Somewhere inside Cisco Systems' headquarters in San Jose, in 2013, an internal presentation circulated among the company's sales force. Pasted at its center was a photograph of a woman's face, superimposed onto a bull's-eye pierced with arrows. The accompanying directive read: "Arm the field, stop the bleeding and fire back." The woman was Jayshree Ullal, who five years earlier had been feted at a farewell party in a packed Cisco ballroom by John Chambers himself — the CEO who had built Cisco into, briefly, the most valuable company on earth. Chambers had celebrated her ability to make complicated things simple and wished her well. He had not expected her to succeed quite this spectacularly.
By the time that bull's-eye was making the rounds, Ullal had already transformed a 50-person startup with negligible revenue into a company stealing marquee customers from Cisco's most profitable business line — the very switching division she had once run. Microsoft and Facebook were buying Arista's switches instead of Cisco's. The student had consumed the master's lunch, and the master was furious. Cisco would eventually sue Arista for patent infringement, launching a multiyear legal war that cast a shadow over what should have been a period of pure celebration. The lawsuit arrived almost simultaneously with Arista's IPO in June 2014, a timing that felt less like coincidence than like revenge.
This is the tension at the heart of Jayshree Ullal's career: the insider who became the insurgent, the loyalist who became the apostate, the executive who spent fifteen years building a $10 billion business inside a corporate giant and then walked away to build something better from nothing. It is a story about the difference between managing scale and creating it — and about how the people who understand a system's weaknesses most intimately are precisely the ones positioned to exploit them.
By the Numbers
Arista Networks Under Ullal
Part IIThe Playbook
Jayshree Ullal's career offers a set of operating principles that are both deeply specific to enterprise technology and broadly applicable to anyone building in competitive markets. What follows are the lessons distilled from her trajectory — from semiconductor engineer to the CEO of a $100 billion company.
Table of Contents
1.Learn the incumbent from the inside before you disrupt it.
2.Bet on architectural re-platforming, not incremental improvement.
3.Earn the right to lead by spanning the full stack.
4.Use patient capital to build what venture timelines won't allow.
5.Serve the customers who are building the future, not the ones maintaining the past.
6.Choose speed and conviction over completeness of data.
7.Treat litigation as validation, not distraction.
Build a culture of engineering-led leadership, not management-led engineering.
In Their Own Words
Building a good company takes constant hard work and heavy lifting. Making a great company is an even harder work-in-progress, demanding tenacity.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
Going from a good to great company takes a monumental series of changes that are not incremental but exponential.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
The ability to question the status quo of engrained habits and even oneself to navigate the right direction is vital.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
What destroys many companies is complacence, stifling innovation, denying new trends.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
An undisputed champion of an industry today can become stagnant in merely a few years.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
Short-term profits, growth, momentum stock price and exit strategies are rewarded instead of differentiated engineering investment and customer quality.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
A dedicated strategy and commitment to execute upon it defines courageous leadership with a culture of innovation and teamwork setting the tone.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
Sometimes you need that re-architecture or significant shift in direction.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
In the pursuit of greatness, the leader must be unassailable as a visionary and a person of action.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
Our industry and those who judge us emphasize the wrong metrics in the evolution to greatness.
— Good to Great: Hallmarks of High-Tech Leadership, 2017
Jayshree Vedantham was born on March 27, 1961, in London — not because her parents were British but because her father, Sudarajan Vedantham, was a physicist working on academic collaborations between Indian and British universities. Sudarajan was among the founding generation of faculty at the Indian Institutes of Technology, those post-independence temples of engineering that Jawaharlal Nehru built to forge a modern nation from the wreckage of colonial rule. That his daughter would eventually build one of America's most valuable networking companies carries a symmetry so neat it feels contrived, but it isn't. The IITs were themselves an act of infrastructural faith — the belief that a country could bootstrap its way to modernity by training the right people to build the right systems. Jayshree would spend her career doing something structurally identical, just for data instead of nations.
When she was five, the family moved back to New Delhi. She attended the Convent of Jesus and Mary, an all-girls school known for academic rigor and a quiet insistence on discipline that would prove more formative than any engineering curriculum. Her mother, Nirmala Vedantham, was a homemaker; her father, the physicist, the gravitational center around which the family's intellectual life orbited. At sixteen, when Sudarajan secured a new position in the United States, the family moved again — this time to California, a place that must have seemed as far from Delhi's heat and density as the moon.
She enrolled at San Francisco State University, choosing electrical engineering at a moment when women in American engineering programs were vanishingly rare. "While I was pursuing electrical engineering, I was only one or two of 100 female students in a class of 100," she would recall decades later at SFSU's commencement. "This made cutting class difficult, as we were conspicuous by our absence!" The joke masks something harder: the experience of being perpetually visible, perpetually other, in a field that rewarded blending in. She graduated in 1981 with a B.S. in electrical engineering, then earned a master's in engineering management from Santa Clara University in 1986. The combination — deep technical fluency married to managerial training — would become her signature.
The Semiconductor Years and the Discovery of Markets
Her early career reads like a methodical tour of the semiconductor industry's supply chain. She began at Fairchild Semiconductor, that legendary incubator of Silicon Valley DNA, as a senior strategic development engineer. Then Advanced Micro Devices, where she designed high-speed memory chips for IBM and Hitachi — work that was precise, deeply technical, and almost entirely invisible to anyone outside a fabrication lab.
But something shifted in the late 1980s. Ullal discovered she was interested not just in building chips but in understanding who bought them and why. She moved to Ungermann-Bass in 1988, a networking company pulling in $40 million annually from developing and selling networking products, where she became director of internetworking products. The role was her bridge between engineering and commerce — the first sign of a career pattern in which she would deliberately position herself at the seam between technical depth and market awareness.
Four years later, in 1992, she joined Crescendo Communications as vice president of marketing, working on 100-Mbit/s over copper, the first CDDI products, and first-generation Ethernet switching. The company was small, scrappy, and operating at the frontier of what networking could become. She was, at thirty-one, already doing the thing she would later counsel young engineers to do: "If you're in engineering, you might have to learn some business. If you're in computer science, you might want to learn something about data science and analytics." She was building herself into an unusual hybrid — the engineer who understood markets, the technologist who could sell.
Then Cisco came calling.
Fifteen Years Inside the Machine
In September 1993, Cisco Systems acquired Crescendo Communications. It was Cisco's first acquisition — an event so foundational to the company's history that it essentially defined Cisco's growth model for the next two decades. Crescendo gave Cisco its entry into the switching market, the business that would become its profit engine. And Ullal came with the acquisition, like a key component bundled into the deal.
She joined Cisco and began work on the Catalyst switching business. What followed was an extraordinary run: under her leadership, the Catalyst line grew from its inception in 1993 to a $5 billion business by 2000. She was, by any measure, one of the most productive executives in Cisco's history. As vice president and general manager of LAN switching in the Enterprise group, she oversaw unified communications, IP telephony, content networking, and policy networking. She managed approximately twenty mergers and acquisitions in the enterprise sector — meaning she didn't just run the switching business, she built it through a relentless program of corporate acquisition that was, in effect, Cisco's primary growth strategy.
Eventually, she was named Senior Vice President of Data Center and Switching, reporting directly to John Chambers. The scope was staggering: the modular Nexus and Catalyst Data Center Switching and Application/Virtualization services, collectively generating roughly $15 billion in direct and indirect revenue. She was, by the time of her departure, responsible for more revenue than most Fortune 500 companies generate in total.
Chambers, a soft-spoken West Virginian whose gracious manner skewed more senatorial than Silicon Valley, and Ullal made a strong team — partly, former executives who worked with them said, because they were both extremely competitive. But their styles clashed in revealing ways. Chambers was a managerial guru and a salesman. Ullal was an outspoken engineering and marketing whiz who disliked rigid rules. In a company that had grown enormous through acquisition and process, she was the person who kept asking whether the processes still made sense.
I had gained a lot of experience and growth at Cisco, where I had been for 15 years. But the company had grown from a billion dollars to over 40 billion by the time I left, so it had become a very large company.
— Jayshree Ullal
The subtext is unmistakable. Cisco had become bureaucratic. The middle managers had multiplied. The company that had once moved with startup velocity was now moving with the caution of an institution protecting its position rather than extending it. Ullal would later write, in one of her Arista blog posts, about precisely this failure mode: "What destroys many companies is complacence, stifling innovation, denying new trends, blinding people with overconfidence and even arrogance. Most corporate leaders create middle managers for a steady and stable organization. There is nothing wrong with that, except that they often rely on middle managers excessively, whose primary jobs are preventing immediate problems."
She was describing Cisco. She was also explaining why she left.
The Call from the Founders
Andy Bechtolsheim is one of those figures who keep appearing at the origin points of Silicon Valley's most consequential companies, like a recurring character in a creation myth. Born in Germany, trained at Carnegie Mellon and Stanford, he co-founded Sun Microsystems in the early 1980s and designed the workstations that powered an entire generation of computing. He wrote the first check to Google — a $100,000 investment made out to "Google Inc." before the company was even incorporated. His co-founder at Arista, David Cheriton, was a Stanford professor whose research on distributed systems and networking laid foundational aspects of modern data communication. Cheriton had also invested early in Google and, like Bechtolsheim, possessed the quiet confidence of someone who had been right about enough things that being right no longer surprised him.
Together, in 2004, they founded Arista Networks — originally called Arastra, Inc. — with a vision for high-performance Ethernet switches built on a fundamentally different software architecture. They had the technical vision. They had the capital. What they needed was an operator who understood both the technology and the market, someone who had spent years inside the incumbent's belly and knew exactly where the soft tissue was.
In October 2008, Bechtolsheim and Cheriton appointed Ullal CEO and President. She had known both of them for years — Bechtolsheim had been her customer at Sun when she worked at AMD, and they had crossed paths again when Cisco acquired his company, Granite Systems. "The beauty of knowing the people you work with and creating a professional family is probably the biggest attraction," she would later say.
What she found at Arista was a team of roughly thirty engineers working on good technology with no revenue and no clear path to market. The company was four years old and still pre-commercial. The Extensible Operating System — EOS, Arista's foundational software platform — had been under development since 2004 and wouldn't see meaningful returns for five years. "No venture-capital funded company would have been given the leeway to take the route we took," Ullal later acknowledged. The company's unique financing structure — backed by Bechtolsheim's and Cheriton's personal fortunes, along with investors like Sequoia Capital, Accel, and Kleiner Perkins — gave it the luxury of a long runway that most startups never enjoy.
She didn't join Arista to manage. She joined to build.
The Architecture of Disruption
The technical insight that animated Arista was deceptively simple: the networking industry's operating systems were obsolete. Cisco's IOS and its successors were monolithic software stacks — massive, intertwined codebases where a single bug could crash the entire system. They were the legacy of an era when networking hardware and software were inseparable, when the switch was the network and the network was the switch.
Arista's EOS was built differently. It ran on top of a standard Linux kernel, with each process operating independently so that a failure in one wouldn't bring down the others. It was programmable, extensible, and open — designed from the ground up for the world of cloud computing that was just beginning to materialize. In 2008, when Ullal joined, the word "cloud" was barely in use. "I don't even think we mentioned the word 'cloud networking' when I joined Arista," she later recalled. "I remember writing the first blog on that topic in October of 2008, but the word 'cloud' was barely even used at that point. Many industry luminaries were saying, 'What's a cloud? That's just a marketing term.'"
It wasn't a marketing term. It was the future. And EOS was built for it.
The brilliance of Ullal's positioning was that she understood, from fifteen years inside Cisco, exactly what the cloud titans — Microsoft Azure, Amazon AWS, Google Cloud — needed and exactly what Cisco couldn't give them. These hyperscalers were building data centers of unprecedented scale, and they needed switches that were fast, programmable, and operated on open standards rather than proprietary lock-in. Cisco's model was to sell you the hardware and the software as an inseparable bundle, creating dependency. Arista's model was to give you switches that worked with everything — open-source tools, third-party controllers, whatever the customer wanted.
The early adopters told the story. Major cloud service providers and top financial institutions — organizations like JPMorgan Chase, Goldman Sachs, and Citigroup that needed high-speed, low-latency data transmission — began adopting Arista's 10 Gigabit Ethernet switches. The financials, in particular, were a natural fit: high-frequency trading demanded the kind of microsecond-level latency that Arista's switches could deliver and Cisco's couldn't match.
She took slightly more than an engineering team doing some good technology and turned it into the thriving network switch company it is today.
— David Cheriton, Arista co-founder
The IPO and the Lawsuit
On June 6, 2014, Arista Networks went public on the New York Stock Exchange under the symbol ANET, with an initial offering price that valued the company at approximately $4.7 billion. Ullal stood on the trading floor and rang the bell. The S-1 filing listed her as the principal executive officer, headquartered at 5453 Great America Parkway, Santa Clara, California — an address within striking distance of Cisco's own campus.
Then Cisco sued.
The timing was, depending on your perspective, either strategically calculated or petty. Cisco filed a patent infringement lawsuit against Arista in December 2014, alleging that Arista had stolen its technology — specifically, that Arista's command-line interface and certain networking protocols were derived from Cisco's intellectual property. The lawsuit cast a shadow over what should have been Arista's moment of triumph.
Ullal has spoken about this period with characteristic directness. "We were celebrating and clapping the fact that we had a successful IPO," she recalled, "and then Cisco comes and casts a shadow over it." The legal battle would drag on for years, with each side notching incremental victories and no sign of resolution. Arista denied the allegations, arguing that Cisco had sued not because its technology had been stolen but because it lacked the innovation to compete. Cisco countered that Arista's product was fundamentally built on intellectual property developed at Cisco — a claim that carried particular sting given that Ullal and several other Arista employees were Cisco veterans.
The bull's-eye image from the 2013 internal presentation was eventually revealed in court proceedings, along with the directive to "stop the bleeding and fire back." John Chambers, in court testimony, admitted that Cisco had been "too slow to react to a fast-changing market." It was a remarkable concession from the man who had built Cisco into a colossus — an admission that the company he led had been outmaneuvered by a startup run by one of his own former executives.
The feud had become personal. Chambers couldn't stand to lose sales, especially to someone he considered family. Defeating Arista became a priority for Cisco, a company more than forty times bigger by annual revenue. Chuck Robbins, who succeeded Chambers as CEO, told a gathering of 28,000 partners and customers that Cisco was on "a journey to change everything." But the damage had been done. The cloud titans had tasted something better, and they weren't going back.
The Quiet Operator
There is a particular kind of CEO who generates headlines through provocation — the Musk, the Ellison, the personality-as-brand model. Ullal is the opposite. She is, by her own description, "a very shy, quiet introvert" — a self-characterization that seems improbable for someone who has run one of Silicon Valley's most aggressive companies for seventeen years but that multiple people who know her confirm. On Hacker News, in a thread debating the eccentricities of tech billionaires, one commenter listed Ullal alongside Eric Schmidt and Alex Karp as examples of enormously wealthy people who generate almost no gossip. "There are plenty of people who you don't hear much about," the commenter wrote, "but because you don't hear much about them, the other ones tend to take up all of the air in the room."
This quietness is strategic. Ullal's leadership philosophy, as articulated across years of blog posts and interviews, is almost ostentatiously un-flashy. She writes about "deciding with speed and conviction," about the importance of admitting when you've been wrong, about building a foundation rather than chasing short-term revenue. "Most decisions can be iterated as you discover more," she has written. "One guiding principle that motivates my decisions is the correlation of customers and employees with my own instinct and experience."
What she does not do is perform leadership. There are no viral Twitter threads, no podcast tours, no carefully cultivated founder mythology. She runs the company, writes technical blog posts about Ethernet standards and AI networking architectures, and lets the results speak. In a fireside chat with Satya Nadella, when Ullal brought up the topic of leadership, Nadella responded that he finds people like her — "who have created massive organisations out of nothing" — inspirational. The compliment is revealing for what it implies: even the CEO of Microsoft, who runs a company worth trillions, sees Ullal's zero-to-something trajectory as a different and arguably harder feat than managing an inherited empire.
Her public persona, such as it is, tends to emerge in commencement addresses. At San Francisco State University in May 2023, she told graduates that the institution "can be your rock just like it's my foundational rock." At Santa Clara University's 2025 commencement — where she received an honorary Doctor of Engineering degree — she encouraged graduates to "trust your conscience." These are not the remarks of someone constructing a personal brand. They are the remarks of someone who genuinely believes the institution matters more than the individual.
Building for the AI Moment
If Arista's first act was cloud networking and its second was enterprise campus, its third — and potentially most consequential — is AI infrastructure. The explosion of generative AI, driven by companies training and deploying massive language models, has created an insatiable demand for the kind of high-performance, low-latency networking that Arista has spent two decades perfecting.
The math is straightforward. Training a large language model requires thousands of GPUs working in concert, and those GPUs need to communicate with each other at extraordinary speeds. The network connecting them is not a commodity — it is the bottleneck. If the GPUs are the engines, the network is the nervous system, and a slow nervous system renders even the fastest engines useless. Arista's switches, built on EOS and designed for exactly this kind of scale, are positioned at the center of this infrastructure buildout.
In a conversation with Nicolai Tangen, CEO of the Norwegian sovereign wealth fund, Ullal explained that AI traffic is "fundamentally different from anything that came before" — and that power, not hardware, has become the biggest constraint on AI deployment. This is a characteristically Ullal observation: technical, precise, and focused on the actual limiting factor rather than the one that generates the most excitement.
The company's revenue trajectory reflects this positioning. Arista reported approximately $7 billion in revenue in 2024, a 20 percent year-over-year increase. Its operating margins remain among the highest in the networking industry. And in June 2025, Arista hired Todd Nightingale — previously CEO of Fastly and, before that, the executive vice president who ran Cisco's enterprise networking and cloud division, including Cisco Meraki — as its new president and chief operating officer. The hire was a signal: Arista was preparing for its next phase of growth, and Ullal was building the team to execute it.
Clearly, AI and campus are going to grow and do great guns for us, as it should, because they're two very large TAMs. These two are going to grow substantially in double digits.
— Jayshree Ullal
The Wealth and Its Discontents
By December 2025, Forbes estimated Jayshree Ullal's net worth at $5.7 billion, making her the richest Indian-origin professional manager in the world — ahead of Satya Nadella (approximately $1.1 billion) and Sundar Pichai (approximately $1.5 billion). The comparison is instructive. Nadella and Pichai are hired managers of companies they did not create; their wealth comes from compensation packages. Ullal's wealth comes from equity — approximately 3 percent of Arista's outstanding shares, a stake she has held and built since the company's earliest days. The distinction matters. Equity wealth is founder's wealth. It is the wealth of someone who bet on the thing she was building and was right.
She and her husband, Vijay Ullal — himself a venture capitalist and former president and chief operating officer of Fairchild Semiconductor — live in Saratoga, California, with their two daughters, Adeeti and Tarini. They co-founded a charitable foundation in honor of Ullal's late sister, Susie Nagpal, a Saratoga City Councilwoman who died of lung cancer in 2010. The foundation supports cancer research, social welfare, and hunger relief in India. Some of Ullal's Arista shares are set aside for her two children, niece, and nephew.
The wealth has not changed the public presentation. There are no reports of art collections, mega-yachts, or island purchases. In the Hacker News thread about billionaire eccentricities, Ullal was cited specifically as an example of someone about whom there is nothing weird to report. This absence of spectacle is, in Silicon Valley's current climate, its own kind of radical act.
The Cisco Shadow and the Arista Light
The relationship between Arista and Cisco remains one of the defining competitive dynamics in enterprise networking. In 2018, Arista's operating margin reached 31.5 percent, eclipsing Cisco's 28 percent — a remarkable inversion for a company that was, by revenue, a fraction of Cisco's size. By 2019, even in what Arista called a "down" quarter — one in which a major "cloud titan," likely Facebook, reduced its orders — revenues still grew 16 percent year over year to $654 million. Fortune named Ullal its Businessperson of the Year.
The pattern is consistent. Arista wins not by matching Cisco's scale but by being fundamentally better at the things that matter most to the customers who matter most. The cloud hyperscalers — Microsoft, Meta, Amazon — need networking that is open, programmable, and fast. Cisco's model, built on proprietary ecosystems and decades of legacy architecture, cannot offer that at the same level. And Arista's EOS, designed from scratch for this exact use case, can.
Cisco's response has been to pivot toward automation and security, acknowledging implicitly that it lost the data center switching war. Chuck Robbins, who succeeded Chambers, launched a new line of automated and programmable switches in 2016 and told partners the company was on "a journey to change everything." Cisco's revenue fell for seven consecutive quarters during this transition. The company that had once been the world's most valuable was now scrambling to reinvent itself in the image of the insurgent it had tried to destroy.
Ullal, characteristically, does not gloat. She writes about market transitions and the importance of "questioning the status quo of engrained habits." She counsels against "prolonging decisions or being stuck in analysis paralysis mode." But the subtext is always there: she built the thing that Cisco couldn't, and she built it because she understood, from the inside, exactly what Cisco had wrong.
The Introvert's Empire
In November 2025, Ullal sat down with Rene Haas, CEO of Arm, for the Tech Unheard podcast. Haas introduced her with a recitation of her career history — Fairchild, AMD, Cisco, Arista — and Ullal laughed at the description of the Arm office being far away. "Well, it wasn't that far. It's a mile away."
The exchange captures something essential about Ullal: she is perpetually deflating grandeur, returning the conversation to the proximate, the specific, the concrete. She does not speak in abstractions about disruption or transformation. She speaks about EOS architecture, about Ethernet standards, about the specific technical characteristics of AI network traffic. In her blog posts — she has written them regularly for Arista since 2008 — she discusses EVPN, VXLAN, zero-trust networking, and scale-up Ethernet with the fluency of someone who still reads the technical documentation.
This is unusual for a CEO of a company worth over $100 billion. Most executives at this level have long since delegated the technical details. Ullal has not. She remains, after four decades, the engineer who learned business rather than the businessperson who learned to cite engineering. The distinction is real and it matters, because Arista's competitive advantage is, at its core, a software advantage — and maintaining that advantage requires a CEO who can evaluate technical architecture at the deepest level.
She also serves on the board of directors of Snowflake, the cloud computing company that went public in September 2020 in one of the year's largest IPOs. The board seat is notable less for its prestige than for what it reveals about how Ullal thinks: Snowflake is a company built on the same open, cloud-native principles that animate Arista, and Ullal's presence on its board suggests a strategic attention to the broader cloud ecosystem that extends well beyond networking.
🔄
From Incumbent to Insurgent
Ullal's career trajectory between Cisco and Arista
Arista revenue exceeds $7 billion; market cap surpasses $100 billion
2025
Todd Nightingale (ex-Cisco Meraki, ex-Fastly CEO) hired as President & COO
A Rock, Not a Rocket
At SFSU's 122nd commencement in May 2023, standing before more than 31,000 people at Oracle Park, Ullal told the graduates about arriving in the United States from India in 1977. She talked about being one of two women in a class of a hundred. She talked about being shy. She talked about the professors and students who supported her. And then she offered this: "This great San Francisco State institution shaped me and guided my future. And it can be your rock just like it's my foundational rock."
A rock. Not a launchpad, not a rocket, not a catapult — a rock. The metaphor is revealing. Ullal does not describe herself as having been launched or catapulted or accelerated. She describes herself as having been grounded. The rock is what you stand on; it is what keeps you from being swept away. It is the thing that remains when everything else shifts.
Two years later, at Santa Clara University's 174th commencement on June 14, 2025, where she received an honorary Doctorate of Engineering, she told graduates to "trust your conscience." The valedictorian that day, Grace Davis, spoke about the tension between pride and humility, drawing on Aristotle and Aquinas. It was the kind of philosophical framework that Ullal herself might resist — she is not a philosopher but an engineer — and yet her career is itself a resolution of precisely that tension. She has built one of the most valuable companies in technology while maintaining a public profile so restrained that a Hacker News commenter could cite her as an example of a billionaire about whom there is nothing interesting to say.
But there is something interesting to say. It is just not the kind of interesting that generates viral content. It is the interesting of a woman who designed memory chips for IBM in her twenties, who grew a switching business from zero to $5 billion in her thirties, who walked away from $15 billion in annual revenue in her forties, who built a hundred-billion-dollar company in her fifties and sixties — and who, through all of it, kept writing blog posts about Ethernet standards. Who described herself as a "very shy, quiet introvert" while running one of the most aggressive competitive campaigns in the history of enterprise technology. Who received a bull's-eye on her face from the world's most powerful networking company and responded not with rhetoric but with revenue.
In Saratoga, California, in a house she shares with her husband and daughters, within a mile of the Arm headquarters and a short drive from Arista's campus on Great America Parkway, Jayshree Ullal continues to run the company she has led for seventeen years. The AI infrastructure boom is accelerating. The cloud titans are spending billions. Power, not hardware, is the constraint. And the network — the thing she has spent her entire career building, optimizing, and reimagining — is, once again, the bottleneck through which everything must pass.
8.
9.Stay small in identity even as you grow large in scale.
10.Let the product be the brand.
11.Position for the next transition before the current one peaks.
12.Compound your advantages across multiple cycles.
Principle 1
Learn the incumbent from the inside before you disrupt it
Ullal's fifteen years at Cisco were not wasted time — they were reconnaissance. She learned how Cisco's switching business worked at the deepest level: the product architecture, the sales motion, the customer relationships, the organizational pathologies. She oversaw twenty mergers and acquisitions, which means she understood not just how Cisco built products but how it consumed companies. When she joined Arista, she carried a mental model of every weakness in Cisco's approach — the monolithic software, the proprietary lock-in, the bureaucratic decision-making — and could build against each one with surgical precision.
The lesson is not simply "work at the incumbent before starting a competitor." It is subtler: the most dangerous insurgents are the ones who loved the incumbent enough to spend years understanding it, and left not out of failure but out of conviction that they could do better.
Tactic: Before disrupting a market, invest years understanding its dominant player from the inside — not as an outsider analyzing from afar, but as an operator who has lived within its decision-making architecture.
Principle 2
Bet on architectural re-platforming, not incremental improvement
Arista did not build a slightly better Cisco switch. It built an entirely different operating system — EOS — on top of a standard Linux kernel, with a modular architecture that allowed independent process operation, open programmability, and cloud-native design. This was not a feature improvement but an architectural paradigm shift. Cisco could not match it without abandoning its own legacy codebase, which meant that Cisco's competitive response was structurally constrained.
Ullal has written about this in her blog posts: "What destroys many companies is complacence, stifling innovation, denying new trends." The corollary is that what creates great companies is the willingness to build from a blank slate when the existing architecture has hit its limits. Arista spent five years — from 2004 to roughly 2009 — developing EOS before seeing meaningful commercial returns. Most companies cannot tolerate that timeline. The ones that can build something very difficult to replicate.
Tactic: When entering a market dominated by incumbents, don't optimize within their architectural constraints — rebuild from first principles, even if it takes years to reach commercial viability.
Principle 3
Earn the right to lead by spanning the full stack
Ullal's career arc — from chip designer to marketing VP to SVP running a $15 billion business to CEO of a startup — is unusual in its breadth. She has designed semiconductors, run product marketing, managed enterprise sales, overseen M&A, and led a company from zero to $100 billion in market cap. This span gives her a form of authority that specialists cannot replicate: she can evaluate an engineering decision, a go-to-market strategy, and a financial forecast with equal fluency.
She has counseled this approach explicitly: "In today's world, interdisciplinary fields are more important than ever. If you're in engineering, you might have to learn some business. If you're in computer science, you might want to learn something about data science." The recommendation is not vague self-improvement advice — it is a specific prescription for building the kind of credibility that allows you to lead technical organizations without losing the respect of engineers.
Tactic: Deliberately rotate through technical, commercial, and strategic roles early in your career to build the cross-functional fluency that allows you to lead with genuine authority in each domain.
Principle 4
Use patient capital to build what venture timelines won't allow
Arista's unique financing structure — backed initially by Bechtolsheim's and Cheriton's personal fortunes, supplemented by investors like Sequoia Capital, Accel, and Kleiner Perkins — gave the company something venture-backed startups almost never have: time. EOS took five years to develop before generating meaningful revenue. Ullal herself acknowledged that "no venture-capital funded company would have been given the leeway to take the route we took."
This is a structural insight about innovation timelines. The deepest platform plays — the ones that create durable competitive advantage — often require development periods that exceed standard VC fund cycles. If your technical ambition requires five or more years of building before commercialization, you need a capital structure that can tolerate that timeline. Arista had it because its founders were wealthy enough to fund it themselves.
💰
Capital Structure as Competitive Advantage
Arista's patient funding model versus typical VC timelines
Dimension
Typical VC-Backed Startup
Arista's Model
Time to commercialization tolerance
18–36 months
5+ years
Primary capital source
Institutional VCs
Founder wealth + strategic VCs
Pressure to pivot or exit
High
Minimal
Architectural ambition
Constrained by runway
Unconstrained
Tactic: If your competitive advantage requires a development timeline that exceeds standard venture cycles, structure your capital around patient investors — ideally founders or strategics — who can tolerate a multi-year build before commercial traction.
Principle 5
Serve the customers who are building the future, not the ones maintaining the past
Arista's early customer base — cloud hyperscalers and high-frequency trading firms — was not the largest addressable market in networking. That belonged to traditional enterprise IT departments, which were buying Cisco. But the hyperscalers and trading firms were the customers whose needs most precisely anticipated where the entire industry was heading: massive scale, programmability, open standards, and extreme performance.
By serving these customers first, Arista built products that were architecturally prepared for the market that would eventually emerge. When the broader enterprise began moving to cloud, Arista's products were already proven at the most demanding scale. The strategy inverted the conventional wisdom of starting with the broad market and working up — instead, Arista started with the most demanding customers and worked down.
Tactic: Identify the customers whose current needs most closely match the future state of your market — even if they represent a small initial segment — and build for them first, because the products they require will be the ones everyone eventually wants.
Principle 6
Choose speed and conviction over completeness of data
Ullal has written explicitly about this: "Despite the data-driven era we live in, good leaders must make decisions without all the necessary data and be comfortable with some ambiguity. A clear, crisp extrapolation from available data to a decision is better than sitting on the fence in a prolonged manner." And the corollary: "To admit that it is okay to be wrong and course-correct is the true sign of a mature leader."
This is not an argument against data. It is an argument against data paralysis — the organizational pathology in which the quest for perfect information becomes an excuse for inaction. Arista's decisive moves — entering the cloud networking market before the term existed, pivoting from point products to multiple "places in the cloud" networking in 2014, positioning aggressively for AI infrastructure — were all made under uncertainty. The key was not having complete data but having the judgment to extrapolate from incomplete data and the courage to act on that extrapolation.
Tactic: Set a threshold for decision-making at 70 percent confidence rather than 95 percent, and build organizational tolerance for course-correction as a normal part of execution rather than an admission of failure.
Principle 7
Treat litigation as validation, not distraction
When Cisco sued Arista in December 2014, it could have been devastating. The lawsuit arrived at the worst possible moment — immediately after the IPO, when the company was most vulnerable to reputational damage. Many companies would have settled quickly to make the problem go away. Arista fought. For years. Through multiple courts. Each side notched incremental wins, and eventually Arista emerged with its business not just intact but stronger.
Ullal understood something that many first-time founders do not: being sued by the incumbent is a signal that you've succeeded, not that you've failed. Cisco didn't sue Arista because Arista was irrelevant. Cisco sued because Arista was threatening its most important business. The lawsuit was, in effect, a competitive validation — proof that the insurgent had drawn blood.
Tactic: When an incumbent attacks through legal or competitive channels, reframe the attack internally as evidence of product-market fit rather than existential threat — and invest in legal defense as a strategic function, not a cost center.
Principle 8
Build a culture of engineering-led leadership, not management-led engineering
Ullal has been explicit about her disdain for the middle-management model that dominates most large technology companies. "Most corporate leaders create middle managers for a steady and stable organization," she has written. "There is nothing wrong with that, except that they often rely on middle managers excessively, whose primary jobs are preventing immediate problems. Middle managers are designed to avoid tactical mistakes."
Arista's organizational model is the opposite: flat, engineering-centric, and designed for direct communication rather than hierarchical filtering. The founding team — Bechtolsheim as chief development officer, Kenneth Duda as CTO — remained embedded in the technical work. Ullal herself continues to write detailed technical blog posts about networking architectures. The company's culture privileges the ability to understand and build technology over the ability to manage processes around technology.
Tactic: Structure your organization so that the people closest to the technology have the most influence over product decisions — and resist the institutional creep of management layers that filter information and slow decision-making.
Principle 9
Stay small in identity even as you grow large in scale
One of the most striking aspects of Ullal's leadership is how consistently she frames Arista as a startup, even as its market capitalization has exceeded $100 billion. She told the American Bazaar that "one of the huge attractions of Arista was that it was small and entrepreneurial." She has written about wanting to return to her "entrepreneurial roots." And the company's operational culture — direct communication, minimal bureaucracy, engineering-first decision-making — reflects this orientation.
This is not affectation. It is a deliberate strategic choice. Companies that begin to think of themselves as large inevitably begin to behave as large: they add process, they slow decision-making, they optimize for risk avoidance rather than risk-taking. By maintaining a startup identity, Ullal keeps the organization oriented toward the behaviors that created its success in the first place.
Tactic: Regardless of your company's actual size, maintain the cultural practices of a startup — flat hierarchies, direct communication, fast decision-making — by actively resisting the organizational complexity that scale naturally introduces.
Principle 10
Let the product be the brand
Ullal's public profile is almost comically modest for someone leading a $100 billion company. No viral social media presence. No personal brand strategy. No keynote circuit. She writes blog posts about Ethernet standards and gives commencement addresses at her alma maters. In an era when CEO celebrity is treated as a competitive asset, Ullal's approach is the inverse: the product is the brand, and the CEO's job is to make the product so good that it sells itself.
This works because Arista's customers — cloud hyperscalers, financial institutions, enterprise IT departments — do not make purchasing decisions based on CEO charisma. They make them based on performance, reliability, and total cost of ownership. In markets where the buyer is technical and sophisticated, CEO celebrity is not just unnecessary but potentially counterproductive, because it signals that the company is compensating for product weakness with marketing.
Tactic: In B2B markets with sophisticated buyers, invest your personal brand equity in technical credibility and product excellence rather than media visibility — let customer outcomes be your marketing.
Principle 11
Position for the next transition before the current one peaks
Arista's entry into AI networking infrastructure is not reactive — it is the latest iteration of a pattern that has defined the company since its founding. Arista was built for cloud before the cloud existed. It expanded into campus networking before that market fully opened. And it began positioning for AI infrastructure before the current AI boom reached its peak.
Ullal has articulated this principle in terms of "turning points": "Every technology company must navigate its phases of success and the first phase rarely resembles the next. Next generation leadership needs thinkers who can identify the risks, find the hidden jewels and build upon them." The discipline is to begin investing in the next transition while the current business is still growing — which requires the willingness to cannibalize present revenue in favor of future positioning.
Tactic: Allocate a meaningful percentage of R&D investment to technologies that address the market you believe will exist in three to five years, even when that investment competes with resources for your current growth engine.
Principle 12
Compound your advantages across multiple cycles
The deepest lesson of Ullal's career is not about any single decision but about the cumulative effect of consistent, compounding advantage. EOS, built once in 2004, continues to power every Arista product twenty years later. The relationships with cloud hyperscalers, built in the company's earliest days, continue to drive the majority of its revenue. The technical culture, established when the company had thirty people, persists at scale. The CEO herself — Ullal has led the company for seventeen years — provides continuity that most technology companies lack.
This compounding effect is the source of Arista's durability. Each advantage reinforces the others: the software platform attracts the best engineers, who build the best products, which win the most demanding customers, whose requirements push the platform further ahead. Breaking into this cycle would require a competitor to simultaneously match Arista on software, talent, customer relationships, and institutional knowledge — a feat that no competitor has yet achieved.
Tactic: Build your competitive advantages as interlocking systems rather than isolated capabilities, so that each advantage compounds the others and the aggregate moat deepens over time rather than eroding.
Part IIIQuotes / Maxims
In her words
One of the huge attractions of Arista was that it was small and entrepreneurial, with only 30 engineers. Most importantly, it was people I knew. The beauty of knowing the people you work with and creating a professional family is probably the biggest attraction.
— Jayshree Ullal, on joining Arista
Despite the data-driven era we live in, good leaders must make decisions without all the necessary data and be comfortable with some ambiguity. A clear, crisp extrapolation from available data to a decision is better than sitting on the fence in a prolonged manner.
— Jayshree Ullal, on decision-making
What destroys many companies is complacence, stifling innovation, denying new trends, blinding people with overconfidence and even arrogance.
— Jayshree Ullal, on organizational failure
This great San Francisco State institution shaped me and guided my future. And it can be your rock just like it's my foundational rock.
— Jayshree Ullal, SFSU Commencement 2023
In today's world, interdisciplinary fields are more important than ever. If you're in engineering, you might have to learn some business. If you're in computer science, you might want to learn something about data science and analytics.
— Jayshree Ullal, on interdisciplinary careers
Maxims
The insider makes the best insurgent. The people who understand a system's weaknesses most intimately are the ones positioned to build something better — which is why the most dangerous competitors often come from inside the incumbent.
Architectural bets outlast feature bets. Incremental product improvements create temporary advantage; fundamental platform shifts create durable ones. Invest in the kind of replatforming that competitors cannot match without abandoning their own legacy.
Patient capital is a competitive weapon. The ability to invest for five years before commercial returns is a structural advantage that most competitors, constrained by VC fund cycles or quarterly earnings pressure, cannot replicate.
Serve the future's customers, not the past's. Build for the most demanding buyers in the emerging market, even if they represent a small initial segment — their requirements will become everyone's requirements.
Speed of decision beats completeness of data. Seventy percent confidence with immediate action beats ninety-five percent confidence with delayed action, provided you maintain the institutional humility to course-correct when you're wrong.
Litigation from the incumbent is a growth signal. If the dominant player sues you, it means you've drawn blood. Treat the lawsuit as competitive validation and fight it as a strategic investment.
The CEO should be the last person to stop reading the documentation. In engineering-driven companies, the leader's technical credibility is not a nice-to-have but the foundation of organizational authority. When the CEO stops understanding the product at a deep level, the culture begins to drift.
Identity scales slower than revenue. Maintain the cultural practices and self-image of a startup regardless of your actual size — the moment you start thinking like an incumbent, you start behaving like one.
Compounding advantages beat isolated ones. Build your moat as an interlocking system — software platform, talent, customer relationships, institutional knowledge — where each element reinforces the others and the aggregate deepens over time.
Let the work speak. In markets where the buyer is sophisticated, CEO celebrity is noise. The most durable brand in enterprise technology is a product that works better than the alternative, deployed by a company that delivers reliably, led by a person who doesn't need you to know her name.
