Here's a statistic that should terrify enterprise software investors: 94% of companies say they'll continue spending on AI initiatives even after projects demonstrably fail. It's the definition of throwing good money after bad, and it reveals troubling dynamics about corporate AI adoption.
The finding comes from a Gartner survey of 2,500 IT executives globally, and it exposes the gap between AI hype and AI results. Companies are pouring billions into machine learning projects with no clear path to ROI, no success metrics, and apparently no willingness to pull the plug when things go wrong.
"This is classic sunk cost fallacy at industrial scale," said Svetlana Sicular, Gartner VP analyst. "Executives have publicly committed to AI transformation. Admitting failure isn't politically viable, so they double down."
The numbers are staggering. Enterprise AI spending is projected to hit $110 billion in 2026, up from $67 billion in 2023. Yet success rates remain dismal. Gartner estimates only 20% of AI pilots make it to production, and fewer than half of those deliver measurable business value within two years.
Why the disconnect? Corporate politics and fear of missing out. No CFO wants to explain to the board why they're not investing in AI when competitors are. No CTO wants to be blamed for falling behind on the "next big thing." So spending continues regardless of results.
The implications for enterprise software valuations are significant. Companies like Palantir, Snowflake, and Databricks trade at 15-25x revenue multiples based on assumptions of sustained AI spending growth. But if 94% of projects are zombies—alive on corporate budgets but delivering no value—how sustainable is that revenue?
We've seen this movie before. During the late-1990s dot-com boom, companies spent billions on enterprise software with vague promises of "digital transformation." When recession hit in 2001, those projects got axed overnight. Software valuations collapsed 70-90%.
The AI boom has similar characteristics: massive spending driven by FOMO, vague ROI justifications, and executives who can't admit failure without career consequences. The difference is today's projects have even less measurable output than Y2K-era enterprise deployments.
"At some point, boards will demand accountability," said Martin Casado, general partner at Andreessen Horowitz and former Nicira founder. "When recession or margin pressure forces companies to cut costs, AI projects without clear ROI will be first on the chopping block."
For now, the spending continues. The 94% figure suggests AI has become less about business results and more about corporate signaling. As long as money is cheap and growth expectations remain elevated, companies will keep funding experiments. But when the cycle turns—and it always does—the AI money pit could become a valuation graveyard.
