Enterprise AI spending boom masks fragile revenue for startups

The rapid expansion of artificial intelligence in the enterprise sector is disrupting traditional revenue models for startups, according to new research from venture capital firm Madrona. While market researcher IDC predicts companies will spend $4.25 trillion on technology in 2026, driven almost entirely by AI, the stability of that income is proving more precarious than previously anticipated.

Madrona’s survey of 150 enterprise IT professionals found that 74 per cent plan to expand their AI budgets over the next 12 months, with the remainder holding spending steady. However, the conversion of these budgets into long-term contracts remains elusive. Fewer than half of enterprise AI pilots currently reach full production, a significant improvement from the 95 per cent failure rate reported by MIT last year, but still a low bar for sustained commercial success.

The most significant finding is the erosion of the traditional multi-year contract, which historically provided a "moat of inertia" for software vendors. Madrona reports that 77 per cent of enterprises now re-evaluate their AI vendors every six months or on a rolling basis. This creates a "fast in, fast out" dynamic where switching costs are lower and the re-evaluation cadence is relentless, fundamentally altering the landscape for annual recurring revenue (ARR).

This shift complicates the narrative of rapid growth for AI startups, many of which have reported moving from zero to $10 million in revenue within three months. While enterprise trial budgets fuelled the initial boom in 2025, 2026 was expected to be the year customers settled into long-term commitments. Instead, revenue remains insecure even after a product graduates from the pilot phase, as enterprises retain the flexibility to switch providers frequently.

Further complicating revenue security is the preference for new pricing structures. Research from Andreessen Horowitz, which surveyed 50 technical AI buyers, indicates that more than half prefer pricing models tied to outcomes, such as work produced, rather than usage metrics like token consumption. Partners Tugce Erten and Sarah Wang note that charging for recognizable work, such as reports processed or leads generated, makes the product economically valuable to both sides.

This departure from SaaS-era usage models, which are based on employee counts or data storage, suggests a new era of enterprise experimentation. While this openness allows startups to gain entry more easily, it means that an enterprise contract no longer guarantees long-term revenue. Whether enterprises will revert to their traditional long-term buying habits remains to be seen.