Startup ARR is less secure than ever, new research shows

Artificial intelligence is driving unprecedented shifts in enterprise IT. Companies that have historically maintained a cautious approach and committed to long-term technology investments are on track to spend $4.25 trillion on tech in 2026, market researcher IDC predicts, with the surge driven almost entirely by AI adoption.

New research from venture capital firm Madrona indicates that 74% of 150 enterprise IT professionals surveyed plan to increase their AI budgets over the next 12 months, while the remaining respondents intend to keep spending steady. Yet, despite expanding budgets, these same enterprises report that fewer than half of their AI pilot projects ever graduate into full production.

This completion rate actually marks an improvement over the previous year. MIT famously reported that 95% of enterprise AI projects failed to yield a positive return on investment. While a success rate below 50% remains modest, it still outperforms a dismal 5% benchmark.

Stock photo for illustration only, not from the actual event

However, the most striking insight from the Madrona report is that even when enterprises successfully deploy AI technology, they routinely withhold long-term commitments, shaking the foundation of fast-growing annual recurring revenue (ARR) figures reported by startups.

Enterprise trial budgets initially fueled the AI boom of 2025. This year was anticipated to be the period when major customers settled down and committed long-term to AI startups—contracts that previously allowed many early-stage companies to claim astronomical revenue growth, such as scaling from $0 to $10 million in three months.

Yet, for the first time, enterprise revenue remains insecure even after an AI product successfully exits its pilot phase and achieves corporate adoption. Part of the challenge stems from AI startups struggling to establish effective pricing frameworks for enterprise buyers.

Recent research from VC firm Andreessen Horowitz, surveying 50 technical AI buyers, revealed that over half prefer tying AI expenses to actual work output or business outcomes rather than usage metrics like token consumption.

Charging based on usage metrics like tokens mirrors legacy SaaS business models. Once an enterprise identifies a need for email systems, HR software, or cloud storage, expenditure scales strictly according to employee headcounts or data volume.

"More than half of technical AI buyers want fees tied to the work produced or other outcomes, rather than to usage like the number of tokens consumed."

Andreessen Horowitz Research

Ultimately, AI has ushered in a prolonged era of enterprise experimentation. While this openness lowers entry barriers for startups seeking initial trials, enterprise contracts no longer guarantee secure long-term revenue streams, leaving open the question of when or if traditional corporate procurement habits will return.

Source: TechCrunch