The Generative AI Fizzle™
Disclaimer: Anything can happen at anytime in the market; I don’t give stock picks, and as the saying goes, the market can remain irrational longer than you can remain solvent.
Still:
My joke/warning about SpaceX and tulip mania (look it up if you don’t know what it is) 9 days ago:
Gary Marcus@GaryMarcussouvenir $SPCX tulips t-shirt
3:51 PM · Jun 16, 2026 · 2.7K Views1 Reply · 5 Reposts · 12 Likes
That actually looks kinda on track so far:
Thoughts and prayers for those that bought 10 days ago at 225.
Yesterday I coined a new term, the Generative AI Fizzle™. There is no doubt in my mind that a lot of things in AI are overvalued, but that doesn’t mean they will deflate at all once. They might, but maybe what we will see is more like a slow decline, as investors lose enthusiasm for the ratio between hype (insane) and profits (not that impressive thus far). Along these lines, I made a tweet last night, based on yesterday’s market:
Of course one day’s results don’t tell us much, but for what it’s worth most of those stocks are down again today (except Micron which just had a big bounce after a great earnings report).
and most of them are down significantly for the month:
Maybe this is how GenAI declines, not with a bang, but a whimper?
To be sure I don’t expect LLMs to disappear anytime soon but the valuations and market caps we are currently seeing may eventually be seen as ludicrous.
Tulips didn’t disappear; astronomical tulip prices did.
For a really long time I have been warning that the almost-universal obsession across the AI industry with LLMs meant basically (i) no technical moat for anyone, and (ii) that would make profits hard to come from. I warned (iii) that LLMs would become commodities, and (iv) that we would expect price wars. I have been banging that drum here for three years, as in this January 2024 warning about OpenAI:
In my view, the world has indeed gone in this direction. LLMs have become commodities, prices wars have been significant, profits for the LLM providers have been elusive. Yesterday a new Chinese, open-source model came out that might make all of this drastically worse for the US LLM companies:
George Noble@gnoble79Big Tech's $1 trillion AI moat just got DESTROYED by a free Chinese download. Microsoft, Amazon, Google, and Meta are pouring fortunes into chips and data centers because they have been told that whoever builds the biggest model wins, and that lead becomes a fortress no one can11:10 PM · Jun 24, 2026 · 53.6K Views43 Replies · 99 Reposts · 330 Likes
§
And yeah, profitability is still a struggle:
(By the way, Elson interviewed me yesterday for a lively episode of the Prof G Pod that will appear Friday.)
§
If GenAI-mania does fall apart, slowly or suddenly, it’s a big deal, especially given the growing amount of debt.
How big? Here a quote I just saw via Bloomberg:
Even if the bubble doesn’t pop, maybe it has already started to fizzle?
Stay tuned.
Also, watch for a new oped from yours truly in the Financial Times, coming very soon.
As I told someone at work yesterday, they have no moat and no business model, but they do have alllll the money.
As Karen Hao pointed out, it was in their interests to believe that scale is all you need. We need all the money to build AI god, and everyone who believes us can take a cut, hey Goldman 👋 Hey VCs 👋
I hope it is slow instead of a crash, maybe it will hit us all with less velocity.
It seems like, while insisting that we must do this before China, they just handed all our IP to China via their lack of moat. Sigh.
We'll see I guess, markets are unpredictable.
Any of the businesses involved in the ML & LLM investment bubble (eg Micron) with better earnings, where they're just selling hardware to intermediaries, but there's no end user demand to ever actually pay for it...
Are just making the investment bubble and its consequences worse and worse... More losses for someone coming in future...
This insanity has carried on for 3+ years already, so you would be a fool to guess exactly when it'll end.