Back in September, we introduced the Stripe SaaS Index, a same-business measure of the revenue health of non-AI software-as-a-service (SaaS) businesses. The index showed that software revenues had not only survived the “SaaSpocalypse” of early 2026 but in fact accelerated and thrived.
Today, we are releasing a counterpart to the SaaS Index, the Stripe AI Index. Like the SaaS Index, it tracks payments to businesses on Stripe in a way that is robust to Stripe’s own business mix and also prevents any one company from being overly dominant. Unlike the SaaS Index, however, the AI Index covers businesses whose core product or service depends on artificial intelligence—for example, an AI coding assistant, an image generator, or a service built specifically to support AI companies.1 It covers nearly 20,000 such businesses weekly. It is a proxy for the health of AI companies, but as with the SaaS Index, since companies can earn revenues outside of Stripe, it cannot replace the signal from audited revenue statements or other company data.
A few observations follow from our new index. First, global same-business AI revenues have seen explosive growth over the last year, with perhaps some signs of cooling on a higher-frequency basis. Growth over the 13 weeks ending November 8 was about 80% at an annual rate, lower than its 140% pace just a few months ago and lower than the 108% annualized 26-week growth.
Second, like the SaaS Index, the typical AI business has seen lower (but still positive) growth than the aggregate. The Stripe Median AI Index rose 41% over the 52 weeks ending November 8, compared with 103% for the aggregate. Larger businesses appear to be doing better, though the typical AI business’s revenue growth has been consistent throughout 2026.
Third, the US leads the four geographic groups in annual growth. Over the 52 weeks ending November 8, payment activity rose roughly 104% for AI businesses located in the US, 83% in the UK, 67% in the European Union (EU), and 76% in the rest of the world.2 The wedge between US and non-US growth appears to be widening; this is even clearer when looking at the levels rather than the growth rates of the AI Index by region.
Fourth and finally, AI businesses are growing faster than non-AI SaaS businesses, even at the median. The Stripe AI Index aggregate rose 103% over the 52 weeks ending November 8, compared with 33% for the Stripe SaaS Index. The median AI and SaaS indices rose 41% and 15%, respectively, over those same periods.
We think this new AI Index will prove to be a useful high-frequency barometer of the health of the industry. Going forward, we will make it available on Stripe Economics, updated weekly alongside our SaaS Index.
Methodological Appendix
Population and payment measure
The Stripe AI Index adapts the weekly Stripe SaaS Index to AI-native organizations. Eligibility requires the current Stripe organization-level AI-company classification, a canonical domain, and a first Stripe charge date. AI-native describes businesses building AI products; using AI tools alone does not qualify a business. The payment measure is pay-ins volume (PIV: net of refunds and gross of chargebacks), using the source metric's fixed exchange rates. Payments processed elsewhere, accounting revenue recognition, costs, profits, and inflation are outside the measure.
Weekly Fisher indices
Weeks run Monday through Sunday. For each pair of adjacent weeks, construct the earlier-week and later-week organization sets. Set an organization's missing opposite-week payment amount to zero, matching the SaaS Index convention. The first leg divides current payments for earlier-week organizations by their earlier payments. The second leg divides current payments for later-week organizations by their earlier payments. The Fisher relative is the geometric mean of the two positive legs, chained over time.
The flagship Fisher index is PIV-weighted, with an explicit cap that no business may comprise more than 5% of aggregate PIV each week. The median diagnostic instead chains the unweighted approximate median of current/prior payment ratios among organizations with positive payments in both weeks.
Seasonal adjustment and filtered trend
Each series is fitted independently with multiplicative BOIWSA in R, version 1.1.4, using r = 0.8, corrected Akaike information criterion (AICc) model selection, automatic additive-outlier search, and the SaaS outlier threshold of 3.8. Automatic outlier regression is retained as part of seasonal estimation; it does not cap or replace the organization-level contributions used to construct the raw Fisher index.
After BOIWSA, the SaaS logged smooth-trend unobserved-components model uses exact-diffuse initialization, and it selects among a stochastic 52-week seasonal component, a one-harmonic 52-week trigonometric component, and a nonseasonal fallback. A seasonal form must converge in both the full and cutoff samples and pass the existing seasonal-cancellation, recurrence, and localized-distortion guards. The chart displays the exponentiated one-sided filtered log-level state. It does not display the two-sided smoothed state. Estimated parameters and seasonal factors still use the full sample, making the history revisable rather than a real-time vintage.
We identify these businesses through automated reviews of their websites, supplemented by human review. Note that a conventional business adding an AI feature or using AI internally would generally not qualify. Moreover, the AI Index measures all payments to qualifying AI businesses rather than transactions for AI products.
We classify AI businesses by their geographic location; revenue growth reflects global demand for those businesses.


