Enterprise customers are no longer sticking with AI vendors for the long haul, fundamentally destabilizing the revenue models that have powered AI startups' explosive growth. According to new research from venture capital firm Madrona, 77% of enterprises reevaluate their AI vendors every six months or even on a rolling basis, creating what researchers call a "fast in, fast out" dynamic that's radically different from traditional enterprise software deals.

This shift matters enormously because AI startups have built their valuations on the promise of predictable, long-term recurring revenue. The phenomenon of startups claiming they went from zero to $10 million in annual recurring revenue (ARR) in just three months relies on enterprises committing to multi-year contracts. But that's no longer happening. "Enterprise trial budgets fueled the initial AI boom of 2025," explains the research. "This year was supposed to be the year these big customers settled in and started committing long term to AI startups. Yet, for the first time ever, enterprise revenue remains insecure, even after a startup's AI product graduates out of a pilot phase and gets adopted by a company".

The good news for enterprises is real: fewer than half of their AI pilots make it into full production, which is actually an improvement from last year when MIT reported that 95% of enterprise AI projects failed to deliver return on investment. But the bad news for startups is equally real. When enterprises do roll out AI technology, they're treating it like a trial that never quite ends.

What's Driving the Constant Reevaluation?

Part of the problem stems from how AI startups price their products. New research from Andreessen Horowitz surveyed 50 technical AI buyers and found that more than half of them want AI fees tied to the work produced or other outcomes, rather than to usage metrics like the number of tokens consumed. Tokens are the small chunks of text that AI models process; pricing by tokens is essentially a carryover from traditional software-as-a-service (SaaS) business models.

The disconnect is significant. With traditional enterprise software like email or HR platforms, once a company knows it needs the product, pricing is straightforward: you pay based on how many employees use it or how much data you store. But with AI, enterprises struggle to understand whether they're getting real value for their money. "Charging for usage like tokens is basically a SaaS-era business model," the research notes. "For AI, pricing around the recognizable work is what helps the startup prove its worth to the customer".

"When the fees revolve around, say, how many reports are processed, or tickets closed, or leads generated, this makes the product economically valuable to both sides," stated Tugce Erten and Sarah Wang, partners at Andreessen Horowitz.

Tugce Erten and Sarah Wang, Partners at Andreessen Horowitz

How Can AI Startups Stabilize Their Revenue?

• Outcome-Based Pricing: Shift from charging per token or per API call to pricing based on measurable business results, such as the number of customer support tickets resolved or leads generated, which gives enterprises clearer ROI visibility.
• Longer Pilot Periods with Clear Metrics: Design pilot programs that establish specific success criteria and timelines, helping enterprises feel confident enough to commit to longer contracts once they see results.
• Vertical-Specific Solutions: Build AI products tailored to specific industries or use cases rather than horizontal platforms, making it harder for enterprises to justify switching vendors every six months.

The implications are profound for the entire AI investment landscape. Madrona's survey of 150 enterprise IT professionals found that 74% plan to expand their AI budgets in the next 12 months, with the rest holding spending steady. That's a massive opportunity, but only if startups can convert trial customers into committed ones. The challenge is that enterprises now have options. The AI market is crowded, switching costs are low, and reevaluation cycles are relentless.

This creates a paradox: the very factors that made AI startups attractive to investors in 2025 and early 2026 may not hold up under scrutiny. Fast revenue growth looks impressive on a pitch deck, but if that revenue can evaporate in six months when a customer reevaluates, the underlying business model is fragile. Startups that can prove durable, outcome-driven value to their customers will likely survive this transition. Those that can't may find their ARR numbers tell a very different story than their growth rates suggest.