Insight Partners Says It Is Going Earlier, Not Heavier, on AI’s Biggest Labs

The rush to back the biggest AI labs is testing a basic venture-capital rule: diversify or concentrate. Insight Partners is sticking with diversification, co-head Devin Parekh said in a newly published TechCrunch interview, even as some funds pitch putting 35% to 40% of a fund into OpenAI or Anthropic.

Parekh’s argument is not that those companies cannot succeed. Insight itself holds stakes in both. His concern is portfolio construction over many funds: a concentrated wager may improve returns in a particular moment, he said, but long-run data and the preferences of most limited partners favor spreading risk.

The firm’s position comes as venture valuations rise at a pace Parekh compared with 2021. Follow-on rounds would normally offer investors more operating evidence and therefore lower risk, he said. Instead, deals are moving so quickly that investors can pay more with little added data to support the higher price.

The logical response is to go earlier.

Devin Parekh, Insight Partners co-head, speaking to TechCrunch

Parekh said Insight is responding with earlier investments and smaller opening positions. Its scale fund can write initial checks of roughly $20 million to $25 million rather than about $500 million, then invest more heavily in companies that prove themselves. He said that approach has generated a disproportionate share of the firm’s returns.

• Insight has not completed a major buyout since 2024, citing high rates, weak software debt markets and lower exit multiples.
• The firm returned more than $20 billion to limited partners over the past two years through strategic sales and IPOs, with several billion dollars more expected, Parekh said.
• At the Series A and B stages, Parekh said Insight avoids directly competing investments and uses information-sharing restrictions.

The emphasis on smaller early bets is paired with a sharper focus on cash distributions. Parekh said many funds that raised large sums have not returned money to their limited partners, putting future fundraising at risk. His advice to managers with a fast-rising position: take out the original investment even if they believe the company can climb further.

That does not mean Insight treats all rivals alike. Parekh said later-stage holdings in OpenAI and Anthropic are possible because Insight is off the board and not directing governance. Early-stage conflicts are different, he said, when investor access to information and influence can make competing stakes harder to manage.

The unresolved test is whether patient portfolio management can hold its appeal if concentrated bets in the leading labs keep outperforming. Parekh acknowledged that a larger Anthropic allocation would make Insight’s current returns look better. His answer is that one fund is not the horizon the firm is managing for.