Ofir Ehrlich runs a $4 billion company that sells enterprises a way to keep and reuse their own data, on the argument that data is the one asset a competitor cannot buy from an AI provider.
Companies selling into the AI boom generally advertise themselves as AI companies. Ehrlich said he deliberately did the opposite, and then found the market had turned him into one anyway — while the capital spending going on around him began to look to him like the year 2000.
"The technology is insane and it's growing as fast as someone can do something. The only question is are the valuations right?"
He co-founded CloudEndure, sold it to Amazon Web Services, spent four more years running it inside the company, and has since put money into more than 60 startups as an angel investor.
I listened to the full interview so you can skip it. 42 minutes of audio, 24 minutes of reading.
Here are the 12 takeaways that matter.
👤 Guest: Ofir Ehrlich, Co-Founder and Chief Executive of Eon, a $4 billion data resilience and protection company, who earlier co-founded CloudEndure and sold it to Amazon Web Services, and invests in startups as an angel
🎙️ Host: Maggie McGrath, Senior Editor at Forbes, where she edits ForbesWomen
📰 Published: 9 September 2026
🔴 YouTube | 🟣 Apple Podcasts | 🔗 Episode page | ⏱️ 42 min | ✅ Time saved: 18 min
Key Takeaways
AI is real, and in his telling that is exactly why the capacity being built for it gets overbuilt
His comparison is the year 2000, when the internet was also real and the infrastructure built for it took years to use up
The companies he thinks are mispriced are not Anthropic or OpenAI but the ones riding behind them
In his account the label alone now earns a high valuation, especially in Silicon Valley
Most of the new independent AI data-center operators will end up consolidated, he said
They have to find electricity, buy scarce chips and build their own sales effort, all of which the established clouds already have
His whole company is a bet that a customer's own accumulated data is its only durable advantage
Every rival can buy the same models from Anthropic, OpenAI, Databricks or Google
His last company was profitable, never spent its $13M round, and he says that is why it left money on the table
Investors reward growth over early profitability, and he learned it after the sale rather than before
The question to ask an investor is not what company they want but what company they would tolerate
He treats fun as a management metric and asks his direct reports how much of it their teams are having
He will not hire a genius who is toxic to the team, however good the engineering
He invests in the team first and the market second, because nearly every startup pivots
More than 60 angel investments, now alongside Sequoia, Lightspeed, Greylock and Index
Silicon Valley now has a nickname for a trillion-dollar company: the Kilocorn
He puts Eon's addressable market above $100B and says the investment community's consensus is that he has the shot
1. Data Is the Only Edge Left
McGrath introduced Eon as a quiet powerhouse in data resilience, data protection and backup, and explained the word quiet: "And I say quiet because I think if I were to go out on Fifth Avenue and ask anyone on the street, have they heard of you? They'd probably say no." Ehrlich agreed the company is young, and described what it sells as the infrastructure that sits underneath the applications a company's own customers see.
His pitch is that a customer's accumulated data is the only asset its competitors cannot also go out and buy. Every company building on the big AI systems — he named Anthropic, OpenAI, Databricks and Google — is drawing on the same underlying material, he said, so the differentiator has to be the data the customer created itself over the years
Eon's job, in his description, is to gather all of the data across an organization and its history and make it secure, reliable, efficient and usable by whatever system the customer wants to point at it
"Taking your data, making it yours and making it usable and viable"
He said that is what has made the company appealing to the customers now using it, and that whatever technology arrives next, the data a company has built up over years is the most important asset it owns
2. Same Gaps, Any Vintage
Ehrlich named Chick-fil-A, SoFi and AlphaSense as customers he is free to talk about, and said some of the large Fortune-listed companies using Eon have not cleared him to use their names. McGrath picked at the range: she assumed, on stereotype and founding date, that a company as young as SoFi would have tighter data-protection practice than one as old as Chick-fil-A, and asked why chief technology officers were coming to him at all.
His buyers are broader than the CTO, he said — chief information officers, chief data officers, and now chief data and AI officers
The original reason companies bought was not AI but the move to the cloud, which he dates to around 2020 and 2021 at large companies. When everyone started working from home, he said, running a company's systems in someone else's data center suddenly made sense
Because the cloud is less fragmented than the on-premises world it replaced, the weaknesses repeat from customer to customer. "It doesn't matter which company, which type of companies you are, you will basically have the same gaps"
He said he started the company as, in his own framing, a crazy person building a company that was not an AI company in an AI world — and that the market then taught him that making data usable is exactly what AI systems need
AI is not the only reason a company comes to Eon, he said, but it is the compelling event that makes them act
3. Two Transitions at Once
The credential he leaned on for all of this is the last company: it built migration and disaster-recovery technology for companies moving into the cloud, and Amazon bought it. That gave him what he called a front-row seat as businesses that had worked the same way for 20 years moved into someone else's data center and had to change every habit they had.
He described AI as a second version of that same shift, running much faster. He said the AI shift "kind of resembles the transition that happened in 2020, but on steroids," and that it is bigger than the one before it: "It's way, way, way larger"
Asked whether the catalyst for Eon was his own experience with data or a read on where the world was going, he said it was the experience — the cloud first and AI second — and that the company's job is to help customers take advantage of both
The transition is not only technical: "So we're facing two transitions not only of technology but of skill sets and knowhows and it's new for everyone"
His sales argument is relevance rather than efficiency. "You read the news, you generate the news," he said, adding: "If you're not staying relevant working with the top technologies today, if you're not leveraging the real revolution around artificial intelligence, you're not going to be relevant even in next few years"
He said the reset cuts both ways — a company that was relevant three years ago has the opportunity now, and one that was not has a chance to become significant
He does not claim to know which technology matters next, and says the data argument survives either way. "And I'm saying the word AI, but maybe tomorrow it's quantum computing and in 3 days something else"
Asked whether being a non-AI company meant his own staff avoid AI tools, he said the opposite: "When we started, we said we're contrarians. We're leveraging AI, but we're not solving AI problems." Every problem they ended up solving turned out to be AI-related, and he said the company then immersed itself in being a hardcore AI company
4. Why AI Looks Like 2000
McGrath put it to him that he has expressed bearish views about AI valuations, and that he has said valuations are too high and that at some point in the next two years there will be a downturn. His answer separated the technology from the price.
The 2000 comparison
He said the technology is not in question. "I think that first and foremost the technology is real and amazing and unbelievable and a day before the ChatGPT moment which was on November 2022 I wouldn't have imagined that we'll ever get there in our lifetime," he said, adding that as a technologist it is stranger still to realize how well it works
"So I think it's real and it's here to stay forever. The question is about whether we are valuing all those companies in the right way"
The mispricing he sees is not at the model developers. "I'm speaking about many other companies which are tagging along for the ride," he said, and described a market where the label does the work: "I think it feels like when you're saying you are an AI company, you automatically get a high valuation"
The precedent he reaches for is the dot-com bust, and his point is that the internet was real then too. "And it feels like what happened in the year 2000 and the bubble," he said. The technology worked and people used it — "The thing is that the infrastructure then was overbuilt." He said some of that capacity was only paid off a few years ago, because it took that long to use
"The risk is that we're overbuilding as a world in AI"
He cited capital-spending plans by five of the Magnificent Seven that he said run to more than two and a half times their expected revenue, prefacing the figure with "If I remember correctly and I hope I'm not misquoting." He said investors are encouraging that spending because they expect demand to outrun capacity
Too many new data centers
He named two problems with the buildout: whether the demand is there, and whether the industry is fragmenting instead of consolidating. The first is a return-on-investment question about the dollars going into data centers
The second is a reversal of the previous decade: "Before AI, you saw consolidation of data centers. You didn't see a lot of small data centers," he said, and everything looked to be converging on a few large clouds
Now, he said, there is a wave of new cloud operators building their own data centers without the prior knowledge or the existing infrastructure that Amazon, Google and Microsoft have. Each has to find electricity, buy chips that are a scarce resource and harder for them to get, and staff its own marketing and sales — duplication the incumbents do not have to repeat
His forecast is consolidation, and lower valuations with it. "I think some will survive. I think the vast majority will eventually consolidate," he said, adding that as efficiency improves the world needs less capacity, that the current investments would be very hard to justify, and that valuations will drop for many of those companies as a result
5. A $100B Addressable Market
McGrath noted that Ehrlich has said raising money is not the goal: "It's a means to an end, and it's also a loan that you need to pay back at either 10x, 20x, or 100x." She asked how big Eon could get. In her introduction she had said his goal is to grow it into a $100 billion company.
His fear starting out was not failure, which he called the easy outcome, but being wrong about the thesis — grinding for a few years and selling for a few hundred million dollars, an outcome he was careful to say he admires in others
He says the market is large enough to make the ambition credible. "We have an addressable market of maybe over a hundred billion dollars today growing extremely fast and I have a once in a lifetime opportunity to create a formidable company"
He is explicit that he is normally the person rolling his eyes at this kind of talk. "I lost my innocence," he said, describing himself as a very active angel investor who hears founders promise enormous companies constantly
What he thinks he has is one shot at a household-name company, and he says the investment community agrees. "A company that could be a formidable household name company worth 100 billion, 200 billion, 300 billion, even maybe what's fondly called a Kilocorn, which is a new name for a trillion dollar company"
"It's going to take a very long time. I don't doubt it, but we have an enormous market opportunity"
The technical defense he points to is patents: "We've been filing so many patents. We got so many of them already approved and we'll get to a few hundreds next year"
He turned the size of the opportunity into personal exposure. He said that "if we're missing this opportunity I need to go and live under a bridge and I can't show my face anymore and I kind of like showing my face so I better execute." When McGrath observed that he puts that pressure on himself, he said the team, the market and the vision are all in place: "If we fail, that's on us. And I think it's a positive pressure"
He tells employees the same thing, he said, and it is why they join
6. Fun Is a Soft KPI
Asked how much of that he articulates to staff, Ehrlich said he repeats it constantly — and then gave the management principle he says sits underneath it. His hires come from the incumbents and from the storage and security industries: "We have ex employees from Google from Amazon from Microsoft from storage companies like VAST Data and XtremIO, and from security companies like Palo Alto Networks."
His stated method is to hire the highest-caliber people he can find, point them at the right problem and make sure they enjoy it. "So my management philosophy is take fun seriously and this is a soft KPI," he said. Hierarchy and process without enjoyment produces nothing: "But if you get really smart people to want to wake up every morning and come to work and work with their colleagues and bring other people and solve very hard problems then all the other things will fall in line"
He asks his direct reports about it by name. "I'm speaking with my direct subordinates and I'm speaking with them about how much fun do their employees have," he said, conceding: "It's hard to measure"
Pushed by McGrath on what he actually looks at — productivity, logged hours, demeanor in meetings — he ruled out the first two. Long hours do not mean good work, he said, and productivity is close to meaningless in invention work, where one genuinely good invention in a year can beat a stream of small ones
What he watches instead is energy, whether colleagues like the person and the person likes them, how they behave when something works, how they react when they fail and how quickly they recover. He also asks directly, through his own conversations and through managers talking to their teams
The philosophy dates to the middle of his last company, and came out of pattern-matching rather than instinct. He said he went back through what had worked and what had not, then interviewed vice presidents of research and development, product, sales and marketing at other companies about who their top performers were and why. The answer that came back was that the top performers were the people enjoying the work
He also applies a hard exclusion. On hiring: "I never want to hire someone even if he's really smart that is toxic to the environment that people don't like him and that is not easy to work with even if they're geniuses"
7. Hours Are a By-Product
McGrath explained why she had asked about hours in the first place: "I only ask about the hours because two founders who've sat in your seat in the last few weeks have talked to me about the hours that they work at their companies and it is surpassing the 996 weeks that have become so popular in parts of Silicon Valley." She said she had pushed both of them on it and both had smiled and said they were having fun, and added her own view: "If you're going to work hard, if you're going to work long hours, it at least better be enjoyable."
Ehrlich's answer, which he flagged as a paraphrase of Jeff Bezos, was that hours are not what makes work feel bad. He said Bezos, in an interview, said that "people are not feeling pressure because they work hard. They feel pressure because of uncertainty." Given certainty and enjoyment, he said, people naturally work long hours most of the time
He set one limit against it. "Now I think it's a marathon and not a sprint," he said, adding that the constraint is people with families — that a partner should not come to resent the job and children should see the parent
He is skeptical that long hours mean output. He described people spending 15 hours in the office who drink coffee for two hours in the morning, tell themselves they have plenty of time, take an hour and a half at lunch, tell themselves the same thing again, and end up ineffective
He treats it as a matter of individual wiring rather than virtue. "So some people are really good with working long hours all the time and they enjoy that. Others simply can't because of how the personality is built"
He said he knows the 996 trend — the much longer working weeks the host was referring to — and is hearing of harsher versions, that Silicon Valley always has its trends, and that he tries to count hours less because hours follow from whether people are enjoying the work
8. Code, Then Soft Skills
McGrath turned to his resume, starting with his first job as a software engineer and the detail that he started programming at nine. She asked whether he had always intended to be an entrepreneur.
He says he never wanted to manage anything. "I enjoy writing code. I don't want to do anything else. I'll write code for the rest of my life," he told his father, who had asked why he would not become a manager. The two laugh about it now
What he liked was the gap between understanding every part and being surprised by the whole. He wrote 3D graphics code as a boy, understood the math behind every piece, and still found the spinning object on the screen astonishing — "writing code is like poetry," he said
The move into management was instrumental, not a change of taste. He said he became an entrepreneurial manager because activating other people around his own vision was how he could scale himself and build bigger things
Asked for one hard skill and one soft skill from his years as a researcher, developer and team leader in Israeli military intelligence, he chose the soft one and said hard skills are the easier half. The image of the lone super-genius nobody can work with does not survive contact with real work, he said
His answer was teamwork, trust and empathy, and he ranks it above technical ability. "The empathy that you must have for your peers, for your subordinates, for your managers is extremely important and it's way more important than any other hard skill you can learn," he said, adding: "Soft skills are too many times underrated and I'm putting a lot of emphasis on soft skills"
He said engineers are all characters, and that the job is to be relatable across people whose values differ, because the combination can be better than either side alone
McGrath offered a story of her own from a project she edits, which in its first year involved art, development and design teams together. She considered herself a good communicator as the wordsmith, and still turned to a colleague and said, in her words, "I feel like we're just speaking two different languages to each other" — the two sides had never learned the soft skill of talking to each other
9. DigiCash's Real Lesson
His first company, DigiCash, started in September 2005 while he was still in the military and became his full-time job late in 2006. He said he did not understand what he was doing, including what a promise to an investor actually is.
The money came anyway, and he still cannot explain it. "We somehow raised $6 million for this venture," he said, adding that to this day he does not understand how, and that "I think the entire world was more naive than today"
He argues that not knowing is often an asset, and that the visible evidence is filtered. "I would say ignorance is bliss," he said of the 20-year-olds starting companies in Silicon Valley, then named the bias directly: "We're seeing only the success stories. You don't see all the Stanford dropouts who started a company." The founders who succeed, in his account, are often the ones innocent enough to discover they were good at it
He said he still finds it hard to articulate what the company did — marketing technology that helped e-commerce sites sell more — and that at the time he assumed his own coding ability and a co-founder who had managed at Microsoft would be enough
The lesson he draws is the one that runs through the rest of the interview. "The most important thing I've learned there is that technology is not enough," he said. What he learned much later, and credits to Amazon, is to start from the pain, need or want of a person rather than from what he could build, and then work backwards to something that solves it
10. Money Left on the Table
McGrath asked about CloudEndure, founded in November 2012, which he ran as co-founder and vice president of research and development for 10 years — six as a standalone company and four inside Amazon Web Services after the acquisition. Her question was about a line on his own LinkedIn profile: that the company raised less than $20 million from venture capitalists before AWS bought it.
The company was deliberately built to be profitable, and he now thinks that was the wrong optimization. "In CloudEndure, we didn't raise a lot of money," he said. It raised $18.3 million and was profitable — and it left the money it did raise untouched: "We actually never spent any dollar from our round A," he said of a round that was effectively the A and B together. "It was $13 million. We never spent any dime of it." In hindsight, he said, they should have spent it, raised more and become a bigger company as a result
His explanation is that investors do not reward early profitability, they reward the option of it. "One thing I've learned, investors don't appreciate being profitable early on," he said — they want a company that could be profitable to use money to grow faster instead. "Investors appreciate growth, at least on earlier stages, way more than they appreciate profitability"
He applied the same logic to the AI leaders: when people note that Anthropic and OpenAI are not profitable, he said, that is a function of stage, and investors care more about growth and the eventual path to profit than about profit now
The question he now tells founders to ask is different from the one they do ask. "And the question to investors shouldn't be what kind of company do you want, but what kind of company would you tolerate and still invest?" he said, because every investor will say they want profitability and then also ask for growth and everything else
The exit is the evidence he uses against himself. "We've built a tremendous business. It was really good. We're profitable," he said of the years before the sale, and of the years after it: "Not only that, we were profitable after selling to AWS, we grew the business internally to over a billion dollars in revenue rate. So it was a really good buy for AWS." The price, in his telling: "And when people in AWS realized for how much we sold the business, which was according to newspapers a bit over $200 million, they always laughed at us that we left so much money on the table"
He said the cause was not the technology or the business but the scoreboard: "And the reason was that we didn't necessarily understand how we're going to be measured by investors and acquirers when building and selling the business"
Asked what he would change, he said nothing. "I actually wouldn't change anything," he said, because the puzzle of the price is what sent him into investing. "And then I have friends with their cyber companies, not doing almost any money, just burning lots of cash" — companies with perhaps a million dollars of annual recurring revenue selling for $200 million to $500 million. "And this got me to try to understand how the world works"
11. Team First, Space Second
McGrath said her notes had him invested in about 30 companies. He corrected her: over 60. Asked for the single thing he looks for, he gave an ordering.
He ranks the founding team first and the market second, and treats everything else as provisional because the plan will change. "I'm investing mainly in team, second in space," he said. "The rest doesn't really matter because it'll pivot almost always. When did you last see a company that didn't even slightly pivot? Almost never happens"
What he looks for in the team is grit, and a founder who can be aggressive and kind at the same time
He wants the team complete before he invests, because he does not believe a critical function can be outsourced. Technology, go-to-market, product and vision all have to be inside the founding group, alongside founder-market fit and a market other investors will want to fund again — the next company in that space, and the one after that, since he says the first is the easiest to raise for today
His early investments were more contrarian, he said; today he invests more often alongside Sequoia, Lightspeed, Greylock, Index and other top-tier firms
The exits, looked at in hindsight, keep pointing at the same two variables. Two days before the interview, he said, one of his portfolio companies was acquired by 1Password, and he had breakfast with the founder that morning — a founder he had originally connected with his investors. "So the world changes, technology changes, trend changes eventually. The team and space are the most important things in my opinion"
12. Venture as a Hobby
Asked whether he expects to be an angel investor for the rest of his career, Ehrlich said he is not doing it for the returns, which he said are not significant relative to everything else. He invests in a lot of companies, is involved with several venture funds and advises a number of startups, and he gave three reasons for all of it, none of them financial. The first is that it teaches him how the world actually measures a company.
The advice he singles out as the worst he ever got is the advice founders hear most. "The worst advice I ever got in business is build a great business, money will come," he said. "What do you mean when it will come? How do you define a great business? I don't know. I have no clue"
His replacement for it is to learn the scoreboard first. "You need to understand how you will be measured and those who will measure you will be the investors, the acquirers or the underwriters," he said — and if the plan is an initial public offering, to learn that now and plan accordingly
"For example, don't be profitable too early. Depends on the industry"
"For example, focus on generating more annual recurring revenue as opposed to one-time revenue"
The second reason is that watching other people's decisions is cheaper than making them. "The other thing is that it's scaling me," he said. He gives founders his opinion, data and logic at decision points, then watches: "The great thing is I may be in agreement with them or not, but I'm not accountable to their decision no matter what happens." He checks back in three months and in six. "You have a simulation of life where you can see a whole bunch of decisions being made and you're not accountable for them"
He said what he learns there changes how he thinks about his own company — profitability, what to measure, which roles matter, what a great entrepreneur or manager or visionary actually is, what has to be a core competency in-house and what can be outsourced
The third reason is simply the people. "In addition that I just really love people. I really love speaking with people. I really love meeting crazy smart fascinating people," he said — founders backed by Sequoia, Lightspeed, Greylock and Index, who he said think radically differently from everyone else and would be hard to meet any other way
He said it makes him a better chief executive, and he hopes a better husband, father and person. And he put it in the category he thinks it belongs in: "Other people have hobbies like knitting or watching Netflix or playing poker. I used to love playing poker. I don't have time today. My hobby is venture"
Bonus Insights
McGrath opened the episode by asking whether listeners had ever wished they could clone themselves to get more done, and framed Ehrlich as someone who thinks becoming an entrepreneur and an investor is the easiest way to do it. Ehrlich used the same word, unprompted, for both the move into management and the angel investing: scaling himself
He dislikes the name of the company he sold to Amazon, and picked it himself. "I hate the name CloudEndure, by the way. I invented the name," he said — he wanted an available .com domain, was looking for something combining endurance and cloud, and told himself he would replace it within months. Amazon still uses it internally
On the temptation to re-run past decisions with today's knowledge, he said: "I could go to ninth grade and get a higher score in a test, right?"
Ehrlich's bottom line is that the technology is permanent and the prices are not: AI will stay, the capacity being built for it looks to him like the internet infrastructure of 2000, and the only advantage a company holds that its competitors cannot buy from a model provider is the data it accumulated itself.
Products, Companies & Tools Mentioned
Eon (Ehrlich's company, valued at $4 billion, which collects an organization's data and makes it secure, reliable and usable by other systems)
Anthropic, OpenAI, Databricks and Google (The AI systems he says every company is now building on — which is why he argues the customer's own data is the differentiator)
Chick-fil-A, SoFi and AlphaSense (The customers he is free to name; he said some large Fortune-listed users have not cleared him to use theirs)
CloudEndure and Amazon Web Services (His previous company, which built cloud migration and disaster recovery, raised $18.3 million, was profitable, and sold to AWS for what he said newspapers reported as a bit over $200 million)
ChatGPT (The November 2022 moment he says he would not have believed possible in his lifetime the day before it happened)
Amazon, Microsoft and Google (The established clouds whose existing infrastructure and know-how the new data-center builders have to replicate from scratch)
VAST Data, XtremIO and Palo Alto Networks (The storage and security companies his engineers came from, alongside Google, Amazon and Microsoft)
Sequoia Capital, Lightspeed, Greylock and Index Ventures (The firms he now invests alongside, after starting with more contrarian bets)
1Password (Acquired one of his portfolio companies two days before the interview)
DigiCash (His first startup, begun in September 2005 — marketing technology to help e-commerce sites sell more, and the source of his rule that technology alone is not enough)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: