Why 90% of AI Startups Will Be Dead in a 2026

We’re drowning in AI tools. An AI for writing emails, an AI for planning your holiday, an AI for debugging your code. It feels like an explosion of innovation. But what if most of these tools are already ghosts at the feast, made obsolete by a change that’s happening right now?

There’s a clear historical precedent for this. In the early days of the PC, a whole industry existed to sell utility software for Windows. There were tools that compressed files or defragmented hard drives. With each new version, Microsoft simply absorbed those features into the operating system, making entire categories of software redundant overnight.

This didn’t kill all software, of course. Complex applications that solved specific problems, like Photoshop, thrived because they weren’t just simple utilities, they created new capabilities on top of the operating system. The same distinction applies today. The companies at risk are the ‘AI utilities’, the thin wrappers that just offer a single function of a large language model.

This is why the AI bubble is going to pop. Not because the technology is a fad, it’s the most important change in a generation. It will pop because 95% of the AI businesses currently scrambling for funding are building simple utilities destined to be absorbed by the foundational platforms. The 5% that survive won’t lead the market, they will be the market, providing a universal capability that makes thousands of specialist tools obsolete overnight.

The Detonator

The latest disruptions. Your inbox. Once a month

We don’t spam! Read our privacy policy for more info.