Gili Raanan - Cybersecurity Investment Playbook - [Invest Like the Best, EP416]
What this covers
This is a conversation between an interviewer and Gili Raanan, founder of Cyberstarts, who moves from the threat landscape to the mechanics of investing in it. The first stretch covers offense and defense. Attackers choose targets by ROI-driven attack selection: the most damage for the least effort, cost and time. Raanan notes that something as simple as taking down name servers for a country's power stations could cut electricity for days. He treats LLMs as dual-use technology. The model that designs drug recipes can be coaxed into describing how to kill, and open-source releases like DeepSeek mean guardrails cannot be centrally enforced. He says attackers already use AI agents, and that defenders will eventually have to hand control to AI to respond in time. He also argues, contestably, that there is no standalone cybersecurity market, only security derived from each new technology.
The second half is about investing. Cyberstarts' Sunrise methodology delays writing software until dozens of customer conversations have verified a pain point, which can take six months. Those conversations weigh what buyers have actually done over what they say. He also describes the four buyer personas and why Wiz worked with a CISO holding all the roles. Wiz also went against Silicon Valley convention by selling to large enterprises from day one. Raanan defends high seed valuations, using Airbnb and Instagram as examples, though he concedes a small ownership stake can leave a company short of investor attention. He selects founders on adversity rather than intelligence, and draws on his own early CAPTCHA company, which had no business model.
Raanan argues cybersecurity is being redefined by AI and geopolitical conflict, so defenses must be AI-native, and that a repeatable pain-first process (Sunrise), applied to founders chosen for adversity-driven resilience, produces outsized early-stage venture returns.
- AI-driven attacks run at a speed and scale that make prior defense architectures (rule-based and behavior-based) obsolete, so companies must be AI-first
- Cyberstarts' returns came from a single-sector thesis, founder selection for adversity, and validating real customer pain before writing code
- Wiz's early success followed the same logic: an urgent pain point, one buyer persona who held all four roles, and enterprise-first selling
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
The best offensive tactics are driven by return on investment: applying the highest damage for the smallest effort at the lowest cost and fastest speed, with the definition of damage depending on the target.
“The best offensive tactics are always ROI driven.”
There is no standalone cybersecurity market; cybersecurity is always derivative of a new technology or business, so cloud, autonomous vehicles, and AI each create their own security category, and a focus on picking the right talent will let the best domains emerge.
“There's no cybersecurity market. Cybersecurity is always a derivative of something else”
The same LLM that can design drug recipes for diseases such as Alzheimer's can be asked for recipes for killing people, and guardrails can be bypassed, so a bank's defensive AI agent can be tricked into describing how to breach that bank.
“You can use the same LLM and ask for recipes for drugs that can kill people.”
The traditional venture model is broken for both investors and entrepreneurs, because investors ask seed founders about markets, products, and technology that change within weeks, so the dialogue becomes a game that adds no value to the people building solutions.
“my conclusion back in 2018 was that the venture model is completely broken.”
AI will not simply improve cybersecurity; it will redefine it, because the same LLMs and AI agents used to predict and prevent attacks can be used to deliver attacks at a scale and sophistication not seen before.
“AI is not going to improve cybersecurity, it's going to redefine it.”
Against Silicon Valley's common wisdom of product-led growth with low entry prices, Wiz chose large enterprise deals from the start, which brought committed sophisticated customers, outside investor attention, and top go-to-market talent, growing ARR from $1 million in the first quarter to about $25 million by the fourth quarter.
“Unlike the common wisdom within Silicon Valley, we decided to go for the large deals, scrap PLG”
Companies that become expensive are expensive from the seed round onward, as with Airbnb and Instagram, and Wiz was expensive from the start yet became the fastest unicorn and fastest to reach $100 million and $500 million, so a high price can be a good sign when justified.
“the expensive deals, the pricey companies, are pricier from day one.”
Open-source models such as DeepSeek mean AI is no longer confined to a walled garden, because anyone can download and modify the model, so guardrails cannot be centrally enforced.
“Now you have open source models, and Deepseek is just one example of this, where AI is not in a walled garden anymore; it's in the wild.”
Building important companies is expensive, so founders should raise enough capital to build the right product and hire sales teams, which requires high valuations, and high prices are justified for companies with proven product and team; passing on deals only for price means passing on good companies such as Airbnb and Instagram.
“So if you're going to pass on an opportunity just because of price, you're going to pass on all the good companies.”
A cybersecurity product faces four personas: the person with the pain, the budget holder, the decision-maker with authority, and the end user; mapping all four to one person is a mega hit, two people is very nice, three is borderline, and four distinct people means the team should pivot, which is why Wiz succeeded with the CISO holding all roles.
“if those four personas map into a single person in real life, that's a mega hit.”
Attackers are already using AI agents to take control and launch attacks, and LLMs can quickly generate attack plans that defenders did not practice for, so the Terminator-style autonomous future is already here rather than science fiction.
“They're already using AI agents to take control and launch those attacks.”
Wiz sourced engineers whose hobby is writing code rather than those with elite pedigrees, reasoning that a great developer is about ten times a good one, so a hundred such engineers act as a force multiplier equal to 300 or 400 strong engineers.
“He would source people whose hobby is writing code.”
When interviewing buyers about pain, Cyberstarts discounts what customers say and looks at what they have done to address it, since a claimed top pain that has triggered no spending or action is probably not as important as stated.
“words are cheap. So don't look at what the customer says; look at what the customer does.”
Beyond Networks, the early Wiz company, had a secure-access idea that addressed an important but not urgent problem, and a small venture should prioritize urgent pain points because urgency drives faster customer engagement, learning, and revenue, so the team pivoted to cloud security and renamed the company Wiz.
“it was important, but not urgent.”
Raanan's first startup invented and built the first working CAPTCHA, yet could not find a business model for it, which taught him that starting from technology and then searching for a problem is the wrong order, leading to the pivot into web application firewalls.
“we couldn't find a business model for the technology. That taught me a lesson”
When interviewing security leaders, customer statements are unreliable, so Sunrise weighs what they have actually done about a pain point, such as downloading open-source tools, hiring consultants, or evaluating vendors, rather than their opinions.
“words are cheap. So don't look at what the customer says; look at what the customer does.”
Attackers can cause large-scale harm with simple methods, such as taking down name servers for a country's power stations to cut power for several days, which requires no army of operatives and can happen to any country on any day.
“taking down name servers for their power stations and leave that country without power for a few days”
To reach a statistically meaningful pain-point conclusion, a team must talk to dozens of organizations, make a choice, then run another round of conversations on solution design, and only after that build software, which sometimes takes six months.
“I think that to have a statistically meaningful outcome, you need to talk to dozens of organizations.”
Sunrise is not just a product-market fit exercise but a program that takes an entrepreneur from inception through the first three or four years of running the business, with the early phase focused on finding the most important market pain point by speaking with large organizations.
“It's actually a program that takes an entrepreneur from day one—from inception through the first three or four years of running the business.”
Military technology evolves through stages of better analysis, human augmentation, workflow automation, and finally AI taking control; human augmentation may suffice for now but is not a long-term solution, so defenders must eventually give AI agents control to respond in time.
“It started with better analysis, better decision-making process, then it moves to human augmentation.”
Cyberstarts committed to not questioning the market, technology, product, or competitive advantage at investment time, and instead to focusing on the founder as the single thing that matters.
“to completely avoid questioning the market, the technology, the product, or the competitive advantage—and to focus on one thing only, which is the person ahead of me.”
The ability to overcome adversity is the greatest signal of a terrific founder, more important than raw intelligence, because startups are an insanely painful journey and people who did not have a straight-line start are the ones who persist.
“The most important thing, in my view, is the adversity and the ability to overcome it.”
The old systems of rule-based and behavior-based security, used to fix misconfigurations and patch bugs, are gone, so founders must build AI-first and AI-native cybersecurity companies to have any chance of lasting value.
“founders would have to create cybersecurity companies that are AI-first and AI-native just to have a chance”
Unlike typical startups that hire and build as soon as VC money arrives, Sunrise delays software development until the chosen pain point is verified by dozens of customers, which yields solutions customers love and a high success rate.
“Sit tight, don't get too excited. This is your last chance to pick the right problem to go after.”
High valuations are a double-edged sword because when an investor owns only about 5 percent of a company, the company is unlikely to become the most important one in that investor's portfolio, so founders must compete for attention, bandwidth, and access.
“when investor ownership goes down significantly, it's hard to become the most important company in the portfolio when an investor owns 5% of their cap.”
To reach a statistically meaningful pain point, a team must talk to dozens of organizations, using the framing that it is an offer of about $100 million of engineering budget off the buyer's balance sheet, then iterate over multiple rounds.
“I think that to have a statistically meaningful outcome, you need to talk to dozens of organizations.”
Sunrise uses simulations that assume failure, such as a customer not taking a product evaluation, to force founders to plan go-to-market, pricing, channel strategy, and even headquarters location for the next three to four years.
“We use a lot of simulations that assume failure and force the founders to really analyze this situation”
The best founders are often unaware of their superpowers because they use them naturally, and the superpower a founder claims often reveals how they view their weaknesses, since the claimed skill is usually one developed to compensate for a deficit.
“the superpower they claim they have is not necessarily the real superpower they possess, but it shows you the way they view their weaknesses”
Defensive methods developed over roughly the past eighty years, starting in World War II, would be useless against AI-driven attacks because their speed, scale, and sophistication exceed anything previously seen.
“the methods that we developed, maybe for the past eighty years starting in World War II, would be useless”
Sunrise reverses the power balance in customer conversations by presenting the team as a new Cyberstarts company that will spend about $100 million on engineering over three years, asking customers what single pain point they would have it solve, which gives executives a free, high-stakes frame to share real priorities.
“We are giving you essentially $100 million off balance sheet to solve one pain point”
Offensive cyber has become a lethal or strategic weapon for states in conflicts like Ukraine and Gaza, and capabilities drift downward from state-level weapons to criminal organizations and then to script kiddies.
“Offensive cybersecurity became a lethal weapon or even a strategic weapon for states”
After losing his 19-year-old daughter in a tragedy before starting Cyberstarts, Raanan coped by continuing to work, not fighting the pain but waiting for it to subside, and found that working with younger people on innovation kept him younger and healthier.
“I taught myself to live with the pain—not fight it, just wait for the pain to subside.”
The first fund's success came from luck plus four drivers: a clear single-sector thesis that Cyberstarts owns, founder selection criteria, eliminating market, technology, product, competition, and business plan considerations early, and a commitment to fast product-market fit through Sunrise, plus strong partners.
“So luck, thesis, founder selection criteria, and then I would add two other factors.”
Early-stage cyber investors should not try to identify the best cybersecurity domain, since security categories will introduce themselves; instead they should pick the right talent, partner with them, and use a Sunrise-style system to build what customers love, because if they do that everything else will fix itself.
“If we do that, everything else will fix itself.”
An AI agent can generate thousands of attack scenarios and execute them simultaneously, and this is current technology rather than a future concept.
“an AI agent that's able to produce thousands of scenarios like that and execute all of them simultaneously”
Shortly after Fireblocks' first investment it entered the first crypto winter, and six of its first ten design partners went bankrupt, forcing founder Michael Shaulov to rethink the customer plan entirely.
“Six out of his first ten design partners went bankrupt.”
Cyberstarts' first fund of $50 million invested in nine teams, all with Cyberstarts as lead, including re-bets on people without ideas, and the portfolio was valued at over $25 billion by 2021, roughly $2 billion of value attributable to the fund.
“the portfolio of those nine teams is valued at over $25 billion, converting the $50 million into close to $2 billion.”
In evaluating founders, the important part of their life story is the why behind decisions, such as why they chose one job or partner over another, because that reveals how they make decisions, analyze situations, and what drives them.
“what's important in the story is not the what, it's the why.”
Each of Cyberstarts' four active funds has delivered over 100 percent IRR to investors, showing the outperformance is systematic rather than the product of a single successful first fund.
“Each of the four active funds delivered over 100% IRR to investors.”
Cyberstarts sees two pricing markets at seed: first-time founders typically raise at $15 to $20 million post-money, while repeat founders range from $40 to $50 million and sometimes go much higher, and this price level has barely changed in seven years.
“For first time entrepreneurs, we've seen seed deals anywhere between $15 to $20 million post-money and for repeat entrepreneurs I see a broader range starting from the $40 to $50 million range.”
Cyberstarts manages over $700 million across five funds and its portfolio is worth about $45 billion, roughly 50 percent of the worldwide market cap of private cybersecurity companies, achieved in under seven years, with each of the four active funds delivering over 100 percent IRR.
“The portfolio value today, cumulatively, is around $45 billion, which is 50% of the worldwide market cap of private cybersecurity companies.”
Cyberstarts invested $3 million in June 2018 as the seed for Fireblocks, which is now valued near $8 billion and serves institutions such as BNY Mellon, BNP Paribas, and ABN AMRO.
“In June 2018, we invested $3 million, and that was the seed for Fireblocks. It's an $8 billion company today.”
Luck matters in venture returns, but Warren Buffett's observation that the harder one works the luckier one becomes applies, so effort and thesis discipline amplify luck.
“Warren Buffett once said that as he works harder and harder, he becomes luckier and luckier.”
When Cyberstarts was founded in 2018, cybersecurity was widely considered a boring portion of IT, but by the mid-2020s geopolitical and AI forces had made it a high-priority domain.
“Cybersecurity was considered a boring portion of IT.”
A favorite Sunrise question for finding pain is who is the vendor you hate the most, because disliking a vendor creates strong motivation to displace it.
“Who's the vendor you hate the most?”
Raanan's self-confidence developed over time as compensation for a feeling of being unappreciated as an insecure child in a small Israeli town, and it solidified when he joined the Israeli signals intelligence unit at 18 and found a group he belonged to.
“it took time and served as compensation for being a very insecure child”
Fireblocks founder Michael Shaulov, who immigrated to Israel as a child from the former Soviet Union and was raised by a single mother, learned to handle uncertainty, which helped him survive the first crypto winter when six of his first ten design partners went bankrupt.
“Six out of his first ten design partners went bankrupt.”
Wiz built a product that during the first prospect call requested AWS credentials and showed real value in the customer's own environment rather than in a demo, which accelerated sales.
“allowed them to ask for the AWS credentials. And within the call, show them real value for their own organization rather than in a demo environment.”
The Wiz founders, Assaf, Ami, Roy and Nir, began their journey with Adallom, sold it to Microsoft within about three years, spent five years at Microsoft building a cloud security business to roughly a billion dollars in revenue, and left in 2020 as, in Raanan's view, the most experienced cloud security team worldwide.
“they were the most experienced team in my view worldwide in running a cloud security business.”
The Sunrise approach is applicable to any field with relatively uniform buyers, such as fintech, gaming, or medical devices, but it also needs an active ecosystem including M&A so companies can be acquired.
“I believe that in any field where there's a relative uniformity of buyers, that's applicable.”
The kindest thing anyone has done for Raanan is that his son said he had realized after much thought that Raanan is his best friend, which Raanan considers the best thing anyone could say.
“my son told me that he had spent a lot of time thinking about it and realized that I'm his best friend.”
Cybersecurity today is a 'perfect storm' because global conflicts (Ukraine, Gaza) have made offensive cyber a strategic weapon, and AI advances are creating unprecedented threats, so the threat environment is far more dangerous than a few years ago.
“we actually the perfect storm in cybersecurity for various reasons”
Achieving amazing venture returns has always been very hard and will remain hard, but the cybersecurity perfect storm, the AI-driven redefinition of the category, and rising talent quality keep the opportunity open for Cyberstarts.
“Achieving amazing returns has always been super hard. It will continue to be hard”
Raanan is never satisfied with achievements and always looks for the next peak, and he finds satisfaction only in making an impact on founders and partners he cares about.
“Making an impact on people that I really care about.”
Cyberstarts treats co-investors as partners who have repeatedly backed its companies, because even a great firm cannot be the only smart party involved, and Raanan is highly involved in helping portfolio companies finance growth.
“We think at Cyberstarts of co-investors as partners.”
It is not enough for an investor to be first; they must also be right, and being right is what separates a terrific investor from a regular one.
“it's not enough just to be first; you also need to be right”
The best investors treat their area as a life project rather than a wealth-generation exercise, and Raanan regards cybersecurity as a life project and not merely an investment area.
“do something that's a life project for you.”
Wiz reached a $32 billion valuation within five years of founding, and Raanan predicts that within ten years a new cybersecurity company will break that record because the field has market size, an evolving pain point, fast change, and talent.
“within the next decade we would see that record being achieved and broken by a new player.”
A hit rate of four or five unicorns, including one decacorn and several companies acquired for hundreds of millions, is proof that the thesis and model are sound, because early investment in terrific companies drives outsized outcomes.
“a hit rate of 4 or 5 unicorns, including one decacorn”
Major cybersecurity platforms such as Wiz, Sierra, and Island invest heavily in applying AI, and this is re-architecting the product to remain effective rather than an additive feature.
“That's not an additive capability to the product; that's re-architecting the product”
Cyberstarts sold five portfolio companies in the past twelve months at a cumulative value of about $2.5 billion, showing an active M&A ecosystem matters.
“In the past 12 months, we've sold five portfolio companies at a cumulative value of two and a half billion dollars.”