Part IThe Story
The lightning struck in February 2007, and it found the one thing in Musgrave that could not afford to be hit. A broadcast tower in a working-class neighborhood of the South African city had, for eight years, been the hub of a small nonprofit internet network. Costas Criticos had built it for students and low-income families, sharing a single high-speed connection among neighbors at a discount, because broadband in South Africa cost several times what it cost in the United States. The network was a wheel. Every home was a spoke, and every spoke ran back to the tower. When the bolt came down there was a sharp crack, the kind people remember, and eight years of work shorted out at once. That is how a wheel fails. All of it, together.
What happened next is the reason this story begins in Musgrave and not in Cambridge or Mountain View or the boardroom on Sand Hill Road. A friend of Criticos's, visiting from the Massachusetts Institute of Technology, had brought along several small gadgets from a Silicon Valley startup called Meraki. They were not powerful. The flagship, called the Mini, cost $50 and was, by most accounts, neither the fastest nor the strongest wireless router on the market. But the boxes did not need a tower. Each one talked to the others, passing data from box to box, so that a page or an email could zigzag across a neighborhood instead of beaming out from a single point. If one link broke, the rest noticed and found another way around. "In really just a few hours, we had the whole community up and running," Criticos told a reporter from The Christian Science Monitor later that year. "Not to mention, we set up the whole thing without the help of technicians, and our costs have never been cheaper." A hundred Meraki nodes spread across a two-mile area. Subscribers paid $7 a month for broadband speeds. Dial-up, which was slower, had cost $30.
The CEO of the company that made those boxes was 25 years old. His name was Sanjit Biswas, and when the Monitor asked him about Musgrave he could recount a dozen stories like it: housing projects, poor neighborhoods, villages in countries he had never visited, each one wired together by people who had plugged in a box and found that it worked. He told the stories, the reporter noted, with a pride more like a parent's than a businessman's. Then he ran out. "Well, those are all the stories I can think of right now," he said.
Almost two decades later, Biswas runs a public company worth something in the neighborhood of $20 billion, depending on the week, whose sensors and cameras ride in millions of commercial vehicles and log roughly 90 billion miles of driving a year. Before that he sold his first company to Cisco for $1.2 billion. Forbes, in 2023, put his net worth at $3.6 billion and his source of wealth, with the brutal economy of a list, at "sensor systems." None of this is what a reasonable person would have predicted from the mesh network in Musgrave, which was a nonprofit, ran on $50 boxes, and was meant for people who could not afford the internet. And yet the architecture in that neighborhood—no center, every node deciding for itself, the failure of one part routed around by all the others—turns out to be the clearest diagram anyone has drawn of how Biswas thinks, how he builds, and why the things he builds tend to keep running after something has hit them.
By the Numbers
Two Companies, One Architecture
$1,500 → $50Cost of a mesh router at MIT vs. Meraki's Mini
150 nodesSize of MIT Roofnet, spread across six square miles of Cambridge
$1.2BCisco's price for Meraki, announced November 18, 2012
$11.5BSamsara's valuation at its December 2021 IPO, priced at $23 a share
$1.9BSamsara annual recurring revenue at fiscal year-end, growing 30%
90 billionMiles of driving data Samsara captures annually
2–10 wattsPower budget of the edge devices running Samsara's AI inference
Six square miles of rooftops
Before there was a company, there was a roof. Then another roof, then 150 of them.
MIT Roofnet was an experimental 802.11b/g mesh network run out of the university's Computer Science and Artificial Intelligence Laboratory, and the ambition of the thing was almost comically literal: put radios on rooftops across Cambridge, Massachusetts, and see if they could form a working network on their own. Biswas and a fellow graduate student named John Bicket led it. They built it, as Biswas would later describe it, from the ground up—radio chipsets, firmware, management software, the routing protocols that decided where every packet went. By the time they were done, Roofnet covered six square miles with 150 nodes. Along the way it taught them every way a network in the physical world can fail. Nodes had to configure themselves, because no one was going to climb onto a triple-decker in February to type commands into a router. Routes had to be found and lost and found again. The weather did what New England weather does.
The routers they started with cost $1,500 apiece. That figure matters, because nearly everything that came later is a story about making it smaller.
The intellectual core of Roofnet was a routing protocol, and its key insight was a matter of manners. In most mesh networks of the era, every node that heard a packet simply passed it along, a crude kind of shouting. In Biswas's version, as MIT Technology Review explained when it named him to its TR35 list of innovators under 35 in 2007, the nodes talked to each other first. They compared notes on the packet's destination and on the strength of their own signals, and then decided among themselves which one should forward it. The protocol accounted for a network that would not hold still: users signing on and off, conditions shifting minute to minute, or—the example the magazine chose, and one worth holding onto—a passing truck that blocks a node's radio signal.
A network, in other words, that expected interruption and was built to negotiate its way around it.
Biswas came to that problem honestly. He was born in January 1982 to a Bengali family and grew up in the South Bay, graduating from Lynbrook High School in San Jose in 1998, in the years when the valley around him was being rewired by the commercial internet. By about age ten he was running early versions of Linux and dialing into bulletin board systems, which is to say he discovered networks in their most handmade form—one modem, one phone line, one stranger's server in someone's spare bedroom. As a teenager he helped get his high school online, a project he still brings up in interviews, the first instance of a pattern that would repeat: a place that was not connected, a kid who could see how to connect it, and a willingness to do the unglamorous wiring. He took a degree in computer systems engineering at Stanford, finishing in 2002, and then went east for graduate work at MIT, where he earned a master's in electrical engineering and computer science and enrolled in a doctoral program he would, technically, never leave. Every official biography he has had since includes some version of the same clause: on leave from the Ph.D. program. A leave of nearly twenty years. It is less a credential than a running joke at his own expense, or perhaps a door he likes to keep open.
John Bicket is harder to sketch, which is itself a sketch. The two met as graduate students in the same lab, and Bicket has been beside Biswas at every turn since: co-lead of Roofnet, co-founder and chief technology officer of Meraki, co-founder of Samsara. He is the partner who appears in every company history and almost none of the quotes. When he does surface in Biswas's stories, it is usually doing something slightly out of character with tremendous commitment—a detail we will come to.
Something else was happening in that lab in 2006, and it is one of those juxtapositions that seem planted. A group of MIT researchers, including the computer scientist Hari Balakrishnan, published a paper on a system called CarTel, which proposed putting commodity sensing units on ordinary vehicles to monitor road traffic and road hazards. Years later the paper would win a test-of-time award from the Association for Computing Machinery's mobile computing group, which noted that the concept had been widely adopted in popular map applications. Biswas was not an author. He was, that year, busy leaving to start a Wi-Fi company. But it is hard to read the CarTel abstract now without a small shiver of recognition: the trucks were already there, in the building, in 2006, waiting for him to come back around to them. When he returned to MIT for a fireside chat in March 2023, in room 32-123, Balakrishnan was one of his two hosts.
A thousand routers and a thesis advisor
Origin stories multiply as companies get valuable, and Meraki has at least two.
In one, recorded by Technology Review in 2007, a nonprofit group approached Biswas in 2006 and asked whether the Roofnet technology could help bring internet service to poor communities. He was intrigued enough to take a leave of absence. In another, which Biswas told at a networking-industry event in 2012, the turn came after he presented the Roofnet research at Google. Afterward, an enthusiastic person in the audience came up and wanted to buy routers. The product did not yet exist in any form you could ship.
Doug Leone, the Sequoia Capital partner who would sit on Meraki's board, supplied the name and the scale: Chris Sacca, who was at Google at the time, was enthusiastic, and Google bought 1,000 routers and invested. (One retelling of Biswas's version renders the order as $1,000 worth of routers, a difference of roughly three orders of magnitude; Leone's figure is the one with a venture capitalist's signature on it.) Both origin stories are probably true at once, the way most origin stories are—a moral pull from one direction and a commercial pull from the other, arriving in the same season. The tension between them would take three years to resolve.
There was a problem of geography. Biswas and Bicket were in Boston. Y Combinator was a few months old. The angel investors who in Palo Alto would have written a first check over coffee simply did not exist in Cambridge in any meaningful density. So they did the thing graduate students with a working prototype and no money do. "We went out to early customers," Biswas recalled on the CEO.com podcast, "and we resold the product, telling them, 'We'd like to build this. If you put an order down, we can go make it happen.' And that's how we got the company off the ground." The company bootstrapped its first year on product sales. "We didn't actually have venture money," he said elsewhere. "We managed to sell enough product to get off the ground." The founding team had a third member, Hans Robertson, who would run product management; it is a small and rather perfect irony that, as of this writing, Wikipedia's entry for Cisco Meraki tags Robertson's co-founder status with citation needed, and flags the entire article—about a company Cisco bought for $1.2 billion—as possibly failing the site's notability guidelines. Networks, when they work, are invisible. Apparently so are the people who build them.
The introduction to Sequoia came through Rajeev Motwani, a Stanford professor best known in the valley as the thesis advisor to Larry Page and Sergey Brin. Motwani was, in the language of the thing Biswas built, a high-signal node: he did not originate the packet, but he knew which neighbor should forward it. Leone remembered the founders as "a bunch of MIT PhDs who had built a very proprietary solution as part of their own thesis," and he remembered something else more vividly. He bought one of their products to test it. "It was so easy to use, I set up a wireless network myself in minutes," he wrote. "You just plugged in the box and it worked. That's all we needed to see."
Biswas's own memory of the pitch is of being seen through. "Doug and Jim had quickly cut through all the info we gave them and said, 'OK, what's unique here is that these guys are making wi-fi networks easy to deploy.' And that was not the case in 2006."
Leone loved the team and did not love the plan. Meraki's founders wanted to sell their gear to building owners, who would offer free Wi-Fi to tenants as a way to compete on rent. "We love the guys, we want to partner with you," Leone told his partners, "but we have to agree that the current go-to-market plan needs to be changed." He was, by his own admission, not sure. "It took my partners to get me over the line." The partner who did it was Jim Goetz, whose argument was a single sentence and, in retrospect, the entire investment thesis: "If you've got a special team like this, they'll figure it out, so let's go."
That sentence is worth sitting with, because it is the opposite of how venture capital likes to describe itself. Goetz was not betting on a market, or a product category, or a business model. There was no business model. He was betting that the people who had made 150 radios on Cambridge rooftops find each other in bad weather would, given time, find a customer.
The next billion
For a while, though, the customer they found was the world.
In the summer of 2007 the journalist Christopher Mims went to Harlem to see what Meraki was doing for Scientific American. He sat in the neighborhood's first Starbucks, at 125th Street and Lenox Avenue, and found that of the three wireless networks within range, only one was open. Then he went walking with Michael Lewis, who ran a budding nonprofit called Wireless Harlem. Ten blocks north they sat on a bench in the shade of the first apartment complex Wireless Harlem had hooked up, and Lewis pulled out a laptop and laid out the arithmetic. The median income in Harlem was $35,000 a year. "At the end of the day when people make a decision about what they're going to spend money on," he said, "it's not going to be Internet access." Lewis was a man running a network for people the market had priced out, and Meraki was the only hardware that made the math work.
Mims described Biswas as "a polite and bespectacled Massachusetts Institute of Technology student-cum-CEO on permanent hiatus from the pursuit of a doctoral degree in computer science." He also discovered, on his own and later, what the company's name meant—meraki is a Greek word for putting a piece of yourself into the thing you make, doing it with love—and he noted, with a dry precision that is hard to improve on, that no one at the company had ever mentioned this to him. "There is such a thing as being too earnest," he wrote.
The earnestness was real. Biswas told Mims his goal was to "connect the next billion people." He told the Monitor the same thing in December, in fuller form.
— Sanjit Biswas, to The Christian Science Monitor, December 2007The mission of the company is to bring affordable Internet access to the next billion people. We've always felt a social obligation in this work.
And for a stretch the numbers backed him up in a way that would have flattered any startup, social mission or not. "We now have more than 1,000 networks around the world," he told Mims, "and all that growth was through word of mouth." Meraki did no advertising. By the time Technology Review wrote him up, Meraki-based networks were running in 25 countries, from Slovakia to Venezuela, serving more than 15,000 users. By December 2007 the Monitor counted several thousand networks across 70 countries. The model was elegantly human: one person plugs a router into a broadband connection, the neighbors stick routers to their windows, and a mesh of up to hundreds of people forms on its own. Users could give access away or sell it to their neighbors. The company's engineers were almost all programmers, and yet the thing they sold was hardware—cheap, commodity hardware built by outside vendors and animated by Meraki's software into what Mims could only call "a kind of magic."
Then came San Francisco, and the magic briefly took on the shape of a dare.
Municipal Wi-Fi was, by late 2007, one of the great collapsed enthusiasms of the decade. EarthLink had cooled on it. So had AT&T and even Google. San Francisco's own big plan had been, in the view of the wireless industry analyst Andrew Seybold, "doomed to failure before the first access point was installed." Seybold had spent years watching city after city announce a network and then discover that almost nobody would subscribe. He had concluded that municipal Wi-Fi's moment had passed or, more pointedly, "that there never was a time for it."
Into this graveyard walked Meraki, offering to blanket San Francisco—roughly 50 square miles—with free Wi-Fi at 1 megabit per second, with no money from the city and no outside investment beyond its own venture partners. The Register ran the story under a headline that captured the absurdity of the moment: a Google-funded startup would offer free Wi-Fi to the whole city. The plan required 10,000 to 15,000 devices, nearly all of them repeaters placed inside homes and businesses, which meant that Meraki would essentially have to go door to door recruiting volunteers to stick boxes in their windows. From many places in the city, Biswas acknowledged, you could already see 40 or more access points competing for the air. He said Meraki had a "magic sauce" for the interference.
Seybold called him in January 2008 to ask the obvious question. What was the business model?
He was, he wrote afterward, startled by the answer. Biswas told him there was no business model. San Francisco was a test bed.
You can read that exchange two ways, and the reading tells you more about the reader than about Biswas. To a skeptic like Seybold, it was the sound of a startup that had raised a fresh $20 million round from Google and Sequoia and did not yet know what it was for. To someone who had sat through the Roofnet years, it sounds like a researcher describing an experiment—which, in a sense, is exactly what it was. Biswas would say much later that in the beginning, "We originally didn't think of it as a startup. We thought of it a little more as a project, which is, hey, this would just be cool if we could see this research have impact." The San Francisco network was a 50-square-mile version of the Cambridge rooftops, only now the rooftops were living rooms and the grant was venture capital.
It is not clear the distinction was obvious to him at the time. It was about to become very obvious.
A walk in the woods
The phrase belongs to Leone, and Biswas has adopted it with evident affection: the "walk in the woods." It describes the period after Sequoia's Series A when Meraki had a working product, a gifted team, a celebrated humanitarian mission, a free network spreading through San Francisco, and no idea what market it was actually in.
The product was the problem, or rather its generality was. "The product was so general that people were finding all kinds of creative applications for it," Biswas said, "and that meant we were getting pulled in all different directions." Every market was plausible. None was obviously right. Leone, recounting it, ticks off the failures like a man reciting a list of ex-girlfriends: building owners, then Wi-Fi for municipalities—"that wasn't a market"—then Wi-Fi for the rest of the globe—"that wasn't the answer"—then free, ad-supported Wi-Fi—"that wasn't a business"—then Wi-Fi for the home, sold through the Best Buys of the world, "and that would have just been too painful." Elsewhere the list grows to include retail, small enterprises, large corporations, and "even the developing world."
— Doug Leone, Sequoia CapitalIn some ways, our greatest value to the company was to provide them a setting for an extended period of time where they could continue to iterate on their go-to-market strategy. Our greatest value was that we did absolutely nothing.
The length of the walk depends on who is telling it. In the oral history Sequoia later published, Leone says it took "five tries and two or three years to find the right business model." Biswas calls it "a three-year period after Sequoia made our Series A investment." But on November 18, 2012, writing the congratulatory note on the day Cisco announced its acquisition, Leone put it differently: "We wasted nearly a year evaluating different markets." A year, or three. Memory compresses the wandering when the destination turns out well, and expands it when you want to make a point about patience. Both versions are told by the same man, which is somehow reassuring.
What ended the walk was not insight. It was fear.
In October 2008, with the financial system coming apart, Sequoia gathered the CEOs of its portfolio companies and walked them through a presentation that has since become a minor artifact of valley history, a slide deck whose title page read R.I.P. Good Times. Biswas was in the room. "At that point things got a lot more serious," he recalled. "We realized, 'Hey, we can't just keep experimenting and running science projects. We need to focus on a single segment and prove we've got a business here.'"
Science projects. It is the most self-incriminating phrase in his vocabulary, and he uses it without flinching.
They chose small and medium-sized businesses—the SME market, in Leone's shorthand—for product fit and speed of revenue. The choice looks inevitable in hindsight and was anything but at the time: it meant walking away from the mission that had made the company a darling of Scientific American and toward branch offices, hotels, restaurants, retailers, and schools, customers with no romance and plenty of cash. "Meraki took off," Leone said. "We couldn't add salespeople fast enough."
Somebody, though, had to be the first salesperson, and in the beginning there wasn't one. So the founders did it themselves. This is the detail about John Bicket, and Biswas tells it with obvious delight: "He literally put on a suit. He was like a method actor." The two MIT engineers got call lists, picked up the phone, and called. Inside sales, cold, by the co-lead of a research network that once spanned six square miles of Cambridge rooftops. Biswas would come to argue that this was not a detour from the engineering but a continuation of it. As a technical founder, he said, "you can talk in a deep way about the product and you can get ideas about what to build next from talking to customers." The call list was a sensor network. Every rejection was data.
There is a version of this story in which the turn to small business is a betrayal of Musgrave and Harlem—the moment the mission died and the company was born. Biswas has never told it that way, and the more interesting truth is that he did not need to give up the architecture to give up the mission. The thing that made Meraki work for Costas Criticos—you plugged in a box and it worked, with no technicians—was exactly what a regional restaurant chain with one overstretched IT person wanted. The poor neighborhood in South Africa and the dentist's office in Sacramento had the same problem. Neither could afford an expert.
The box that phones home
Leone called it "the smartest move," and it came in 2009, the year after the slide deck: Meraki released business-class, cloud-based software to control its hardware.
To understand why it mattered you have to remember what managing a network looked like before it. Leone painted the picture with the precision of a man who had heard the complaint many times: "Before Meraki if you got a call at 1am you'd have to go into the office, access a control panel and fix the issue yourself, or even travel to the location of the problem." Networking gear was configured box by box, on site, by specialists who knew the incantations. Meraki inverted it. The boxes stayed in the closets and on the ceilings, carrying the customer's actual traffic locally, while a stream of management data flowed up to Meraki's data centers. An administrator could sit anywhere with a browser and see every deployment, every device, across every branch, on one dashboard. If the link to the cloud dropped, as Biswas explained to a room of network engineers in January 2012, the local network kept working. The cloud was the brain; the local box kept breathing on its own.
It was, again, the Roofnet bargain: let the nodes make their own decisions, but let them talk. And it was multi-tenant, the only offering of its kind according to Leone—one shared platform serving every customer, which meant constant updates, no maintenance for the buyer, and the economics of scale. In the language that would become fashionable a few years later, Meraki had turned networking into software-as-a-service while its competitors were still shipping boxes. "They simplified a complex ecosystem," Leone wrote, "and pioneered networking as a service."
The company's attention to the person at the other end of the dashboard became a kind of signature. Meraki ran usability tests. It shipped feature updates continuously. Its interface included a small feature called "Make a wish," a box where any customer could type what they wanted the product to do, which went straight to the people building it. Network engineers—a notoriously hard audience—asked about it specifically when Biswas appeared before them in February 2013. A feedback loop built into the product itself. A user, deciding which neighbor should forward the packet.
The financial story of those years is less romantic and more instructive. Revenue doubled every year, including through the crisis. Meraki raised about $80 million across four rounds, with Sequoia leading, and did not spend all of it. The R.I.P. Good Times deck had done its work. "We didn't know if the next round of capital would be there," Biswas said. "It was very important for us to be operating right around cash flow break-even. In retrospect, a lot of lessons were learned there as well about financial discipline and always having a plan that was rock solid." This is a company that a year earlier had offered to give an entire city free Wi-Fi with no business model. Now it ran at break-even on purpose, with money in the bank it chose not to touch.
The product line grew outward from the wireless access point the way a mesh grows outward from the first node: branch routing, security appliances, VPN and firewall, WAN optimization, mobile device management, and in time gigabit switches. By early 2014 Meraki counted more than 10,000 customers worldwide. And Biswas, who had once put down the management books because he found them useless, kept leading the way he knew how. "I ended up just kind of putting them all down at the beginning," he said, "and saying, look, this is actually just a big project, right? So let me do what I know how to do, which is lead a project, and kind of get it to the next stage."
Leone's verdict on the founders, written at the end, has the ring of a man who did not expect to be writing it: "They made very few mistakes."
When was the last time you plugged in an Ethernet cable?
On November 18, 2012, Cisco Systems announced it would acquire Meraki for approximately $1.2 billion in cash and retention-based incentives. Meraki had somewhere between 300 and 400 employees. Six years had passed since the founding in Mountain View; the company had since moved to San Francisco, the city it once proposed to blanket in free Wi-Fi.
Leone opened his blog post that day with a question for the reader: "Quick: when was the last time you plugged in an Ethernet cable?" It was a good line, and it contained an irony he was too gracious to name. Cisco was the company that had sold the world its Ethernet. Sequoia had been an early partner to Cisco too, a quarter-century earlier, and Leone noted with evident satisfaction that the two "transformational companies" were coming together "25 years later." The venture firm had, in effect, financed both the wire and its replacement, and now the replacement was being bought by the wire.
The founders stayed. Biswas became vice president and general manager of what Cisco named its Cloud Networking Group, headquartered in San Francisco, in a new building. Bicket and Robertson stayed in their roles too, which Leone singled out as a point of pride. Under Biswas, the division grew more than 100 percent, making Meraki one of the fastest-growing product lines in Cisco's history. To Biswas the experience was, by his own account, a second education—Cisco was 30 years old, had survived several technology cycles, and was a living textbook in how a very large company actually functions. He went to the 2013 Wireless Field Day to reassure skeptical engineers that the platform would survive inside the giant. They asked him about the culture. They asked about the wish box.
In April 2014 he gave a talk in Santa Clara at the TiE Conference Center, an evening event hosted by the entrepreneurs' organization founded by South Asian technologists. The program listed him as the VP/GM of Cisco's Cloud Networking Group and described his trajectory under the title "MIT Research to a Cloud Leader." There was dinner at six, the program at seven, questions at a quarter past eight. A young man from a Bengali family in San Jose, now a Cisco executive, telling a room of aspiring founders how a research project became a company. The setting is not incidental. TiE exists precisely to be a mesh for people who came from somewhere else—to make sure no one has to route through a single hub.
He did not stay long. In 2015, he and Bicket left.
A coda belongs here, though it happened after they were gone. On August 3, 2017, Meraki engineers made a change to the company's North American object storage service, applying what Cisco described as "an erroneous policy." It deleted customer data: custom floor plans, photos of where devices had been mounted, hold music, voicemail greetings, the custom logos customers had uploaded to brand their dashboards. Within days Meraki told customers it did "not expect to be able to recover most assets." The cloud that had saved administrators the 1 a.m. drive to the office had, for one week in August, demonstrated what it means to put every customer's memory in one place. The mesh can route around a lightning strike. A central policy cannot be routed around. It is not a story about Biswas, but it is a story about the bet he made, and it rhymes with the tower in Musgrave in a way that is hard to unhear.
We've never driven a truck
Most founders who sell a company for $1.2 billion before the age of 31 do something else next. Biswas did something else next, but not in the way the phrase usually implies.
The second company started backward. Meraki had begun with a technology—"essentially our PhD research from MIT put into a box," as he put it—and then spent years searching for someone to sell it to. Samsara began with the someone. "We started with the customer in mind, the market in mind," Biswas said, "and said, we want to help and be useful in the world of physical operations. We didn't know the first thing about it."
— Sanjit Biswas, on the CEO.com PodcastTo be honest, we didn't know the first thing about operations. We've never driven a truck or a forklift or worked in a warehouse. So, a lot of learning along the way, just getting out there, spending time with customers, and then going back and building products to solve their problems.
The world of physical operations—trucking fleets, construction sites, utility crews, warehouses, cranes—was enormous, unglamorous, and, as Biswas would later frame it, full of neglected, high-value problems that software had largely ignored. A former Sequoia podcast host described the pivot as Biswas going from domain expert to newcomer, carried by little more than curiosity about how infrastructure actually works. The two founders who had once spent their nights on Cambridge rooftops were now spending their days learning how a forklift operator logs a shift.
They had advantages this time and were suspicious of them. They had run a startup, run a division of a giant, and had money of their own, and they used it to fund Samsara's earliest days. When outside capital came, Andreessen Horowitz led the Series A and Marc Andreessen took a seat on the board that he still holds; General Catalyst came in a few rounds later. Forbes, in May 2015, framed the new venture with the logic of the old one: the Meraki founders thought they could reshape sensors the way they had reshaped switches. How did they pitch firms like that? Biswas has said they mostly didn't. They showed a working product, told stories about customers, and told investors to call the customers. "We didn't spend a lot of time pitching," he said. "We spent most of our time doing."
He was aware of the danger in the résumé. "If you try to do a startup in easy mode, you're probably not going to succeed," he said. "Don't let the Meraki win kind of taint the way that you approach the second one. And if anything, use it for pattern-matching recognition."
The name is worth a pause. No account in the record explains it, and the founders seem in no hurry to. But the first company was named for a Greek word about pouring yourself into your work, and the second takes its name from an old Sanskrit word for the cycle of return—the wheel that comes back around. It is either a coincidence or the most understated joke in enterprise software. A Bengali engineer, whose first company was named in Greek for the love of the craft, names his second for the thing that happens again.
What happened again was the box. Samsara sold sensors, cameras, and cloud software—hardware as the Trojan horse for recurring revenue, the same bargain Meraki had struck, now aimed at vehicles and job sites instead of office ceilings. By June 2017, when it closed a $40 million round, Samsara was emphasizing sales through the channel. By September 2019 The Wall Street Journal reported it was raising at a $6 billion valuation. By June 2021 it had raised $930 million in total. That December, Samsara listed on the New York Stock Exchange under a ticker that was also a category: IOT. It sold 35 million shares at $23 each, raising $805 million at a valuation of $11.5 billion.
Two years later, in the quarter reported at the start of December 2023, Samsara crossed $1 billion in annual recurring revenue—less than a decade after founding. By the end of a later fiscal year it stood at $1.9 billion in ARR, growing 30 percent, and had reported its third straight quarter of GAAP profit on quarterly revenue of $478.8 million. Biswas credits focus, and he credits the customers. "Companies pay to have their problem solved," he said. "And so when we talk about things like revenue and revenue growth, for me, it's actually impact and problem-solving growth." His test for product-market fit is the moment a customer calls someone else into the office, or in from the hallway, and says: you need to see this.
The company that once had two founders working a call list in borrowed suits now hires executives on the scale of a mature industrial. In a letter dated May 29, 2023, Samsara offered Lara Caimi the title of President of Worldwide Field Operations, reporting to Biswas, at a base salary of $500,000, an annual bonus of up to $500,000, and restricted stock units worth approximately $23 million, vesting quarterly over four years. It is a remarkable document mostly for how unremarkable it is—the standard paperwork of a company that has stopped being a project and become an institution. Biswas, by then, was among the ten youngest people on the Forbes 400. He was 41.
The third shift
Ask Biswas what he thinks about at the edge of his imagination, and he tends to start talking about the hours most people sleep through.
"If you think about it, there's, like, a third shift between midnight and 8:00 am roughly, right, that people tend not to work because they're sleeping," he told Sequoia's Training Data podcast. "Imagine if operations like logistics could still run during that shift." A field technician who needs a part could simply have it delivered. The cost of moving things falls, and when it falls, he argued, demand rises: "Yeah, I could use that part, but I don't need it delivered if it's going to cost 50 bucks for someone to drive it to me. If it costs five bucks or no bucks, like, how awesome would that be?" The arc from $1,500 to $50, now pointed at the price of moving physical things.
The technology underneath this ambition has the same shape it had on the Cambridge rooftops. Samsara's AI runs, in large part, not in a distant data center but on the devices themselves—edge hardware with a power budget of two to ten watts, compressed and distilled models doing task-specific work on the truck, in the moment. Biswas has argued that for autonomy workloads, the edge beats the centralized data center. The data it feeds on is everything the clean world of screen-based AI is spared: weather, road conditions, and what he calls the long tail of human behavior. The diversity of real-world data, he says, is both the biggest challenge and the greatest asset. In 2007, the passing truck that blocked a radio signal was noise for his protocol to work around. Now the truck is the signal.
The company says its products help prevent more than 200,000 crashes a year. Biswas describes the payoff of all that sensing in strikingly humane terms—real-time coaching for drivers, recognition for frontline workers, an effect he connects to workforce morale. A careful reader will notice what goes unsaid. The camera that coaches a driver is also a camera pointed at a driver; the same stream that can recognize a safe maneuver can document a careless one. The "next billion" language of 2007 imagined networks that liberated people from expensive gatekeepers. The physical-operations network of the 2020s sees the people inside it with unprecedented clarity, on behalf of the companies that employ them. Biswas talks about it as care. It is also, unavoidably, oversight. Both things are true about the same two-to-ten-watt box, and the tension is not one his company's materials spend much time on.
He has built a way of running the company that looks designed to survive his own blind spots. Samsara allocates its R&D on a rule he calls 70-20-10: seventy percent toward making the core products and platform better, twenty percent toward near-term upgrades the company is fairly confident will matter to customers, ten percent toward long-shot bets, many of which will fail, a few of which will graduate into the twenty. Feedback runs through formal and informal loops, culminating each year in a customer conference called Samsara Beyond, where the company spends an entire day sitting with customers in small groups, workshopping ideas. "The great thing about this is as your customer base grows, the amount of feedback you get grows," he told Turner Novak. The trick is to look for patterns. If a request matters to most customers in a segment, build it. If it matters to the other twenty percent, it may still be a good idea—just one better served by a partner, or by an API that lets the customer build it. The wish box, grown up.
And he has a habit of firing himself, in a sense. Once a year Biswas runs a 360-degree review on his own performance and asks what the company needs most from him next—a practice he calls rehiring himself. He audits his own calendar, color-coded, comparing where he planned to spend his time against where it actually went, a process now automated. A node checking its own signal strength before deciding whether to forward the packet.
The deepest difference between the two companies is not technical. Meraki, from fairly early on, had an obvious buyer on the horizon. Samsara does not. Biswas has said he sees no single dominant competitor foreshadowing an acquisition, and the market is large enough to keep building. Without a defined end, the company thinks in longer arcs about technology and hiring and culture, runs more experiments before committing, and—in one of the clips from a long conversation he gave the How Leaders Lead series—operates on the conviction that shortcuts today create problems tomorrow. Another clip from that conversation gave the whole interview its title, and it may be the best four-word summary of his method: big eyes, big ears.
His life outside the company has drifted, quietly, toward the same project. Biswas is married to Hope Biswas, an infectious disease epidemiologist; they have three young children, and he has talked publicly about the problem of raising kids with a work ethic inside an affluent life. In 2023 the couple started the Biswas Family Foundation, a grant-making foundation that funds teams of scientists at research institutions. In March 2024 they announced, with the Milken Institute, $15 million for research at the intersection of AI and health, alongside a $5 million grant to the Gladstone Institutes betting on computational biology and AI to change the future of cancer. The engineer who wanted to connect the next billion people to the internet now funds the people trying to connect sensor data to disease. He reads biographies of technology leaders to unwind. His reading list, as recorded elsewhere, includes Matt Ridley's The Rational Optimist and Yuval Noah Harari's Sapiens—two books about how humans, in large numbers, figured out how to pass things to one another.
The passing truck
In October 2025 Biswas came back to Stanford, where he had graduated twenty-three years earlier, to speak at the Entrepreneurial Thought Leaders series in the NVIDIA Auditorium on Via Ortega. The lessons he offered were deliberately plain. Go deep on the right problem. Build with a world-class team. Stay grounded in your customers. He had given a version of the same talk to a Stanford class a few years earlier, where a lecturer's summary rendered it as: customer feedback loops, because it is easier to course-correct than to guess perfectly; build for the long term, because it takes years to make a big impact; and win as a team, because research, product, and startup building are all team sports. The instructor introduced him, a little generously, as Dr. Biswas. The doctorate is still on leave.
The advice is so modest that it almost disappears, which is perhaps the point. Biswas has never been a founder of declarations. "People get very intimidated by the entrepreneurial process," he has said. "But if you return to your roots as an engineer and say, 'Hey, I just like to build things,' whether it's products or companies, you can just give it a shot and experiment." The downside, he argues, is not that big. The upside is enormous, because products scale.
There is a temptation, writing about someone like this, to find the hidden drama—the rupture, the betrayal, the crisis that remade him. The record does not oblige. The founders who met in a lab are still together. The board member who doubted the go-to-market still speaks of them with something like paternal pride. The mission that sounded too earnest to say aloud became, without much fanfare, a business, and then a second business, and then a foundation. The dramatic events in his story mostly happened to other people's infrastructure: a lightning strike in South Africa, a financial system collapsing in 2008, an erroneous storage policy at a company he no longer ran. In each case the lesson was the same lesson. Don't build a tower.
Somewhere tonight, in the hours between midnight and eight, a truck is moving down a highway with no one watching it but a small box behind the windshield, drawing a few watts, making its own decision about what it sees. It does not need a technician. It does not need a hub. It phones home when it can and keeps working when it can't. Eighteen years ago, a graduate student wrote a protocol that treated that truck as the thing that got in the way—the shadow that passed between two radios on a Cambridge roof and broke the link. The box on the truck has been listening ever since.
Part IIThe Playbook
What follows distills the operating logic visible across two companies, two decades, and one unusually consistent architecture of thought. These are not the slogans Biswas gives on stage, though some overlap. They are the patterns that recur when you lay Meraki and Samsara side by side—what he did when things were uncertain, what he refused to do when things were easy, and how the engineering instincts of a mesh-network researcher became a management philosophy.
The principles
- 1.Sell the box before you build the company.
- 2.Run the company as a project until it forces you not to.
- 3.Take the walk in the woods—then let a crisis end it.
- 4.Move the complexity onto yourself.
- 5.Leave capital in the bank on purpose.
- 6.Start the second company from the market, not the lab.
- 7.Put the engineers on the phones.
- 8.Build the feedback loop into the product.
- 9.Allocate like a portfolio: 70-20-10.
- 10.Rehire yourself every year.
- 11.Build as if no one is coming to buy you.
- 12.Let the customers do the pitching.
S
The Arc of Sanjit Biswas
From dial-up bulletin boards to edge AI on millions of vehicles.
1998
Graduates from Lynbrook High School in San Jose after helping get the school online.
2002
Completes a B.S. in computer systems engineering at Stanford.
2006
Takes leave from MIT's Ph.D. program to co-found Meraki with John Bicket and Hans Robertson, built on MIT Roofnet.
2007
Named to MIT Technology Review's TR35 at age 25; Meraki networks spread to 70 countries.
2008
Sequoia's "R.I.P. Good Times" meeting ends the walk in the woods; Meraki focuses on small and mid-sized businesses.
2009
Meraki launches business-class cloud-managed software.
2012
Cisco announces the $1.2 billion acquisition on November 18; Biswas becomes VP/GM of the Cloud Networking Group.
2015
Leaves Cisco to co-found Samsara with Bicket; a16z leads the Series A.
2021
Samsara lists on the NYSE as IOT at $23 a share, an $11.5 billion valuation.
2023
Samsara crosses $1 billion in ARR; Biswas and Hope Biswas start the Biswas Family Foundation.
2025
Returns to Stanford's Entrepreneurial Thought Leaders series with three lessons: the right problem, the right team, the customers.
Principle 1
Sell the box before you build the company
Meraki's first year was funded not by investors but by customers. In Cambridge in 2006 there were no angels to speak of and Y Combinator was months old, so Biswas and Bicket went to early buyers with a promise: put down an order and we will make it happen. That presale bootstrapped the company until Sequoia arrived. The order from Google—a thousand routers, in Leone's telling—did the same work at larger scale, converting a research talk into evidence of demand.
The lesson is not merely "get revenue early," which is an old saw. It is that a purchase order is a different kind of signal than an investor's interest. Investors are betting on you. Customers are betting on the thing. When Leone finally met the team, he didn't evaluate a deck; he bought the product, plugged it in, and set up a network himself in minutes. Biswas arrived at Sand Hill Road already holding the most persuasive argument a hardware startup can make: people had paid for this before it fully existed.
Tactic: Before you raise, find three customers willing to put money down on the product you describe, and treat their willingness as the only validation that counts.
Principle 2
Run the company as a project until it forces you not to
Biswas famously put down the management books. Facing the job of CEO with no experience, he decided the company was "actually just a big project," and that he would do what he knew how to do: lead a project to its next stage. Meraki was conceived the same way—not as a startup but as a way to see research have impact.
This framing is underrated as a cure for founder paralysis. A project has a next milestone, a team, and a technical problem; a "company" has a thousand imagined obligations borrowed from people running other companies. Biswas has said that aspiring founders get stuck in analysis paralysis, planning without moving. The project frame collapses the gap between thinking and doing. But it has a limit, and Biswas found it. A project mindset is what produced the San Francisco free Wi-Fi experiment with "no business model." The frame works until the stakes demand a market, at which point the project must become a business. Knowing which phase you are in is the actual skill.
Tactic: Define your company's next stage as a single project with one deliverable, and revisit the frame the moment survival depends on choosing a market.
Principle 3
Take the walk in the woods—then let a crisis end it
Meraki spent somewhere between one and three years wandering among markets: building owners, municipalities, the developing world, ad-supported Wi-Fi, consumer retail. Sequoia tolerated it; Leone later called doing nothing "our greatest value." The patience was real and it mattered—Biswas credits it as the turning point that made the company worth a billion dollars.
But the exploration did not end through insight. It ended when the 2008 downturn put the founders in a room with a slide deck titled R.I.P. Good Times, and Biswas realized they could no longer "keep experimenting and running science projects." Constraint did what curiosity could not. They picked small and medium businesses, and the company took off.
The non-obvious lesson is that exploration and focus are not opposites but phases, and that founders rarely end the first phase on their own. Generality is seductive—a product that can do anything finds admirers everywhere and customers nowhere. The healthiest response is to use external shocks as permission: the downturn, the runway, the board's warning are the occasion to stop wandering.
Tactic: Give your market exploration an explicit deadline, and if an external shock arrives first, use it as the forcing function to commit to one segment.
Principle 4
Move the complexity onto yourself
One clip from Biswas's How Leaders Lead conversation carries the title "Make it simple for the customer, even if it's hard for you," and it might be the master principle of his career. Roofnet's nodes configured themselves because no one would climb onto a roof to fix them. Meraki's Mini worked in Musgrave without technicians. The 2009 cloud dashboard turned the 1 a.m. drive to the office into a browser tab. Samsara compresses AI models to run on two-to-ten-watt devices so that a fleet manager never has to think about inference.
Each of these was an engineering burden assumed by the company so the user wouldn't have to carry it. The commercial effect compounds: products that are easy to deploy spread by word of mouth—more than 1,000 Meraki networks before the company advertised at all—and products that are easy to manage get expanded. The hidden difficulty becomes the moat, because competitors would have to absorb the same pain to match the experience.
M
Where the Complexity Went
Each generation of Biswas's products moved the hard part away from the user.
| Before | Biswas's approach | Who absorbs the complexity |
|---|---|---|
| Hub-and-spoke networks with a central tower | Self-configuring mesh nodes that route around failure | Protocol |
| $1,500 research routers | $50 Mini built on commodity hardware | Software |
| On-site, box-by-box configuration and the 1 a.m. call | Multi-tenant cloud dashboard in a browser | Cloud platform |
| Raw sensor data overload for operators | Edge AI distilling streams into actionable coaching | Edge models |
Tactic: List every task your customer must perform to get value from your product, and assign an engineer to eliminate the most painful one each quarter.
Principle 5
Leave capital in the bank on purpose
Meraki raised about $80 million across four rounds and did not spend it all. After 2008, the company deliberately operated near cash-flow break-even, because, in Biswas's words, "We didn't know if the next round of capital would be there." All the while revenue doubled every year.
The discipline here is psychological as much as financial. Most founders treat raised capital as a budget to deploy; Biswas treated it as insurance. Break-even operations meant the company's survival never depended on a market it couldn't control, which in turn freed it to make decisions on product merit rather than fundraising optics. When Cisco came calling, Meraki was selling from strength, not need. Samsara reported consecutive quarters of GAAP profit as a public company—the same instinct, scaled.
Tactic: Plan your operating budget as if your next round will not close, and treat any capital beyond break-even as a reserve rather than a spending target.
Principle 6
Start the second company from the market, not the lab
Meraki was "PhD research from MIT put into a box," and it took years to find the right buyer. Samsara inverted the sequence: start from a market—physical operations—and learn it from scratch. "We've never driven a truck or a forklift or worked in a warehouse," Biswas admits. That ignorance was the starting condition, not an obstacle.
The insight is that technical founders tend to begin from capability, which produces the walk in the woods. Market-first founding puts the hard problem—what does someone urgently need?—at the front, and lets technology follow. It also guards against the second-time founder's trap. Biswas warned against attempting a startup in "easy mode" and against letting a past win "taint" the approach to the next. Pattern recognition, yes. Assumption of expertise, no.
⇄
Two Companies, Two Directions
The same founders, opposite starting points.
| Dimension | Meraki (2006) | Samsara (2015) |
|---|---|---|
| Starting point | Technology first Roofnet research | Market first Physical operations |
| Founders' domain knowledge | Deep experts in wireless networking | "Didn't know the first thing about it" |
| Early funding | Customer presales, then Sequoia and Google | Self-funded, then a16z-led Series A |
| Time to market focus | One to three years of wandering | Customer-defined from the outset |
| Endgame | Acquired by Cisco for $1.2B | Independent Public as IOT |
Tactic: Before writing a line of code for your next venture, spend weeks inside the target customer's daily work until you can describe their worst problem in their own words.
Principle 7
Put the engineers on the phones
When Meraki needed to prove its market, the founders did inside sales themselves. John Bicket "literally put on a suit," Biswas recalls, "like a method actor," and the two engineers worked through call lists. Biswas argues that technical founders hold an unfair advantage in sales: they can discuss the product deeply and they hear, firsthand, what to build next.
This is not about founders being scrappy. It is about information fidelity. A hired salesperson filters what customers say through quota and pitch; a founder hears the raw signal and can redesign the product on the drive home. Biswas has spoken of learning to love sales as a technical founder—and of the Samsara go-to-market requiring thousands of people and meticulous execution. The founders' phone calls are the template that the thousands later scale.
Tactic: Have every technical founder personally run at least twenty customer sales conversations before hiring the first salesperson, and log what each call reveals about the product.
Principle 8
Build the feedback loop into the product
Meraki's dashboard included a "Make a wish" feature that sent customer requests straight to the builders. Samsara holds an annual conference, Beyond, where the company spends a full day workshopping prototypes with customers in small groups, and it runs both formal and informal loops throughout the year. "The great thing about this is as your customer base grows, the amount of feedback you get grows," Biswas said.
The design choice here is subtle: feedback is not a department but a structural feature, and it scales automatically with the business. Biswas pairs this with an 80/20 filter—build what matters to the majority of a segment, and route the long-tail requests to partners or an API. Without that filter, a large feedback stream becomes noise; with it, the stream becomes a roadmap. "Customer feedback loops," as he told a Stanford class, are "easier to course correct than guess perfectly."
Tactic: Add a direct request channel inside your product, review the submissions monthly for patterns, and build only what recurs across the majority of a customer segment.
Principle 9
Allocate like a portfolio: 70-20-10
Samsara divides R&D effort on a 70-20-10 rule: seventy percent to improving the core products and platform, twenty percent to near-term upgrades the team is confident will matter, ten percent to ambitious long-term bets. The ten percent is explicitly expected to produce failures; the goal is to plant seeds, some of which graduate into the twenty.
What distinguishes this from generic innovation-budget talk is the graduation mechanism. Ideas earn their way up the stack through evidence, which keeps the moonshots honest and the core business protected. It also institutionalizes the walk in the woods without letting it consume the company—a lesson clearly carried from Meraki's open-ended exploration years. Exploration gets a fixed, bounded share.
Tactic: Tag every R&D project as core, near-term, or long-shot, and enforce a ceiling on the long-shot share while setting criteria for promotion.
Principle 10
Rehire yourself every year
Each year Biswas conducts a 360-degree review on himself and asks what the company needs most from him next. He audits his calendar, color-coded and now automated, comparing intended time allocation against reality. He calls the practice rehiring himself.
In hypergrowth the CEO's job changes faster than the CEO notices. The person who worked call lists in 2009 cannot be the person who signs a $23 million equity grant for a president of field operations in 2023. Rehiring is a mechanism for forcing that recognition before the board or the market does. The calendar audit is the quiet genius of it: intentions are cheap, but time spent is evidence, and the gap between them is where misallocation hides.
Tactic: Once a year, write the job description your company needs from its leader now, compare it honestly to how you actually spent the last quarter, and change your calendar before you change anything else.
Principle 11
Build as if no one is coming to buy you
Meraki had a natural acquirer in view; Samsara, by Biswas's account, does not, and its market is large enough to keep building indefinitely. That absence changed how the company decides. Without a defined end goal, Samsara weighs the long-term consequences of choices about technology, hiring, and culture, and runs more experiments before committing. One of Biswas's stated principles is that shortcuts today create problems tomorrow.
Founders often unconsciously build for the exit—optimizing for the metrics an acquirer will scan, deferring the expensive foundations. Building for permanence produces a different company: one that invests in platforms rather than features and in culture rather than headcount. Biswas's own observation that it takes years to make a big impact is the philosophical underpinning. The 2017 data loss at Meraki after the founders' departure is a cautionary aside about what centralized systems demand over decades.
Tactic: For every major decision, ask how it will look in ten years if the company is still independent, and reject options that only make sense in an acquisition scenario.
Principle 12
Let the customers do the pitching
Meraki's first 1,000 networks came through word of mouth, without advertising. When Samsara raised money, Biswas showed a working product, told customer stories, and told investors to call the customers themselves. "We didn't spend a lot of time pitching," he said. His definition of product-market fit is a customer calling a colleague in from the hallway to see the product.
The principle reframes persuasion as a byproduct of performance. Founders who spend most of their energy on narrative risk building a better pitch than product; Biswas spent his time "doing," and let the evidence travel through the network of customers—a mesh, not a tower. Leone setting up a network himself in minutes was the original version of this: the product made the argument.
Tactic: Replace one slide in your next investor meeting with a list of customer references, and invite the investor to call any of them before your second meeting.
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
— Sanjit Biswas, on the 20VC podcastCompanies pay to have their problem solved. And so when we talk about things like revenue and revenue growth, for me, it's actually impact and problem solving growth.
— Sanjit Biswas, on starting a second companyIf you try to do a startup in easy mode, you're probably not going to succeed. Don't let the Meraki win kind of taint the way that you approach the second one. And if anything, use it for pattern matching recognition.
— Sanjit Biswas, on Meraki's finances after 2008We didn't know if the next round of capital would be there. It was very important for us to be operating right around cash flow break-even.
— Sanjit Biswas, on the entrepreneurial leapPeople get very intimidated by the entrepreneurial process. But if you return to your roots as an engineer and say, 'Hey, I just like to build things,' whether it's products or companies, you can just give it a shot and experiment.
— Jim Goetz, Sequoia Capital, on investing in MerakiIf you've got a special team like this, they'll figure it out, so let's go.
Maxims
- Don't build a tower. Systems with a single center fail all at once; design so that the loss of any one part is routed around by the rest.
- A purchase order outranks a term sheet. Customers paying for an unfinished product is the strongest evidence a founder can carry into any room.
- Generality is a trap. A product that can do anything will be admired everywhere and bought nowhere until you choose a segment.
- Let the crisis end the wandering. Exploration needs a deadline, and an external shock is the best permission to commit.
- Absorb the complexity. Every hard thing you take off the customer's plate becomes a moat competitors must also pay to cross.
- Treat raised money as insurance. Running near break-even keeps the company's fate in its own hands.
- Start from the people, not the patent. A market-first company skips the years a technology-first company spends searching for its buyer.
- Make feedback structural. Build the request channel into the product and filter it by the 80/20 rule so volume becomes a roadmap.
- Rehire yourself. The leader the company needed last year is rarely the one it needs now; audit your calendar to find out.
- Build as if no one will buy you. Permanence changes every decision about platform, people, and culture—for the better.
