Why "AI winter" keeps getting predicted
Every previous AI winter followed the same pattern: a wave of hype outran what the technology could actually deliver, funders got burned, and investment collapsed for years. People pattern-match today's AI boom to that history because the hype is real and highly visible, chatbots that write essays, agents that promise to run your business, valuations that seem to defy normal logic. It's a reasonable instinct. I just think it's wrong this time, for a specific reason: the gap between hype and delivered value has been closing, not widening.
The industry has already moved past the pure hype phase
What I tried to get across on the podcast is that AI stopped being a purely speculative bet some time ago. Real products, used by real companies, are already changing how work gets done in customer service, software development, logistics, and marketing, not as a promise on a slide but as something happening in production right now. Previous AI winters hit because the technology couldn't yet do what people expected of it. This time, the technology is doing quite a lot of what was promised, and in some areas doing more than most predictions expected a few years ago.
Growth that's sustainable, not just loud
I don't think today's momentum depends on hype holding up, which is the real test of whether a boom is a bubble. Revenue from AI products is real and growing, adoption inside large enterprises keeps climbing rather than stalling, and the applications multiplying across industries aren't dependent on any single company's stock price staying elevated. A winter happens when the money and attention were never backed by genuine utility. That's not the situation I see when I look across the industry today.
Why this matters beyond investor sentiment
The stakes in this argument aren't just about funding rounds. If people convince themselves a winter is coming and pull back research and safety investment along with speculative capital, that's exactly the wrong moment to lose momentum on the harder questions Conscium exists to work on, whether AI systems are behaving the way they should, and what happens as they get more capable. Sustained progress, not a boom-bust cycle, is what gives us time to get the safety work right alongside the commercial deployment.
The UKTN podcast, hosted by Jane Wakefield. Posted on 23 September 2025