Startup Valuation Models in the AI Cycle: What Has Actually Changed
Revenue multiples, ARR growth bands and burn-multiple discipline have all been re-written by the AI capital cycle. A working framework for separating durable franchises from cycle artefacts.
The AI cycle has produced a generation of companies whose growth rates do not fit the valuation frameworks venture investors spent the last decade refining. Revenue multiples that would have been considered indefensible in 2021 are routinely paid in 2026 — and in many cases the underlying businesses are growing into them.
The useful diagnostic is not the multiple itself, but the composition of revenue underneath it. Recurring software revenue from production deployments inside enterprise customers continues to deserve a premium. Consumption revenue tied to model usage deserves a more cautious read, because consumption can compress as model costs fall. And revenue that depends on a single foundation model partner deserves the most cautious read of all.
Burn-multiple discipline has returned as a frame, but with a wrinkle: the AI cohort tends to burn capital on infrastructure rather than headcount, and infrastructure burn produces a different operating-leverage profile as the company scales. Investors who underwrote the previous cycle's headcount-burn template have had to relearn the math.
The companies that will compound through the next decade are unlikely to be the ones with the most aggressive 2026 multiples. They will be the ones whose revenue composition, gross margin trajectory and customer concentration look most like durable enterprise software — independent of how AI-native their narrative happens to be.
More from Research
Real World Assets: The Institutional Thesis for Bringing Off-Chain Value On-Chain
From treasuries and private credit to real estate and commodities, the tokenization of real-world assets is moving from thought-experiment to balance-sheet reality at the largest financial institutions in the world.
Tokenized Real Estate: From Pilot Project to Recognised Asset Class
After half a decade of pilots, the legal, custodial and platform infrastructure required for institutionally credible tokenized real estate is finally arriving — and reshaping how property is owned, traded and financed.
What Is RWA? A Primer for Institutional Allocators
A working definition of Real World Assets, the categories most active today, and the operational questions allocators should be asking before committing capital.