For over a decade, SaaS valuation ran on a familiar script: annual recurring revenue, net dollar retention, and a multiple pulled from whatever the last comparable deal fetched. That script is being rewritten in real time, and the authors are not entirely sure of the ending.
Artificial intelligence has not just changed what startups build — it has changed how quickly they can build it, how cheaply they can serve customers, and how defensible any single feature really is. The result is a valuation environment where the old inputs still matter, but no longer add up the same way.
“We’re not throwing out the SaaS playbook. We’re asking it questions it was never built to answer.”
The multiple compression nobody saw coming
Through most of 2024 and 2025, AI-native companies commanded a premium over traditional SaaS peers. That premium has started to narrow, and in some categories, reverse. Investors we spoke with across three continents pointed to the same underlying tension: if a competitor can replicate a workflow using an off-the-shelf model in a weekend, how much is the workflow itself actually worth?
What's replacing the old signal is a sharper focus on proprietary data, distribution advantages, and switching costs that have nothing to do with the underlying model. Firms like Sequoia and Insight Partners have quietly rewritten parts of their diligence memos to ask less about what a product does and more about what it knows that a competitor cannot easily learn.
Three questions every term sheet now asks
- What happens to your margins when inference costs fall by 90% — does your moat shrink with them?
- How much of your retention is habit, and how much is genuine switching cost?
- Could a well-funded incumbent rebuild your core loop in two quarters?
None of this means the fundamentals of company-building have changed. Founders who can answer those three questions with evidence, rather than conviction, are still closing rounds at strong terms. The difference is that "AI-powered" is no longer, on its own, a story that clears diligence.
What this means for founders raising in 2026
The practical takeaway from every investor conversation was consistent: lead with retention and unit economics, not model architecture. Boards are increasingly skeptical of pitches that spend their first ten slides on capability and their last two on numbers. Flip that order, and the capital tends to follow.
The firms moving fastest right now are the ones treating this as a genuine repricing of risk rather than a temporary correction — building financial models that assume commoditizing infrastructure costs, and betting the venture math on everything built on top of it.
