Thrive Capital's 2022 early-stage fund, which put $516 million into OpenAI, SpaceX and Anduril, is now marked at more than $3.7 billion, and the firm is trimming part of its position. That single line captures the defining tension in venture right now: the gap between what portfolios are worth on paper and what actually returns cash to investors.
A roughly sevenfold markup looks spectacular, but it is unrealized until someone writes a check. The decision to sell a stake is the more revealing signal. It says the smartest allocators are no longer content to let AI valuations compound on spreadsheets. With most 2020-2022 vintage funds sitting on near-zero distributions to paid-in capital, limited partners are demanding proof of liquidity, not another upward revision. Selling into strength, before a private valuation meets public reality, is a discipline most firms have avoided because it caps the headline number.
The global implication is a structural shift toward the secondary market as the primary exit path. IPO windows remain narrow, and the largest AI names are staying private far longer than the fund lives that back them. That forces a new plumbing layer: continuation vehicles, structured secondaries, and tender offers that let early backers crystallize gains while late money buys in at reset marks. The risk is concentration. When a fund's return is driven overwhelmingly by one asset, its reported value is really a bet on a single company's next round, not a diversified portfolio.
For Japan, the read-through lands squarely on institutional allocators. GPIF, mega-bank affiliates, insurers and the wave of Japanese corporate venture arms that raised AI exposure over the past three years are holding fund positions now marked up on the same logic. Those markups flatter internal reporting but say little about repatriable yen. Japanese LPs should be pressing GPs on realized DPI and on whether the manager is willing to sell, rather than celebrating paper IRR.
The deeper gap is domestic: Japan's private secondary infrastructure is thin, so local funds lack the exit optionality US peers are now exercising. For Japanese CVCs and enterprise strategists, the lesson is to treat AI stakes as positions to actively manage, building relationships with secondary buyers and continuation-fund sponsors before valuations normalize, instead of assuming an IPO will arrive on schedule.
