The exit window reopened, but only for the very top of the AI stack

The buyers are the story. Neither of these was a strategic acquirer picking over a stalled asset. SpaceX closed the largest buyout ever of a venture-backed startup and paid for it entirely in stock ^[1], and Axios reported Stripe's OpenRouter deal as a mix of cash and stock ^[2]. Newcomer's read is that Stripe is now using its shares the way a public company would, which works because the company deliberately keeps secondaries possible ^[9]. That is the mechanism worth noting: private operating companies with liquid-enough paper can buy at IPO scale without an IPO, and sellers will take the paper.

The multiple explains why the practice will be copied. About $320 million invested against more than $8 billion of outcomes is roughly 25x on that capital ^[3][4], which for a firm that raised $1.25 billion for its first dedicated AI infra fund in 2024 and $1.7 billion more at the start of this year ^[6][7] means two deals can carry a sector fund. a16z's 2009 debut fund returned 6x DPI ^[5]. The two exits here, on the money actually deployed, cleared that ratio, and the timing was compressed enough that Newcomer called the speed extraordinary ^[1].

The DPI problem is the reason this matters beyond one firm's scoreboard. Most a16z funds raised after 2016 had not delivered meaningful cash back to limited partners as of the end of 2024, according to return data obtained by Newcomer ^[11], and the firm now reports $106 billion in assets under management ^[10]. LPs have been carrying paper gains on Stripe and SpaceX and accepting that exits are not everything ^[8]. Cursor and OpenRouter do not resolve that so much as show which door is open: a very small number of infrastructure companies sitting directly in the path of large operating buyers.

For everyone else, the IPO route remains blocked at the top end. OpenAI and Anthropic are showing strong revenue but still have hang-ups ahead of their planned mega listings ^[12]. If the two largest AI liquidity events of the moment both ran through private acquirers paying in their own equity, the practical question for a founder is not what public comparables suggest, but whose stock they would be willing to hold.

One caution on the a16z narrative. The credit for OpenRouter is contested in the sources: Anjney Midha led the seed in 2025 with Martin Casado on the memo, then left the firm shortly after, and a16z has publicly emphasised Casado and Chris Dixon instead ^[13][14]. Matt Bornstein is credited by many investors with initially leading Cursor ^[15]. A megafund model that depends on a named practice producing repeatable 25x outcomes is exposed when the named people move.