Private equity sponsors are converting legacy infrastructure holdings into continuation vehicles rather than liquidating positions in AI-critical assets, according to secondaries transaction specialists tracking GP-led deals. The pivot reflects a calculation: the capital deployed between 2019 and 2022 on fiber, edge compute, and data center builds is now worth materially more than the original underwriting assumed, and selling into this cycle would crystallize gains before the next phase of enterprise spending arrives.

Continuation vehicles—structures that allow GPs to transfer portfolio companies from maturing funds into new entities—processed an estimated $47 billion in transaction volume through Q3 of this year, with AI-adjacent infrastructure representing the largest single category by deal count. Sponsors who acquired colocation operators, subsea cable networks, or GPU hosting platforms during the pre-ChatGPT era now face limited partners who want liquidity and a forward view that says the real monetization window opens in 2026 when hyperscaler capex migrates from model training to inference deployment at scale. The continuation vehicle solves both: it offers LPs an exit at a marked-up NAV while allowing the GP to retain economic exposure and collect another layer of management fees on the rollover capital.

The structure matters because it changes how allocators should think about vintage-year exposure. A fund that was supposed to distribute in 2024 is now effectively a 2025 or 2026 vintage if the crown jewel assets roll into a CV. That extends J-curve dynamics, delays cash back to LPs, and—critically—resets the return hurdle for the GP. Franklin Templeton's secondaries group reported $3.5 billion in AUM within twelve months of launching its continuation-focused strategy, a fundraising velocity that signals institutional appetite for these restructured positions. StepStone's recent 10-K filing shows advisory revenue growth tied directly to structuring GP-led transactions, which now represent nearly a quarter of its deal flow by fee volume.

The second-order effect is valuation discipline. When a sponsor moves an asset into a CV, that asset gets re-marked by the secondaries buyer who provides the liquidity to exiting LPs. Those marks are showing 15-20% premiums to the previous quarter's NAV on AI infrastructure, but flat to down marks on legacy industrial or consumer holdings in the same fund. That spread tells allocators which sectors have genuine pricing power and which are being carried at stale valuations. It also creates a selection problem: the best assets stay with the GP, and the weaker ones get sold outright or languish in the tail end of the original fund.

Operators and allocators should watch three follow-on developments. First, whether the continuation vehicles themselves face pressure to exit before 2027 if LP appetite for illiquid extensions weakens—a risk if public markets rally and alternatives look less attractive on a relative basis. Second, how secondaries buyers price the next wave of AI infrastructure CVs, particularly if data center utilization rates plateau or hyperscaler capex guidance disappoints in the back half of 2025. Third, whether LPs start negotiating consent rights or fee caps on continuation vehicles during new fund formation, which would constrain the GP's ability to use this lever in future vintages. Franklin Templeton and StepStone both report that institutional LPs are now modeling continuation optionality into their private markets allocation frameworks, treating it as a feature rather than a bug.

The data center operator that was underwritten at 12x EBITDA in 2021 and re-marked at 18x in a continuation vehicle last quarter is now the anchor asset in a structure that doesn't have to sell until the next fund cycle. That's not financial engineering. That's a sponsor with a forward view on inference deployment and the patience to let the thesis compound.