Startup ARR is less secure than ever, research shows

Enterprise IT spending is on pace to hit $4.25 trillion in 2026, according to market researcher IDC, almost all of it driven by AI. New research from venture capital firm Madrona finds that 74% of 150 enterprise IT professionals it surveyed plan to expand their AI budgets over the next 12 months, with the rest planning to hold spending steady — yet these same enterprises say fewer than half of their AI pilots ever reach full production.

That’s actually an improvement: MIT reported last year that 95% of enterprise AI projects had failed to deliver ROI, making a sub-50% pilot-to-production rate a low bar that’s still better than a 5% success rate. The more telling finding, according to Madrona, is that enterprises don’t commit to AI vendors even once they do roll them out: some 77% reevaluate their AI vendors every six months or on a rolling basis. “This creates a ‘fast in, fast out’ dynamic that is fundamentally different from traditional enterprise SaaS, where multi-year contracts provided a moat of inertia,” Madrona wrote in the report. “In enterprise AI, switching costs are lower and the re-evaluation cadence is relentless.”

That dynamic has major implications for the fast-growing annual recurring revenue figures AI startups report, including the now-familiar trope of a startup going from $0 to $10 million in three months. Enterprise trial budgets fueled the initial AI boom of 2025, and this year was supposed to be when big customers settled in and committed to AI startups long term. Instead, enterprise revenue remains insecure even after a product graduates from pilot to adoption.

Part of the problem is that many AI startups haven’t settled on a workable enterprise pricing model. Separate research from venture firm Andreessen Horowitz, surveying 50 technical AI buyers, found that more than half want AI fees tied to the work produced or other outcomes rather than to usage metrics like tokens consumed — a shift from the SaaS-era model of charging per employee or per unit of data. Pricing around recognizable work, a16z partners wrote, is what makes a product “economically valuable to both sides,” rather than simply billing for consumption once a customer already knows it needs the tool.

Why it matters

If enterprise AI contracts keep behaving like short-term trials rather than SaaS-style commitments, the sky-high revenue growth many AI startups tout could prove far less durable than their valuations assume.