AI Hype Cycle

DW #128 🟡

Author

Ramsey Shaffer
September 21, 2025 • Estimated Reading Time: 3 minutes

Some are saying the AI bubble is popping. I’m not convinced.

The numbers show AI adoption at many firms declining month over month since June.1 Others claim the business model for the big AI companies is a ‘ponzi scheme’, that they are throttling the quality of AI responses at peak hours to save money.2,3  

But tech adoption has followed a hype cycle since the beginning of civilization - AI being maybe the most transformative, polarizing technology since the invention of fire, it’s hype cycle will be duly extreme.

Here’s my two cents, I think we’re here:

Probably somewhere around the 2nd or 3rd inning - by this point every big firm has tried AI, experienced how unfathomably good it’s gotten, spent their R&D budgets fiddling with different flavors of it, and are now planning their transformation strategy for the next 3-5 years.

As the R&D budgets come down to earth and firms settle in to their workflows, the AI hype will come down with it. That’s what we’re seeing now.

The big AI companies will continue raising stupid-large funding rounds to train increasingly larger models, the marginal increase in performance with each launch decrease.

Will they achieve ‘AGI’ in the next 5 years? Who knows - I’m not convinced the we have the right architectures and enough good training data yet to get there.

But either way, AI isn’t going anywhere. This is still the beginning.

Bet your bottom dollar that over the longer arc all the big firms are doubling down on AI. They are buying up every GPU, board members are stepping in to say they need centralized data teams, they are shuffling internal org charts to accommodate the new AI future.

Hedge funds in particular have historically been a good canary in the coal mine of tech adoption. Some of the big firms we've met with recently are pausing new data projects entirely for a bit while they sort out how to reorganize their data teams for firm-wide AI-first initiatives.

And in the short view this might feel like a ‘recession indicator’ or whatever. But I think realistically it’s the opposite. We are about to see unprecedented investment in AI across all the big firms. Energy demand in the US will increase by multiple percentage points by the end of the decade for data processing alone.

The sea goes out before the tsunami. Steering cruise ship sized organizations take time. Month over month AI adoption rates is not the best proxy for the long-term trend.

Peace,
Ramsey