Secondaries as a Strategy: Navigating the AI Vintage Problem

In private equity, vintage year matters enormously. A fund that deployed capital in 2007 faced a very different environment than one that deployed in 2010. The same dynamic is playing out in AI investing. Funds that committed capital to AI companies in 2021 and 2022 — at the peak of the growth equity bubble — are sitting on portfolios where the marks have been revised downward and the path to liquidity is uncertain. Funds that deployed in 2023 and 2024, after the correction, are in a structurally better position. The secondaries market is the mechanism through which this vintage disparity gets resolved.

GP-led secondary transactions — continuation vehicles, single-asset deals, and strip sales — have become the dominant tool for managing legacy AI portfolios. The logic is straightforward: a GP with a 2021-vintage fund holding a portfolio of AI companies that are still growing but not yet ready for IPO or strategic sale can use a continuation vehicle to provide liquidity to LPs who need it, while transferring the assets to a new vehicle with a longer runway and fresh capital. For the GP, it preserves the relationship with the asset. For the exiting LP, it provides a clean exit at a negotiated price. For the incoming LP, it offers exposure to a curated set of assets with known history.

The pricing dynamics in AI secondaries are currently interesting. Bid-ask spreads have compressed as the market has matured, but there is still meaningful dispersion between assets that have demonstrated revenue traction and those that remain pre-revenue. Infrastructure-adjacent assets — companies providing compute, networking, or tooling to AI developers — are trading at premiums to pure-play application companies, reflecting the market's preference for assets with contracted revenue and lower binary risk.

At Lensbridge, secondaries have been a core part of our investment toolkit since we pioneered some of the earliest LP-led secondary transactions in Asia in the mid-2000s. The AI vintage problem is creating a new generation of opportunities in this space — and we are well-positioned to evaluate and execute on them with the rigour and experience that the market demands.