems of Stanford's schools and departments, informally and then less informally. The campus covered sixteen square miles. It held something like 5,000 computers grouped by building, which is to say grouped into islands: the computer science department ran one set of protocols, the Graduate School of Business another, the medical school yet another. Each island worked. None of them could reliably reach the others.
By June 1980 a research engineer at the medical school named William Yeager had built a working router. Yeager is the first of several people in this story whose names the legend omits. He wrote the software on a PDP-11; the circuitry came from a Stanford graduate student named Andy Bechtolsheim. The resulting device, which came to be called the Blue Box, could forward traffic among the campus's incompatible systems, and its architecture scaled well, a property that would turn out to matter enormously. Yeager's design, one later assessment concluded, "became a key to Cisco's early success." Yeager himself would later say that he had asked permission to sell the Blue Box commercially and been refused.
Bosack arrived in this world in 1979 as a graduate student in computer science and stayed on as staff. By the early 1980s he was running the computer science department's machines, and in one telling the department's Director of Computer Facilities. He was, in the IEEE's later phrasing, a support engineer on a 1981 project to connect all of Stanford's mainframes, minicomputers, LISP machines, and Altos. His specific contribution was to the router that allowed the network under his management, in computer science, to exchange data with the network at the business school. In Bosack's own recollection, the first router he worked on spoke PUP, the Xerox protocol native to the Altos, and among the early headaches were broadcast storms: floods of traffic in which machines shouting to every other machine on a network drowned the network itself.
Yeager's router did not, at first, connect to the Ethernet backbone that had become central to what Stanford was calling the Stanford University Network. In 1982, by one detailed history's account, Bosack, Lerner, and other engineers went ahead without authorization. They built an interface between Yeager's router and the Ethernet, and then they physically pulled the coaxial cable needed to string the campus together. Yeager joined them. So did others. It was a skunk works in the plain sense: unfunded, unsanctioned, and successful.
The founders' later account of this period, given on camera in the 1990s, has the flavor of guerrilla warfare remembered fondly. "It was do-it-yourself networking," one of them said. "If you wanted it, you had better do it yourself because no one else was going to do it for you. You couldn't buy it. We basically pulled wire through manholes. We pulled wire through sewer pipe." They had no money and "certainly didn't have any official sanction." The kicker came with a shrug you can hear in the transcript: the university was allowed not to like it, "but they did get a network out of it."
Bosack has described the arc more drily. They built bridges first, then crude routers, then better routers, and the routers solved for Stanford "the same sort of problem that it solved ten years earlier for ARPA": how to use a computer anywhere you wanted. This lineage matters. Bosack had worked on the PDP-10, a popular compute server on the ARPANET, during his years at Digital Equipment Corporation, and he would become a contributor to that emerging Defense Department network from Stanford. To him the campus project was not a new idea. It was an old idea, internetworking, applied at a scale and with a heterogeneity the original designers had not faced. The novelty was in the translation.
— Nick Veizades, Stanford staff member who worked on the router, to the San Jose Mercury News, 2001One can't define who did what, because it was pretty much of a cooperative effort. People tried to improve certain things so they worked better, and in this way they propagated.
Veizades's description is the most accurate account anyone has given of how the multiprotocol router came to exist, and it is a description of a network. Improvements propagated from person to person the way packets propagate from node to node, each one altered slightly in transit. Nobody owned the route. That fact, which seemed incidental in the Hobbit hole beneath the Quad, would become the central legal, moral, and commercial problem of Bosack's life.
A Man Who Knew How to Use Silverware
Leonard X. Bosack was born in Pennsylvania in 1952, on September 27 according to one registry, into a Polish Catholic family. The X was not his to begin with; it belonged to his father, also Leonard X. Bosack, whose full name would later sit on the letterhead of a charitable foundation. He graduated from La Salle College High School in 1969 and from the University of Pennsylvania in 1973. Records disagree on whether the degree came from the engineering school or from Wharton, and the disagreement is a small, telling one, since nothing about the man suggests a business school. He wanted to be a mathematician. He went into computers, he has said, because that was where the jobs were.
The jobs were good ones. He held technical roles at AT&T's Bell Laboratories and at Digital Equipment Corporation, where he worked as a hardware engineer on the PDP-10, the time-sharing machine that served as a workhorse of the early ARPANET. The PDP-10 is a recurring presence in Bosack's life, the way a first language remains in an accent. A copy of the 2001 newspaper investigation into Cisco's origins survives, of all places, on a website devoted to PDP-10 documentation.
Then Stanford, in 1979, and networking, which became his consuming interest. In Redmond in 1993 his gaze still went to the window and his questions came out point-blank. In a documentary interview from the 1990s he offered a theory of work that sounds like a joke until you realize he means every word.
— Leonard Bosack, documentary interview, 1990sSincerity begins at a little over 100 hours a week. You can probably get to 110 on a sustained basis, but it's hard. You have to get down to eating once a day and showering every other day, things of that sort, to really get your life organized to work 110 hours a week.
Sandra Lerner came from the other end of California and from a more complicated childhood. Her parents divorced when she was four, and she was raised between two aunts. One lived in Beverly Hills. The other ran a cattle ranch at Clipper Gap in the California Sierras, and it was there, at nine, that Lerner bought her first steer. She sold it at a profit two years later and used the money to buy two more. By the time she entered college she owned a registered herd of thirty head, and the herd paid her tuition. She finished a political science degree at California State University, Chico, in two years, in 1975. At Claremont, where she went for graduate work in econometrics, she wandered into the school's young computer lab, kept coming back, and was eventually made its manager. "People left you alone because they thought you were weird," she said later, "and nobody knew anything about what you were doing." Stanford, which in her telling was "dying to get graduate students" who could rescue faculty members baffled by computers, took her into its statistics and computer science program. She finished a master's in 1981 and became director of computer facilities at the Graduate School of Business.
So: a rancher's ward who financed her own education one steer at a time, and a Polish Catholic mathematician from Pennsylvania who found the world's operating systems more interesting than its markets. Each was a person who had wandered into a computer room and stayed because no one else understood what was happening there. One account has them meeting while sharing time on Stanford's minicomputers; another describes Bosack, at the time, as fitting "the description of a 'computer nerd.'" They married in 1980.
Asked on camera years later what had drawn her to him, Lerner gave an answer she prefaced as so bizarre "you'll just have to assume that it's true." Bosack's mother, she said, had done a miraculous job: "Lenny actually knew how to bathe and eat with silverware." She was, she said, "absolutely enchanted." She had not thought "a more perfect man could exist."
This is funny, and it is meant to be. It is also an exact portrait of the marriage's division of labor, delivered in comic form. She noticed the man's manners because in the world they both inhabited manners were rare; he had them because someone else had installed them. The pattern held. Lerner would become the one who dealt with the world outside the machine room. When Cisco had departments, she ran every one of them except engineering.
Did You Feed the Cat?
The legend goes like this. Two young Stanford employees, in love, worked in separate buildings a few hundred yards apart (the distance is given as 500 yards in one account, 500 meters in another, which tells you something about how carefully the details have been handled). Their departments' networks could not talk to each other, so they could not exchange email. Bosack built a box to connect the two networks so that he and his wife could send each other notes. The box became the router; the router became Cisco; Cisco became the internet.
Variations multiply. In one documentary the narrator explains that the couple were "unable to send e-mail messages like, did you feed the cat?" A technology blog, decades later, reduced it further: "They weren't trying to build the internet. They were trying to send love notes."
It is a wonderful story. It is also, according to every careful account, untrue. Cisco's own institutional memory came to concede as much. In 2001, after a San Jose Mercury News reporter named Pete Carey spent months reconstructing the history, a Cisco spokeswoman named Jeanette Gibson told him: "Obviously it was a team of people." Carey's piece began with the observation that "founding legends are a specialty of Silicon Valley, and none is more appealing than that of Cisco Systems," and then took the legend apart. It omitted Yeager. It omitted Bechtolsheim, Veizades, Kirk Lougheed, and the many staff members who touched the router. It omitted a fight with Stanford that nearly killed the company at birth. And it obscured, Carey wrote, "the true collective nature of the innovation that built Silicon Valley long before the hype and froth of the Internet bubble."
The unraveling had begun in 1998, when a PBS documentary called Nerds 2.0.1 credited Bosack and Lerner alone and provoked a long exchange of web postings from people who had been in the Hobbit hole and remembered it differently. Several books followed. Yet the legend held. Even Stanford's own website, in a description of a professorship that Cisco had endowed, continued to credit Bosack and Lerner with developing the device "that allowed computer networks to talk intelligently to one another." The university that had once threatened to prosecute the founders for taking its technology was now, in the copy beneath a named chair, giving them credit for inventing it.
Lerner, asked about all this by email, gave the best single sentence anyone has produced on the subject. "The only person I'm certain had nothing to do with it," she wrote, "is Al Gore."
It is a line that does three things at once. It acknowledges the crowd. It mocks the entire genre of retroactive credit-claiming. And it declines, with perfect deniability, to concede any specific person's share. That last move is worth dwelling on, because Bosack himself has rarely engaged the question at all. In the interviews he has given, he talks about protocols, about broadcast storms, about link-state versus distance-vector routing algorithms, about heat dissipation. He does not litigate his place in history. One reading is modesty. Another is that he has never found the question interesting, the way he did not find the dictionary definition of lightning interesting. The mechanism was collective, and he knew it; the legend was individual, and he let it stand.
The love-note story persists, one suspects, because it solves a narrative problem. A router is an abstraction. Nobody falls in love with an abstraction. Give it a couple separated by a lawn, and suddenly infrastructure has a heart. The irony is that the true story is more romantic, not less: two people, married, both running computer facilities, both pulling cable through sewer pipe without permission, both mortgaging everything they owned. You do not need the love notes. The love is in the mortgage.
The Living Room on Oak Grove Avenue
Word of the routers spread the way the routers themselves spread information: over the network. The crude email of the early 1980s carried news of the Stanford boxes to other universities and research centers, and requests started coming in. Bosack and Lerner converted their living room into an assembly room, building routers and testing them on the campus network and on the Interface Message Processors that connected Stanford to the ARPANET. Then they did the institutionally correct thing. They went to Stanford's Office of Technology Licensing and asked for a way to keep supplying the demand.
The office could not offer anything that would take less than years. Bosack and Lerner had, by the account of one networking historian, "little calling to become entrepreneurs." They were "happy academics willing to give what they had or knew away, resistant to even marking-up their production costs to make a profit." Stanford said they could not build product on university property and sell it to commercial customers. So, with what that historian calls little choice, they did it somewhere else.
On December 10, 1984, they incorporated cisco Systems, lowercase c. The name was taken from San Francisco, the city to the north, which one chronicler describes as the gateway to the Pacific Rim; the logo became a stylized pair of towers from the Golden Gate Bridge. The company's engineers would insist on the lowercase spelling for years, a typographical refusal to capitalize that tells you who those engineers thought they were. Not executives. Not a brand. Just a box that sat between other boxes.
The house, by one account, stood at 199 Oak Grove Avenue in Menlo Park, and early employees assembled routers on the kitchen table and shipped them by UPS. Incorporation changed little at first, except that now they could take orders from any organization. The first product Cisco sold as an independent company was reportedly a network interface card for Digital Equipment computers in 1985, which would have pleased the PDP-10 engineer in Bosack. In the roughly eight months ending July 31, 1985, total sales came to $109,000.
That is not nothing, and it is not enough. The orders kept coming, the living room kept filling, and the cash kept running out. Bosack and Lerner mortgaged their home to the limit, borrowed everything their credit cards would allow, and borrowed from friends. Some accounts describe salaries to friends who worked for them being deferred. Lerner, needing money to pay the household bills, took a day job managing research computing systems at Schlumberger's Computer Aided Systems Laboratory. She was, in other words, moonlighting at a corporate research lab in order to subsidize a startup that was itself subsidized by a mortgage, while her husband kept a university job running the very network the startup's product had grown out of.
The early team was small and remains, in most retellings, nearly anonymous: Kirk Lougheed, a Stanford staff engineer; Greg Satz, a programmer; Richard Troiano, who handled sales; in some accounts Bill Westfield. Lougheed deserves particular notice. Beginning in 1985 he and Bosack undertook a project to formally network the Stanford campus, and Lougheed adapted Yeager's router software into what would become the foundation of Cisco's operating system, IOS. That code was the company's core asset. (Decades later, Cisco would license the name "iOS" to Apple for its phones, so that a trademark descended from Yeager's PDP-11 software ended up on a billion devices that had nothing to do with routing.)
What the company sold, once it started selling routers in earnest in 1986, was the Advanced Gateway Server, or AGS, a revised version of the Stanford router. It was among the first commercially successful routers to support multiple network protocols, and it was sold, at first, almost entirely by word of mouth. In its first month of real selling, Cisco landed contracts worth more than $200,000. By 1987 the company had eight employees and was selling roughly $250,000 worth of routers a month, much of it to engineers at universities, research centers, aerospace firms, and government agencies who ran the same protocols and Unix machines as Stanford and who heard about cisco over the ARPANET itself.
The marketing channel was the product's own network. The customers were people like the founders. And the selling proposition was translation: a box that could speak TCP/IP, the Xerox protocols, DECnet, and whatever else a heterogeneous organization had accumulated, so that the organization did not have to choose. In the protocol wars of the 1980s every vendor wanted its standard to win. Cisco's insight, if it can be called a strategy at all rather than an engineer's natural impatience with the question, was to decline to fight. Let the protocols war. Sell the interpreter.
Whose Software
Here the story turns uncomfortable, and the founders' admirers tend to hurry past it.
The router software that Cisco was selling had been written at Stanford, on Stanford time, by a Stanford employee who said he had been refused permission to commercialize it. Bosack and Lougheed had adapted it. The hardware design descended from Bechtolsheim's Stanford circuitry. By some accounts the founders continued in their Stanford jobs after incorporating Cisco, working on what was, in effect, an improved copy of the Blue Box. On July 11, 1986, Stanford forced Bosack and Lougheed to resign. The university considered filing criminal complaints against Cisco and its founders for the theft of its software, hardware designs, and other intellectual property.
It did not. In 1987 Stanford licensed the router software and two computer boards to Cisco. One later account puts the settlement at $150,000 plus the software licenses, though the figure appears in only a single source. Whatever the precise terms, the outcome was a negotiated peace: the university got paid something and got a campus network out of the whole affair; Cisco got legal title to its origins.
How should one judge this? A later YouTube documentary put it bluntly: the founders "stole internal tech that was used at Stanford." An encyclopedia entry calls the use of the software illegal and the eventual settlement a resolution of "copyright violation charges." The founders' sympathizers point out that the work was collective, that the protocols were largely in the public domain, that Stanford's licensing office had been offered the business and could not move fast enough to take it, and that the university's own website would one day credit the founders with the invention. Lerner, telling the story years later on a podcast, emphasized exactly that tension: the protocols were public, the work had been collaborative, and yet Cisco had still managed to build products people wanted to buy.
All of these things are true at once, which is why the episode is worth more than a verdict. Stanford's network was a commons in the technical sense: improvements propagated, as Veizades said, and no one could say who did what. A commons cannot sell itself. Someone has to draw a boundary around a piece of it and take it to market, and the act of drawing that boundary is always, in some light, a theft from everyone else who stood inside it. Bosack drew the boundary. He was not the only person who could have. He was the one who did, at the cost of his job and very nearly a criminal referral.
Then he did something that sharpens the paradox. On January 29, 1988, Cisco filed an application for a patent titled "Method and apparatus for routing communications among computer networks." The sole named inventor was Leonard Bosack. He assigned his interest to Cisco on April 18, 1988. The patent, number 5,088,032, was granted on February 11, 1992, more than a year after he had left the company. It would be litigated as late as 2003, in the Eastern District of Texas, and its family would be flagged in patent databases as having critical litigation history. The man who had taken a router out of a commons where "one can't define who did what" now had his name, and only his name, on a federal patent describing a method for routing among networks.
There is no evidence that Bosack saw any contradiction. Patents were what companies filed; Cisco was a company now. But it is hard to read the filing date without noticing that in the same season the company's other great act of enclosure was underway: its founders were losing control of it.
Adult Supervision
By 1987 the living room could no longer contain the business. The founders needed capital, and they went looking for it in the place Silicon Valley had built for exactly this purpose. By most accounts they were turned away by many investors before Sequoia Capital agreed to put in $2.5 million. Sequoia's own records list Bosack as a founder it "partnered" with in 1987. The partner was Don Valentine, and Valentine's terms were the terms of a man who understood exactly how much leverage he had over two people with a maxed-out mortgage. One account says the investment bought roughly 29 percent of the company. Another says Valentine demanded and received controlling influence, and that he became chairman. The numbers vary by source; the direction does not.
The episode of Nerds 2.0.1 that dealt with Cisco was titled "Adult Supervision." That was the frame the industry used, and to some extent still uses: technical founders are children; investors bring in grown-ups. Cisco's first chief executive, Bill Graves, held the job from 1987 to 1988. Then came John Morgridge, appointed president and CEO in 1988. Accounts differ on who chose him. Lerner's own biography says that she and Bosack brought him in; others say Valentine did. Either way, Morgridge set about replacing a number of Cisco's managers, many of them friends of the founders, with people who had more conventional business experience. Bosack was given the title of chief scientist. Lerner became vice president of customer service.
Morgridge, an industry veteran, belongs to the type the founders' story requires and resents: the professional manager. He would run Cisco until 1995, and decades later he would deliver the graduation address at the Stanford Graduate School of Business, the institution whose computers Lerner had once managed. The symmetry is almost too neat. The business school where she ran the machines eventually invited the man who had fired her to tell its graduates how to succeed.
What the professionals did, in fairness, worked. In 1988 Cisco expanded its target market beyond universities and research labs to mainstream corporations with far-flung branch offices, each running different networks. For those customers the company built routers that supported an ever-wider array of protocols, until it could claim to support more of them than any competitor. When a commercial market for internetworking emerged in the late 1980s, Cisco was positioned at its center, with reasonably priced, high-performance boxes and a growing reputation among exactly the engineers who chose what to buy. The company was producing what the IEEE would later describe as revolutionary technology: the first multiport line cards designed specifically for routers, and sophisticated routing protocols. In fiscal 1990 sales reached $69.8 million and the staff numbered 254.
On February 16, 1990, cisco Systems went public on the NASDAQ at a market capitalization of $224 million. Some accounts give the offering price as $18 a share. The founders, on paper, were rich. They were also, in practice, increasingly irrelevant to the company's governance, and in Lerner's case increasingly at war with it. The accounts describe frequent disagreements between Lerner and Morgridge, disagreements about management style. She was not, by temperament or by biography, a person who had learned to defer. She had paid for college with cattle.
Making the Cisco Connection, David Bunnell's 2000 history of the company, describes Valentine's investment as one of the great returns in venture history, and it was. Bunnell's book also captures how the industry saw Cisco by the end of the decade: as a company that had "convincingly trounced" competitors like 3Com and Cabletron without the benefit of a near-monopoly handed down by IBM, in the way Microsoft and Intel had been. Read with the founders in mind, the book is a study in absence. Cisco's ascent is told almost entirely as a story of what happened after they left.
August 28, 1990
The firing itself has been told many times, and the essential facts are spare. On August 28, 1990, a little more than six months after the IPO, Cisco's management fired Sandy Lerner. When Bosack heard the news, he resigned. Several accounts say it was the same day. There is no record of deliberation.
What is striking, in every version, is the speed. Bosack did not negotiate. He did not stay on as chief scientist to protect his technical legacy, or to preserve his stock options, or to keep an eye on the code he had carried out of Stanford at such cost. He did not argue that the company needed him, though it is difficult to imagine that the routers did not. He simply left, the way a packet leaves a node, without looking back at the node.
Then the couple sold. Over time they disposed of all their Cisco stock, for a combined figure usually given as $170 million; a contemporary estimate put their wealth at about $190 million in 1990. In the ledger of Silicon Valley regret, this is one of the canonical entries, the founders who "lost everything" by selling a company that would be worth half a trillion dollars at its peak. A YouTube video on the subject bears that phrase as its title. The arithmetic is undeniable and also somewhat beside the point. What the founders gave up was not money they would have spent. They gave most of what they did take away.
It is worth pausing on what happened to the company without them, if only to measure the distance between Oak Grove Avenue and San Jose. Revenues exceeded $1 billion for the first time in 1994. In 1993 Cisco made its first acquisition, Crescendo Communications, and then kept buying: Kalpana, Grand Junction, and in 1996 StrataCom, for roughly $4 billion in stock, at the time the largest acquisition in Silicon Valley history. John Chambers succeeded Morgridge as chief executive in 1995. In 1996 Cisco's sales and net income both more than doubled, to $4.1 billion and $913 million. In 1999 the company bought seventeen businesses, including Cerent, a fiber-optic equipment startup in Petaluma, for about $7 billion. The company that had once refused to pick a winner in the protocol wars now refused to pick a winner in the technology wars: rather than invent everything internally, it used its soaring stock as currency to buy small companies, some little more than research teams, and fold them in. It was, in a sense, the router strategy applied to corporate development. Connect everything. Translate between it. Own the junction.
In late March 2000 Cisco became the most valuable company in the world, at more than $500 billion. Then the bubble burst and the company lost $431 billion of market value, roughly 80 percent, by 2002. As late as 2021 its shares remained 42 percent below their 2000 peak. Juniper Networks, shipping its first product in 1999, took about 30 percent of Cisco's service-provider market within a year. The company that had built the plumbing of the internet learned that plumbing, too, can be priced as though it were magic and then repriced as plumbing.
And then, slowly, it recovered. On December 10, 2025, Cisco's stock closed at a record for the first time since the dot-com peak, a quarter century later. December 10 was the company's forty-first birthday. In fiscal 2026 Cisco reported record revenue of $63.3 billion, with billions in orders for artificial-intelligence infrastructure. It is a member of the Dow Jones Industrial Average, which it joined on June 8, 2009, replacing General Motors. Its products have been entangled in nearly every controversy a networking company can be entangled in, from alleged assistance to Chinese internet surveillance to NSA implants intercepted in shipping. Routers, it turns out, are not neutral. Whoever controls the junction sees everything that passes through it.
None of this has much to do with Leonard Bosack, which is the point. By the time Cisco was worth anything, he was gone.
Weird Science
In 1991 the couple moved to Redmond, Washington. That same year they formed a foundation with $20 million and gave it a name that is easy to skim past: the Leonard X. Bosack and Bette M. Kruger Charitable Foundation. It was named for his father and her mother. Two people who had each, in very different ways, grown up as products of someone else's care put their parents' names on the money.
The foundation's work fell into two categories, and the categories were split by temperament. Lerner chose the animal-welfare grants. Bosack chose what the couple cheerfully called "weird science." In 1992 the foundation backed eight animal-welfare projects, including one in Topeka, Kansas, where prisoners at a correctional institution bathed, groomed, and socialized shelter animals before they were put up for adoption. The couple funded PetNet, a computer network linking Northern California animal shelters so people could find missing pets or animals to adopt. Lerner, who had rescued her favorite cat, Lilly, from a shelter where it faced certain death, made no apology for funding animals while people suffered. "People say, 'There are so many people things — why are you helping animals?'" she told the Seattle Times. "Well, I think it goes hand in hand."
PetNet is the detail that rhymes. Even in philanthropy, the founders' first instinct was to connect isolated systems: a dozen shelters, each with its own records, unable to talk to each other until someone built the interconnect.
Bosack's grants read like the reading list of a restless polymath. He paid for an evolutionary biologist trying to explain why menstruation occurs in humans. He bought vehicles for three field scientists studying tigers in Russia, Indochina, and India. He sponsored a three-year scholarship for a University of Washington student doing atmospheric research. The couple funded research at Harvard and Stanford searching the cosmos for extraterrestrial intelligence, a project NASA had once supported and then dropped. They sent 275 tons of grain to the St. Petersburg Theological Academy in Russia to help feed the hungry. And they bought Chawton House in Hampshire, England, a manor that had once belonged to Jane Austen's brother, Edward Austen Knight, and set about restoring it as a center for the study of early women's writing.
Science for science's sake, they told their visitor, often yields unexpected and exciting discoveries. That was when Bosack asked about lightning.
— Sandra Lerner, to the Seattle Times, December 1993We have an unreasonable amount of money.
The marriage did not survive the decade's first half; the couple divorced in the early 1990s. Their subsequent lives diverged in a manner almost too symmetrical for fiction. Lerner put part of her stock proceeds into a venture partnership called & Capital Partners, which funded a cosmetics company called Urban Decay, whose tagline asked, "Does pink make you puke?" The French conglomerate LVMH bought it in February 2000, a month before Cisco's market peak. In 1996 she purchased Ayrshire Farm, a 600-acre estate in Upperville, Virginia, with a 42-room Edwardian fieldstone mansion, and made it the first farm in Virginia to meet both Certified Humane and Certified Organic standards. The rancher's ward returned to livestock. Chawton House opened in 2003 as the Centre for the Study of Early Women's Writing, 1600–1830; in 2015 she received an honorary OBE in a ceremony held there; in 2016 she resigned from its board and withdrew her foundation's support. In 2011, under the pen name Ava Farmer, for "a Virginia farmer," she published a sequel to Pride and Prejudice.
In 2013 she told CNBC that technology companies had lost the long view. "When I was in graduate school," she said, "all of the big companies had labs, all doing blue sky research." Now, "nobody has a long-term view anymore, it's quarter-to-quarter and nobody is willing to fund blue sky research." And in September 2023 she wrote the foreword to a book called Founder vs Investor, about the toxicity in relationships between venture capitalists and startup founders. Thirty-three years after August 28, 1990, she was still writing about it.
Bosack stayed with the machines.
The Floor Set by Light
XKL began, in the 1993 telling, as a project to speed up the input-output ability of a computer, which is to say to move data faster across a boundary. The company later became XKL LLC, privately funded, with Bosack as its chief executive, and it turned toward optical networking: the business of moving light through glass over long distances. This is the most physical layer of the stack. Routers decide where packets go. Fiber is how they get there.
The company's signature achievement, as described in the citation for the IEEE Computer Society's 2009 Computer Entrepreneur Award, which Bosack received jointly with Lerner, is an in-line fiber optic amplification system that transmitted data across 1,231 kilometers of fiber, roughly the distance from Chicago to New York, with a total latency, fiber plus equipment, of 6.071 milliseconds. Light alone cannot traverse that distance of glass in less than about 4.106 milliseconds. The rest is equipment, and the art is in shrinking it. Bosack's stated conviction, repeated in the award materials, is that the inherent but often untapped physics of fiber optic components can be leveraged to move data faster with devices that use less power, less space, and less cooling.
It is the lightning question again. Do not give him the summary. Give him the physics.
In recent years he has occasionally surfaced to talk. At an industry conference called AutoCon 4, he sat for a podcast with the host Scott Robohn and walked through his life in roughly the order a router processes a packet: header first, payload later. He described the shift from mathematics to computing, the PDP-10, the Altos, the PUP router, the broadcast storms. He recalled the support of ARPA, which had encouraged the commercialization of the Stanford technology. He talked about the IETF and the bruising debates over link-state versus distance-vector routing, and about the creation of BGP, the Border Gateway Protocol that still governs how the internet's large networks exchange routes, a protocol remembered in the industry by its nickname, the "Two Napkin Protocol." He talked about how young people should get into networking, and argued for paths into the field that did not run through traditional college programs, which is an interesting position for a man whose company was born, contested, and finally licensed by a university. Intelligence, energy, a willingness to learn: those, he said, were the requirements. Not credentials.
Then he talked about the present, and the present, for Bosack, is a heat problem. Scaling networks beyond 800 gigabits per second runs into the limits of what existing technology can cool. He argued for innovation in optical networking, and for separating the data plane, which carries traffic, from the control plane, which decides where traffic goes, in order to make networks more secure and more reliable. It is a characteristically unsentimental view of the internet he helped make possible: not a utopia, not a marketplace, but a thermal system with failure modes.
The man on that podcast stage is somewhere in his seventies. His net worth, estimated at about $200 million, is a rounding error beside the company he founded. His name appears on the patent, which expired in 2009, the same year he and his former wife accepted an engineering award together for a company neither had worked at in nearly two decades. The patent's claims, once Cisco's property and before that his, and before that, in the honest version, everyone's in the Hobbit hole, now belong to no one again.
And he is still at work on the gap. Between 4.106 milliseconds, which is what light needs to cross 1,231 kilometers of glass, and 6.071, which is what his machines need, lies 1.965 milliseconds: about two thousandths of a second, the whole remaining territory of the problem. It is the part he can do something about. Somewhere under the floor of a data hall between Chicago and New York, a pulse of light is crossing it right now, and nobody watching it pass could tell you whose idea it was.
Part IIThe Playbook
What follows is not a celebration of Leonard Bosack's choices. Several of them cost him dearly, and at least one exposed him to criminal referral. It is an attempt to extract, from a career that ran from a university basement to the most valuable company on earth and then deliberately away from it, the principles that are durable, along with the ones that came with a price tag. Bosack never wrote a management book and gave few interviews; the evidence is in what he built, what he signed, and what he walked out of.
The principles
- 1.Build for the problem you are standing inside.
- 2.Sell the interpreter, not the language.
- 3.Let the network be the sales channel.
- 4.Enclosing a commons is a debt, not a windfall.
- 5.Unsanctioned work needs an exit plan.
- 6.Bootstrap until you know what you have, then stop.
- 7.Price capital in control, not in dollars.
- 8.A title is not a seat.
- 9.Decide your loyalties before the meeting.
- 10.Correct the legend while the witnesses are alive.
- 11.Divide by temperament, not by task.
- 12.Go down a layer when the stack gets crowded.
⏱
The Arc in Dates
From a Pennsylvania high school to an expired patent: the decisions that shaped the Bosack story.
1973
Graduates from the University of Pennsylvania; joins Digital Equipment Corporation as a hardware engineer.
1979
Enters Stanford to study computer science; later runs the department's computers.
1980
William Yeager has a working router by June; Bosack and Lerner marry.
1982
Bosack, Lerner, and others build an unauthorized Ethernet interface and pull cable across campus.
1984
cisco Systems incorporated on December 10 after Stanford's licensing office cannot move quickly.
1986
Bosack and Lougheed forced to resign from Stanford on July 11; Cisco begins selling routers by word of mouth.
1987
Stanford licenses router software and boards to Cisco; Sequoia invests $2.5 million.
1988
Routing patent filed January 29 naming Bosack sole inventor; John Morgridge becomes CEO.
1990
IPO on February 16 at $224 million; Lerner fired and Bosack resigns on August 28.
1991
Bosack-Kruger Foundation formed with $20 million; move to Redmond and XKL.
2009
IEEE Computer Entrepreneur Award with Lerner; routing patent expires.
Principle 1
Build for the problem you are standing inside
Bosack did not discover the multiprotocol router through market research. He discovered it because it was his job to make the computer science department's network talk to the business school's network, and because the existing tools could not do it. The founders were, in the phrase of one later admirer, "practitioners solving their own constraints." The customers who came first, at universities, research centers, and government labs, were people with exactly the same constraint, running the same protocols on the same kinds of machines.
The advantage of building from inside a problem is not empathy, which can be faked, but calibration. Bosack knew what a broadcast storm felt like at three in the morning. He knew which protocols coexisted badly. That knowledge was embedded in the AGS in a way no outside product manager could have replicated, and it explains why Cisco's early reputation spread among engineers, the most skeptical buyers in technology, rather than among executives.
The risk is equally specific: a product built for your own problem is sized to your own world. Cisco's growth required Morgridge's push in 1988 toward mainstream corporations, a market the founders had not built for. The inside view finds the wedge. It rarely finds the whole market.
Tactic: Before you write a pitch, write down the operational failure you personally experienced last week that your product would have prevented, and find ten people who experienced the same one.
Principle 2
Sell the interpreter, not the language
The 1980s were a decade of protocol wars. TCP/IP, the Xerox suites, DECnet, and others each had partisans, and vendors wagered their futures on which would win. Cisco's routers did something structurally different: they supported as many protocols as possible, and by 1988 the company could differentiate itself by supporting more of them than any competitor. It did not need to know who would win. It profited from the fact that no one had won yet.
This is a position, not just a feature. Intermediaries in a fragmented market collect value from the fragmentation itself; every new standard is another thing they can translate. Cisco later applied the same logic to corporate development, refusing to bet on internal invention and instead buying whatever technology the market produced, from Crescendo in 1993 to Cerent in 1999.
The lesson has a dark side that the Cisco controversies make vivid. Whoever sits at the junction sees all the traffic, and the junction becomes an object of interest to governments, intelligence agencies, and litigants. Neutrality is a commercial posture; it is not a moral exemption.
Tactic: When your market is split among competing standards, ask whether there is a business in connecting all of them, and build that before you pick a side.
Principle 3
Let the network be the sales channel
Cisco's first sales force was the ARPANET. News of the Stanford routers traveled by email to other research institutions; orders came back the same way. The company sold its routers "by word of mouth" beginning in 1986 and landed more than $200,000 in contracts in its first month. By 1987, with eight employees, it was moving roughly $250,000 of hardware a month without anything resembling a marketing department.
The deeper point is that the product and its distribution were made of the same material. Every router installed made the network that carried news of routers larger and more valuable. The customers were a community of engineers who talked to each other constantly, and the product was the thing that let them talk.
Founders today tend to translate this as "product-led growth," which flattens it. What Cisco had was rarer: a product whose usefulness was most visible to precisely the people best positioned to recommend it, communicating over a channel that the product improved.
Tactic: Map where your earliest users already talk to one another, and design your first version so that using it is visible, and useful, inside that conversation.
Principle 4
Enclosing a commons is a debt, not a windfall
The router that founded Cisco was, in Nick Veizades's words, "pretty much of a cooperative effort" in which "one can't define who did what." Bosack took a piece of that collective work out of Stanford and sold it. Stanford forced him out, considered criminal complaints, and then in 1987 licensed the software and boards to Cisco. In January 1988, Cisco filed a routing patent naming Bosack as sole inventor.
Commercializing shared work is often necessary; the Stanford licensing office could not, by one account, offer any path that took less than years, and Lerner later noted that even with public-domain protocols someone had to build a product people would buy. But the founders paid for the enclosure for decades, in a 1986 forced resignation, in a 2001 newspaper investigation, and in a legend that their own company eventually had to correct.
The debt to a commons can be paid in money, in licenses, in credit, or in all three. What it cannot be is ignored. Bosack and Lerner's later practice of funding open-ended "weird science" reads, in this light, as something like a repayment in kind.
⚖
The Legend vs. the Record
How the popular Cisco origin story compares with what the sources document.
| The legend | The record |
|---|---|
| A couple built a router to send each other love notes. | Described as an untrue legend; the campus networking project long predated the founding. |
| Bosack and Lerner invented the multiprotocol router. | William Yeager had a working router by June 1980; Bechtolsheim designed hardware; many staff contributed. |
| The company was financed by credit cards alone. | Mortgage, credit cards, loans from friends, Lerner's day job, then $2.5 million from Sequoia. |
| A clean garage-to-glory story. | Forced Stanford resignations on July 11, 1986; criminal complaints contemplated; 1987 license. |
| The founders lost everything. | They sold for about $170 million and gave roughly 70 percent away. |
Tactic: If your product grew out of shared or institutional work, negotiate the license and the credit before you take the first order, not after the first lawyer's letter.
Principle 5
Unsanctioned work needs an exit plan
In 1982 Bosack, Lerner, and other engineers built an interface between Yeager's router and Stanford's Ethernet "without authorization" and pulled coaxial cable through manholes and sewer pipe. As they put it later, the university "was kind of allowed not to like it, but they did get a network out of it." Skunk works succeed by ignoring permission. That is their function.
But the same habit, carried across the line from building a campus network to selling a commercial product, produced the July 1986 crisis. The behavior that earned gratitude in 1982 earned a forced resignation four years later, because the context had changed from improving the institution to profiting outside it. The founders do not appear to have noticed the line until they were across it.
The lesson is not "ask permission." It is to know, at every stage of an unsanctioned project, what the institution will believe it owns once the project works, and to have a plan for that moment before it arrives.
Tactic: Whenever you build something on an employer's time or infrastructure without formal approval, write down today who would claim ownership if it became valuable, and resolve that question before it does.
Principle 6
Bootstrap until you know what you have, then stop
Bootstrapping gave Cisco two things: evidence and pain. The evidence was $109,000 in sales in the company's first eight months and a flood of orders arriving through the ARPANET. The pain was a home mortgaged to the maximum, credit cards at their limits, borrowed money from friends, and Lerner taking a job at Schlumberger to pay the household bills.
The evidence was valuable; it proved demand and let the founders build something real before outside money arrived. The pain was not. It eroded their negotiating position precisely when they needed it most. By the time they reached Sequoia in 1987, they were people with a working business and no cushion, which is the weakest possible stance from which to negotiate control.
Bootstrapping is a discovery tool. Used past the point of discovery, it becomes a liability that transfers leverage to whoever eventually funds you.
Tactic: Set a specific signal, such as a revenue figure or an order backlog, at which you will raise money, and start fundraising when you hit it rather than when the cash runs out.
Principle 7
Price capital in control, not in dollars
Sequoia's $2.5 million is remembered as one of the great returns in venture history. For the founders, its real price was not equity percentage but governance. Within a year Valentine was chairman, Morgridge was CEO, the founders' friends in management were being replaced, and Bosack was "chief scientist." Two years after that, Lerner was fired and Bosack was gone.
Founders tend to negotiate the visible number, valuation and dilution, and wave through the terms that determine who decides. Bosack and Lerner, desperate for cash, accepted terms that made their removal a matter of board discretion. Lerner's 2023 foreword to a book about founder-investor conflict suggests she has spent a long time thinking about what that cost.
📒
The Founder's Ledger
Key decisions, what they bought, and what they cost.
| Decision | What it bought | What it cost | Verdict |
|---|---|---|---|
| Unauthorized campus networking (1982) | A working network; technical credibility | Set a precedent for ignoring institutional ownership | Build |
| Commercializing Stanford code (1984–86) | A product with proven demand | Forced resignations; legal threat; a 1987 license | Risk |
| Extended bootstrapping (1984–87) | Proof of demand without dilution | Mortgage, debt, weak negotiating position | Mixed |
| Sequoia investment (1987) | $2.5 million and professional management | Control of the company | Risk |
| Resigning on August 28, 1990 | Integrity; freedom | Any further role in Cisco's ascent | Exit |
| Giving away about 70 percent of proceeds | A portfolio of open-ended research | Most of the windfall | Build |
Tactic: In any financing, list every term that changes who can hire or fire executives, and negotiate those before you discuss valuation.
Principle 8
A title is not a seat
When Morgridge arrived in 1988, Bosack received the title of chief scientist and Lerner became head of customer service. Both titles sounded senior. Neither came with the power to stop what happened next. The managers who were the founders' friends were replaced, and the founders were left with roles that described their expertise while removing their authority.
"Chief scientist" is a particularly instructive case. It honors technical contribution while quietly separating its holder from operating control, budget, and the board. For a founder whose leverage derived from deep knowledge of the product, it can be a gilded exit ramp. Bosack's resignation two years later suggests he understood exactly how little the title protected.
The question to ask of any role is not what it is called but what it decides: who reports to it, what it can veto, and whether it has a vote.
Tactic: When offered a new title during a reorganization, ask in writing which decisions you will be able to make or block, and treat any answer of "none" as a resignation offer.
Principle 9
Decide your loyalties before the meeting
The most consequential decision of Bosack's career took, by every account, almost no time. When the company fired Lerner on August 28, 1990, he resigned. There is no record of hesitation, negotiation, or a counteroffer. It is the rare case of a founder whose priorities were so settled that the crisis required no deliberation.
The financial cost was enormous on paper. The couple sold for about $170 million a company that would later be worth more than $500 billion. Whatever one thinks of the trade, it was not the product of panic. It was the product of a decision made long before the meeting, about which partnership mattered more.
This is not a recommendation to resign in solidarity. It is a recommendation to know, in advance, what you would do, because the moment itself will offer no time to think, and the people across the table will have already decided their own position.
Tactic: Write down, privately and before any board conflict, the three outcomes that would cause you to walk away, so that if one occurs you act on a prior commitment rather than an impulse.
Principle 10
Correct the legend while the witnesses are alive
The love-note story, and the broader myth that two people invented the multiprotocol router, persisted for years in Cisco marketing, news coverage, and even on a Stanford web page. It took a 1998 documentary, a wave of web postings by former staff, several books, and a 2001 newspaper investigation before Cisco's spokeswoman would say, "Obviously it was a team of people."
Legends are commercially useful. They also corrode trust among the people who know the truth, and they invite exactly the kind of investigation that exposes uncomfortable details such as forced resignations and contemplated criminal complaints. Lerner's "Al Gore" line was a deft partial correction. Bosack's general silence left the field to others.
Founders who correct their own origin stories early get to tell the complicated version on their own terms. Those who wait get it told for them.
Tactic: Write an honest, named list of everyone who contributed to your first product and publish it while those people can still confirm it.
Principle 11
Divide by temperament, not by task
At Cisco, Lerner ran every department except engineering. At the foundation, Bosack picked the "weird science" grants and Lerner picked the animal-welfare grants. The split was not a formal allocation of responsibilities so much as a recognition of who each person was: one who stared out the window asking about lightning, one who rescued a shelter cat and built a network of animal shelters.
Partnerships that divide work by temperament avoid two failure modes: duplication, where both partners do the same thing, and false equality, where each insists on a share of every decision. The Bosack-Lerner division worked well in the company's early years and again in philanthropy. It may also have been a vulnerability: with Lerner absorbing the company's external conflicts, Bosack had few independent relationships with the investors and managers who eventually decided the founders' fate.
Tactic: With any co-founder, explicitly assign domains according to who is happiest doing what, and then make sure each of you still holds at least one direct relationship with every board member.
Principle 12
Go down a layer when the stack gets crowded
After Cisco, Bosack did not try to build a better router company. He went lower: first to a device for speeding up a computer's input and output, then to optical networking, the physics of moving light through glass. XKL's in-line amplification system achieved 6.071 milliseconds of latency over 1,231 kilometers of fiber, against a theoretical floor of about 4.106 milliseconds. His stated conviction is that untapped physics in fiber components can deliver more speed with less power, space, and cooling.
This is a recurring pattern in infrastructure. When a layer becomes dominated by a giant, in this case the very company he founded, the opportunity often moves to the layer beneath, where the constraints are physical rather than commercial. Bosack's recent focus on heat management beyond 800 gigabits per second, and on separating data and control planes, follows the same logic: find the place where the laws of physics, not the incumbents, set the limit.
Tactic: When an incumbent owns your layer, identify the physical or technical bottleneck one level down that limits everyone above it, and build there.
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
— Sandra Lerner, in an email to the San Jose Mercury News, 2001, on who deserves credit for the routerThe only person I'm certain had nothing to do with it is Al Gore.
— Jeanette Gibson, Cisco spokeswoman, to the San Jose Mercury News, 2001Obviously it was a team of people.
— Sandra Lerner, on computing at Claremont in the 1970s, to San Jose's Business JournalPeople left you alone because they thought you were weird, and nobody knew anything about what you were doing.
— Sandra Lerner, to CNBC, April 2013We're still reaping the benefits of the blue sky research from 20 to 30 years ago. In 20 to 30 years, it's not still going to be there.
Maxims
- Ask for the mechanism. The dictionary definition describes what happens; the useful answer explains why.
- Build where you bleed. The best early products come from failures you have personally had to fix.
- Translate, don't take sides. In a fragmented market, the interpreter can be worth more than any single language.
- Distribute through the product's own network. When your users already talk to each other, make your product part of the conversation.
- Enclosure creates debt. Taking shared work to market is sometimes necessary, and the bill arrives eventually.
- Bootstrap to learn, not to suffer. Self-funding proves demand; prolonged self-funding surrenders leverage.
- Negotiate who decides before what it's worth. Control terms outlive valuations.
- A title without a vote is a farewell. Ask what a role can decide, not what it is called.
- Settle your loyalties in advance. The decisive moment will not give you time to deliberate.
- When the layer is full, go down one. Physics is the only incumbent that never gets acquired.
