# Yes, AI is a financial bubble. No, that won’t stop the technology.
Format: Note
Published: October 2026 (2026-10-19)
Canonical: https://blog.narbeh.me/the-ai-bubble
---
A financial bubble and a technological revolution can co-exist. Investors can lose fortunes and companies can go bankrupt without the technology losing its ability to change how we live and work.

A bubble is a good idea taken too far. In AI, we are investing as if technical breakthroughs translate almost immediately into widespread adoption and extraordinary profits. The data from the last few years, however, shows that the real-world application of these advancements is significantly lagging. Models capable of generating a large economic surplus were already available a year or two ago, though economic models estimated that with current adoption and application, AI would add no more than 1% to total factor productivity over ten years. We can disagree with that forecast and still question how much future growth is already priced in today.

The financial bubble created around AI technologies will burst sooner or later. Nobody knows when, and the unwinding will probably take years. Tighter funding, bankruptcies and cancelled investments will follow. The consequences will spread beyond tech, and the bust will matter for investors, lenders and workers.

However, it will not undo what the technology can already do. The internet kept developing after the dot-com crash, and the foundational infrastructure remained available even after the companies that installed it went bankrupt. Similarly, AI’s capabilities and useful infrastructure will outlast failed investments and hype-cycled products. The data centres, power and grid capacity, models and skills will remain useful after the bust.

What comes after will be consolidation and rebuilding, with more focus on applying AI to business cases that actually make sense. The excess compute capacity will make things cheaper for whoever builds during this time. We will then have an actual widespread rapid advancement period where adoption catches up with what the technology has long been capable of.