# Why Insight Partners won't bet the farm on OpenAI > Devin Parekh on writing small checks early, selling into froth, and why funds stuffing 40% into one AI lab worry him. - Title: Why Insight Partners won't bet the farm on OpenAI · TechCrunch - Summary: Devin Parekh on writing small checks early, selling into froth, and why funds stuffing 40% into one AI lab worry him. Diversification and liquidity beat… - Keywords: venture capital, ai investing, insight partners, liquidity, ipos, technology, startups, Insight, Partners, won't, bet, farm - Source: TechCrunch — https://techcrunch.com/2026/09/13/insight-partners-devin-parekh-on-why-the-firm-is-diversifying-while-everyone-else-bets-the-farm-on-openai-and-anthropic - Author: Connie Loizos - Published: 2026-09-13T21:30:00+00:00 - Read time: 4 min - Topics: venture capital, ai investing, insight partners, liquidity, ipos, technology, startups ## The quiet $90 billion firm Devin Parekh has co-run Insight Partners for 26 years. While other VCs live on X and podcasts, Insight stays off the stage and lets the portfolio speak. > Let the portfolio do the talking. ## AI risk, weighed honestly Parekh grants that a non-state actor could misuse an open-source model to build a bioweapon. He thinks faster drug discovery is the more likely outcome, and he'll take that trade. > There's a risk... But there's an even higher probability we get a massive decrease in the time it takes to develop new drugs. ## Fifty million patient records On the NYU Langone board, he watched AI read 50 million records and flag a patient walking in for something unrelated with a 25% heart attack risk. ## Aging population, too few doctors He frames AI in medicine as a capacity problem, not a luxury. There aren't enough medical professionals to serve an aging population, so scale has to come from somewhere. > We're going to need AI to scale healthcare. ## Not everyone is an expert Parekh's read on VC noise: the same people became epidemiologists during COVID and geopolitics analysts during the Iran war. Insight communicates enough to be known, and stops there. > Every venture capitalist thinks they're an expert on everything now. ## No fixed allocation Early stage, growth, buyout and secondaries shift fund to fund. Look across Insight's last seven funds and the percentages never repeat. > It's temporal, not fixed. ## Buyouts are cold High rates, debt markets that won't touch software, and compressed exit multiples have shut the window. Insight hasn't done a major buyout since 2024. ## Paying more for the same risk Normally a follow-on round buys you more data, so a higher price reduces risk. Rounds now move so fast there's almost no new data, so you pay up and the risk stays put. > you're paying more without reducing risk ## Go earlier, write smaller The answer to frothy late rounds is a $20 to $25 million check instead of $500 million, then doubling down on the ones that work. That's where returns have clustered. > That's where our returns have disproportionately come from. ## The Wiz math Insight wrote Wiz's Series A and kept writing checks, multiplying the gain. Had it failed, the loss would have barely registered against a fund that size. ## Talent went flat, mostly A partner flew to Stockholm to pitch Legora because that's where the founder lived. But AI infrastructure talent still clusters in San Francisco, and fintech talent in New York. > Talent has gone flat globally. ## Losing Legora, and shrugging General Catalyst won that deal. Parekh's read is simply that they sold their value proposition better that time, and he doesn't need every win. > It's a big world; we don't need to win every deal. ## Owning both OpenAI and Anthropic Once you're late-stage, off the board and not touching governance, you're buying a great stock. At Series A or B, Insight still bars directly competing bets. > Once you're at a later stage, off the board, not driving governance, you're just buying a great stock. ## Exclusivity died at $30 billion When labs needed to raise $30 to $100 billion, they lost the power to demand exclusivity from investors. The consumer-versus-enterprise split between OpenAI and Anthropic is blurring too. ## Robots are still science projects Physical AI means betting on whether robotics adoption happens and then on when. Insight is watching, not buying. His 23-year-old VC son thinks that's crazy. > you're making a bet on when robotics adoption happens, layered on top of a bet on whether it happens at all ## Funds with one holding Parekh knows two funds raising their entire vehicle in a month on a pitch of putting 35 to 40% into a single AI lab. He's on fund 13 and thinks in terms of ten funds, not one. > this business has always rewarded diversification over a long horizon ## Take your basis out Many first and second-time funds won't raise again because they never returned cash. Insight sent back over $20 billion in two years through sales and IPOs. > LPs want to know you can turn positions into cash; that's the job. ## De-risk 10 or 20 percent When a founder gets a frothy offer, Parekh asks what happens at the correction he can't time. You don't have to sell everything to be sensible. > you can't compound $40 billion at 50% every two months for two years without becoming the world economy ## Three trillion-dollar IPOs SpaceX, Anthropic and OpenAI could all list within six to eight months. The real test is the next tier, and what bar a four-year run to $65 billion sets for everyone else. > Eventually even these companies become normal-growth companies. ## The Armis $5 million check Insight lost the deal to Sequoia but kept the relationship warm with a small check. Eighteen months later it bought out the whole cap table, then sold to ServiceNow for $7 billion. > The goal is finding the best founders in the best markets. ## Key takeaway Diversification and liquidity beat concentration over a ten-fund horizon, even when concentration looks smarter this quarter.