How AI Is Quietly Reshaping Venture Capital

A new generation of funds is betting entire portfolios on foundation models — and rewriting the rules of due diligence in the process.

Venture capital has always moved in cycles, but the current shift toward AI-first portfolios is different in both scale and speed. Several major funds have quietly restructured their entire investment thesis around foundation model infrastructure, leaving traditional sector diversification behind.

The change is most visible in how due diligence is conducted. Where investors once spent weeks modeling unit economics and customer acquisition costs, many are now evaluating compute efficiency, model architecture decisions, and data moats — categories that barely existed in venture vocabulary three years ago.

Critics argue this concentration creates systemic risk: if foundation model economics shift unexpectedly, entire fund portfolios could be exposed simultaneously. Proponents counter that the upside asymmetry justifies the bet, pointing to historical precedent in mobile and cloud computing waves.

What's clear is that the funds making this pivot early are already seeing it reflected in their fundraising conversations with limited partners — for better or worse.