AI startups are raising capital at a record pace. According to Crunchbase data, artificial intelligence firms have already raised $118 billion worldwide in 2025. So far, the effectiveness looks impressive. AI startups are seeing excellent revenue growth and even $100 million ARR milestone is often achieved.
While this growth is breathtaking, some analysts are starting to query its sustainability. They warn that spending on artificial intelligence may soon peak and that is it unprofitable technology firms could also be hit hardest when the cycle reverses. If this happens, many AI investors will find themselves in a difficult situation.
Predicting a bubble is rarely productive, but preparing for volatility is. It can be smart for each founders and investors to be sure that portfolio firms are sufficiently resilient to face up to a potential market shock.
The key is to evaluate the sustainability of the ARR. In the event of a serious economic downturn, the “development game” quickly becomes a game of survival. History suggests that while a few firms may proceed to grow more slowly, most will struggle or disappear.
The query, then, is tips on how to distinguish sustainable ARR from noise-driven ARR.
What distinguishes sustainable ARR from noise?
Several aspects separate real, sustainable revenue growth from the hype.
The first is customer engagement. Sustainable revenue comes from multi-year contracts, recurring renewal cycles and budgeted spend across core IT or operating lines. When revenues depend on pilot projects, proofs of concept, or amorphous “innovation” budgets, they’ll disappear as corporate priorities change. An organization that touts these short trial periods as ARR is actually seeing momentum, not recurring revenue.
That’s what an investor does Jamin Ball he called experimental fixed revenues.
Traditional software firms can thrive with monthly churn in the low single digits – I think 5% to 7%. But many AI firms do seeing double. This means they have to sprint to remain put, continually replacing users as they move on to the next shiny tool.
Another differentiator? Workflow integration and depth. Sustainable ARR is built into core workflows, data pipelines, or multiple customer teams. It can be expensive and tedious to tear it out. Hype ARR, on the other hand, lives on the surface – lightweight integrations, quick implementations and a limited number of stakeholders. Without unique mental property or deep workflow integration, such products will be replaced with minimal friction.
Finally, real growth is defined by clear added value. True ARR is supported by measurable ROI, well-defined outcomes and long-term customer motion plans.
However, the noise around ARR is because of urgency (we’d like to indicate our shareholders the implementation of artificial intelligence as soon as possible) or an undefined return on investment. In such cases, customers don’t even know tips on how to define success. They test, not engage.
Outside ARR
It’s vital to place ARR traction in context. Investors and founders should focus on a broader set of metrics – conversion from pilot projects to long-term contracts, contract length and expansion, net revenue retention and gross margin trajectory. These indicators show whether growth is sustainable.
It would even be helpful to evaluate the actual impact of the product: e.g. productivity gains (more code, content or customer conversations per working hour), accuracy improvements (e.g. in detecting bad actors), and higher conversion rates. These metrics should exceed customer expectations and outperform alternative tools. This is what signals real value creation and a greater likelihood of converting experimental revenues into sustainable ARR.
After all, artificial intelligence may change the speed at which firms are created and grow, but it has not suspended the basic laws of business.
For founders, the message is easy: rejoice ARR if you wish, but couple it with proof of sustainability, profitability and defensibility. Investors should resist the temptation to chase every striking price. The real competitive advantage in the next phase of AI is stability, not spectacle.