Insight Partners’ Deven Parekh Turns AI Concentration Into a Diversification Bet

OpenAI and Anthropic pulled in roughly $217 billion of the $407 billion that flowed into AI startups in the first half of 2026, according to PitchBook data, about 53 percent of the total from just three funding rounds. Deven Parekh, a longtime Managing Director at the $90 billion firm Insight Partners, says he knows two funds raising their entire pool in a single month whose pitch to limited partners is putting 35 to 40 percent of that fund into just one of the two companies. Against that backdrop, Parekh spent a Thursday-night interview at TechCrunch’s StrictlyVC event making the opposite case: stay diversified, even while holding stakes in both rivals.

The interview, conducted by TechCrunch’s Connie Loizos in New York and published September 13, is candid in a way institutional venture capital rarely is on the record. Parekh discussed a deal Insight lost, a philosophical disagreement with a well-known Khosla Ventures investor over AI risk, and why a firm his size can afford to skip nearly every fight without hurting its returns.

A Firm That Owns Both Sides of the Rivalry

Insight Partners was founded in 1995 and now manages roughly $90 billion in assets across more than 875 portfolio companies. Parekh joined in 2000, has led or co-led more than 130 software investments since, and has spent what TechCrunch describes as 26 years helping run the firm alongside co-founder Jeff Horing. Its portfolio includes Databricks, Wiz, and stakes in both OpenAI and Anthropic.

That last detail used to be disqualifying. “You’re invested in rival companies, OpenAI and Anthropic. That was once taboo in VC,” Loizos noted in the interview. Parekh’s answer turns on timing rather than principle: at the earliest rounds, exclusivity is real and enforced. “Khosla did OpenAI’s Series A, and there’s no way they could have then invested in Anthropic, and if we’d done Anthropic’s Series A, we likely couldn’t have done OpenAI either,” he said. Insight still holds itself to that standard at the Series A and B stage, maintaining information-sharing restrictions and avoiding direct competitors, “though some founders are sensitive even to 2% revenue overlap.”

The rule stops applying once capital needs outgrow it. “Once you’re at a later stage, off the board, not driving governance, you’re just buying a great stock,” Parekh said. “We saw OpenAI as the dominant consumer play and Anthropic as having a clear enterprise strategy; that’s shifting in real time. As these companies needed to raise $30 to $100 billion, they stopped being able to dictate exclusivity.”

Insight is far from alone in holding both. sxz.io reported in June that at least 13 investors, including Amazon, Sequoia, Fidelity, and Insight itself, appear on both companies’ official funding announcements. What Parekh’s interview adds is the reasoning: overlap at the late stage isn’t collusion or a lack of conviction, it’s what happens once mega-round sizes make picking a single side financially irrelevant.

The Industry Is Concentrating. Insight Says That’s a Mistake.

Parekh’s real target isn’t dual ownership, it’s funds that go further and build an entire portfolio thesis around one lab. “I’m an LP in other funds, and I know two funds right now (raising their entire fund in a month) whose pitch is literally, 35 to 40 percent of this fund is going into one of those two companies,” he said. “I’m not saying OpenAI and Anthropic won’t do well. But this business has always rewarded diversification over a long horizon.”

He’s explicit that concentration would look better in the short term. “In this particular moment, if 25% of our fund were in Anthropic, our returns would look better,” Parekh said. “But data over time doesn’t support excessive concentration, and most LPs don’t want that exposure either, though firms like Founders Fund and Thrive have done very well running concentrated strategies. There are always going to be exceptions who execute that well.” Insight, on its 13th fund, frames the decision on a longer horizon: “we have to think in terms of ten funds, not one.”

The PitchBook numbers show how skewed the alternative has become. Vertical AI application startups accounted for 62.9 percent of H1 2026 deal count but captured only 12.9 percent of the capital, while horizontal platforms and frontier labs took 70.8 percent of deal value, about $288.1 billion, largely through a handful of enormous rounds: OpenAI’s $122 billion raise in March, and Anthropic’s $30 billion Series G in February and $65 billion Series H in May.

Where the Diversification Argument Actually Pays: Wiz

Parekh’s clearest example of the strategy compounding is Wiz, the cloud security startup Google acquired for roughly $32 billion in a deal that closed March 11, 2026, with Insight expected to collect about $2.9 billion from the sale. “With Wiz, we wrote a Series A and kept writing checks, so our gain was much larger than if we’d stopped at the first check,” he said. “And if Wiz hadn’t worked out, it would have barely dented a fund our size.”

That’s the mechanical case for a scale fund making smaller, repeatable bets instead of a handful of enormous ones. “With a scale fund, you can make smaller bets: write a $20-25 million check instead of $500 million, and double down on the winners,” Parekh said. “That’s where our returns have disproportionately come from.”

Losing Is Part of the Model, Too

The interview’s most concrete admission of failure involves Legora, the Swedish AI legal-tech startup. “We competed for Legora: my partner Jeff Horing flew to [Stockholm] to pitch the company, because that’s where the founder was. We lost that one to General Catalyst,” Parekh said, adding plainly: “I don’t know the specific reason, but I think they sold their value proposition better than we sold ours that time.” Legora has counted General Catalyst among its backers since an earlier round, according to Crunchbase, and by its $550 million Series D in March 2026, led by Accel, the company was valued at $5.55 billion.

Parekh’s response to losing a marquee deal is the same diversification logic applied to talent, not just capital: “There are plenty of examples where it went the other way. It’s a big world; we don’t need to win every deal.” He also drew a distinction other AI-investing narratives tend to flatten: AI infrastructure talent is genuinely concentrated in San Francisco, he said, but vertical AI investing can be far more geographically distributed, since a fintech-focused company like Ramp draws on New York’s finance talent instead.

Why This Interview, Why Now

The conversation opened on a pointed question about AI risk, framed around what Loizos called “a researcher who’s become the big story of the week.” That almost certainly refers to Jacob Coxon, the Anthropic researcher whose resignation and public warning that AI could pose an extinction-level risk within a decade sxz.io covered on September 10, after Anthropic’s own head of alignment stress testing publicly agreed with him. Loizos noted she had put the same question to Khosla Ventures’ Keith Rabois earlier that night, and that he had been dismissive.

Parekh’s answer took the other side of that dismissiveness while still landing on optimism: “Sure, there’s a risk some non-state actor gets access to an open-source model and creates a biological weapon. But there’s an even higher probability we get a massive decrease in the time it takes to develop new drugs and cure diseases. I’ll take that bet.” He pointed to his own seat on the board of NYU Langone as evidence: “what AI is already doing with patient data is amazing. We can look at 50 million patient records and tell someone walking in for something unrelated that they have a 25% chance of a heart attack. Net-net, I think this is highly positive.”

The rest of the interview reads as a broader case against herd behavior in venture markets generally. Buyouts, Parekh said, aren’t attractive right now: “rates are high, debt markets aren’t receptive to software, exit multiples have come down.” Insight hasn’t done a major buyout since 2024. Venture valuations, meanwhile, are climbing at a pace he compared to 2021, without the added diligence data that normally justifies a higher price on a follow-on round, which is part of why Insight has pushed earlier and smaller instead. On the boom-bust pattern that produces both dynamics, he was unsentimental: “We all do this in our personal lives: stay out of an expensive market until we can’t stand it anymore, and pile in right when we should be pulling back. LPs do the same thing at a macro level.”

Parekh expects the current frontier-lab growth curves to normalize eventually, pushing OpenAI and Anthropic toward public markets within the next 18 months as their growth rates settle into something more ordinary. Whether Insight’s diversified approach outperforms the concentrated bets other funds are making on getting there first is a question that, by Parekh’s own framing, won’t resolve on a one-fund timeline. It resolves on a ten-fund one.