“Why did I argue the “yes, bubble” side if I don’t believe it? Because the best way to understand your own position is to steel-man the opposing argument.”
100% on of the best more general pieces of advice and takeaways here. thanks for the read, enjoyed this one.
I sincerely hope it doesn’t collapse in the true sense of the word, but corrections tend to happen in every market when a new technology is introduced.
Thanks Joel and Laura, for providing both evidence and frameworks for a meaningful dialogue around a topic that people seem very happy to normally come firmly down on one side of the fence or the other arm. As you make quite clear here, it's much more nuanced than that. Even though I tend to side with Laura in thinking that it's too embedded in Enterprise for it to burst, the AI 'bubble' is certainly going to become much more malleable over the next 12 months.
Ha, the tulip mania callback 😄 Every time there's a new technology, someone brings up the tulips. But Laura's point is right, tulips were pure speculation. AI is already inside operations. You can't just walk away from your cloud infrastructure, IMO.
I agree that it is currently not a bubble. AI usage is climbing fast and is set to continue this way.
On the other side I'm not sure if I would relate the current stock market to the AI bubble. The entire US stock market is currently overvalued by itself and a correction of 50% in market value is currently very possible. It may have already happened if it wasn't for the tech companies driving the stock market even higher.
BTW, forgot to mention, between 50% and 90% of the current US GDP growth is estimated to depend on the infrastructure investments done by the big tech companies, if they stop the investment a recession could be on the horizon, plus you would probably have a collapse of the tech stocks for the same reason.
The big seven big tech companies currently account for around 30% of the entire SP500 index
Perplexity is a good tool to research these things.
Here are some basic data:
- of the current GDP growth one part is driven by the tech spending and the other part is driven by consumer spending because of the money that is earned in the stock market growth. So if the stock market declines ( will happen if AI spending stops ) there would be big consequences
- the job market is bad and salaries are not increasing. At the current level it is not yet a big problem but if the data may get worse in 2026 it could be a problem for the economy
- Consumption is misaligned with income, supported by the wealth effect of the stock market growth (vulnerable to equity shocks)
- Tariffs and supply chains: Tariff risk is the big unknown for 2026. Companies with exposure to imports should hedge.
Basically, a recession is not expected in 2026 if anything goes well, it may happen in 2027-2028 if job market continues to shrink.
But if there is a huge correction of the stock market it could possibly lead versus a recession by itself in relation with the other data
I love this new format Joel! I do believe that there are unrealistic valuations at the moment not supported by strong fundamentals or real world impact. In 2026, I think we are going to see a correction in that investors start asking “what has happened to my money”, but AI is becoming too foundational to the modern economy to collapse.
My view is that ‘AI wrappers” die off while the underlying infrastructure, enterprise adoption, and productivity gains keep compounding in the background.
My take is that the jagged frontier is going to get even more stark, and you're going to witness real agentification across orgs in major companies. Some in the same industry will unlock real scale; otherwise will simply lag. A rotation is on the horizon.
As I see it, the current infrastucture bets are projected for years ahead, currently they need to invest on more infrastructure because basically all their computing power is currently used.
If you predict a possible 5-10X increase in AI usage per year in the next 2 years, they in fact need this new infrastructure.
But the big spending in infrastructure is projected for the next 5-7 years ( OpenAI ), they can still cut that spending in future if it will not be needed anymore.
Very thoughtfully laid out arguments! Thank you, Laura and Joel, for sharing.
I think that the current AI situation looks very similar to what was happening in the biotech industry back in 2021/2022. With advanced technologies like gene therapy and gene editing becoming better understood and approved by regulators, there was almost this "gold rush" to set up another gene or cell therapy company.
But that sort of R&D is expensive, never mind the low probability of clinical success, so the funding and valuations for these quickly went down after the initial spike of excitement. But this was actually a "blessing in disguise" – companies that didn't have strong scientific and financial foundations went under, but those with good evidence and fiscal discipline actually weathered the storm.
So with AI, it might play out similarly: companies with weaker AI market positioning will likely disappear, but that gives more opportunity for stronger players to reassess and adjust their current strategies to make sure they come out on top.
The bubble feels real because adoption is being mistaken for competence. Using AI reduces surface effort, but it does not reduce the need for judgment. In fact, it raises the cost of bad judgment because errors now scale faster.
“Why did I argue the “yes, bubble” side if I don’t believe it? Because the best way to understand your own position is to steel-man the opposing argument.”
100% on of the best more general pieces of advice and takeaways here. thanks for the read, enjoyed this one.
So glad! Hope you’re doing well!
If I were to sum up my thoughts in three words: sunk cost fallacy.
Yes!!🙌
Absolutely!
Reality usually sits somewhere between hype and collapse.
I sincerely hope it doesn’t collapse in the true sense of the word, but corrections tend to happen in every market when a new technology is introduced.
Thanks for reading, John!
Sure thing, Laura!
Thanks Joel and Laura, for providing both evidence and frameworks for a meaningful dialogue around a topic that people seem very happy to normally come firmly down on one side of the fence or the other arm. As you make quite clear here, it's much more nuanced than that. Even though I tend to side with Laura in thinking that it's too embedded in Enterprise for it to burst, the AI 'bubble' is certainly going to become much more malleable over the next 12 months.
I also side with Laura ;) I just needed to present the counterargument haha
Thanks for always reading my work, Sam! And I do agree that the next 12 months are decisive.
Ha, the tulip mania callback 😄 Every time there's a new technology, someone brings up the tulips. But Laura's point is right, tulips were pure speculation. AI is already inside operations. You can't just walk away from your cloud infrastructure, IMO.
Very interesting conversation!
Very much agree with you here :)
I really don’t see a way to start a conversation about bubbles without mentioning the first one (though I agree everyone brings up tulips 😅).
Thanks for reading, Mia!
I agree that it is currently not a bubble. AI usage is climbing fast and is set to continue this way.
On the other side I'm not sure if I would relate the current stock market to the AI bubble. The entire US stock market is currently overvalued by itself and a correction of 50% in market value is currently very possible. It may have already happened if it wasn't for the tech companies driving the stock market even higher.
Yeah I don't think it's coming, but I really don't think a 50% stock market is in the books either. Guess we'll see!
Well, the basic numbers about the stock market are:
TRAILING P/E (Last 12 months):
Current: 31.22
Average (1957-2026): 19.69
Peak 1999 (dot-com): ~29
Peak 2000 (peak): 27.55
Peak 2008 (pre-crash): 70.91
2022 (minimum): 22.82
Conclusion: +58% over average
------
FORWARD P/E (Next 12 months):
Current: 22-23.76
Average 30Y: 17.1
Peak 1999 (bubble): ~25
Peak 2021 rally: ~22
Conclusion: +29-39% over average
-----
Ratio Market Cap / GDP (Buffett Indicator)
CURRENT (January 2026):
Buffett Indicator: 222.9-230%
Hist. average (1957-2026): 79.9%
Long average(1960s-2020s): 153%
"Good" range: 126-180%
Peak 2000 (dot-com): ~204.8%
Current peak: NEW MAXIMUM!
Valuation Interpretation:
- 79.9% = Underpriced
- 153% = Average Fair Value
- 180% = Slightly Overpriced
- 230% = Extremely Overpriced
----
So, estimared overvaluation is 20-30% (with forward P/E and Buffett Indicator as main indicators)
If confidence in mega-cap tech breaks or recession: 40-50% similar to 2008/2000
BTW, forgot to mention, between 50% and 90% of the current US GDP growth is estimated to depend on the infrastructure investments done by the big tech companies, if they stop the investment a recession could be on the horizon, plus you would probably have a collapse of the tech stocks for the same reason.
The big seven big tech companies currently account for around 30% of the entire SP500 index
That’s a really interesting perspective, Matija. I wonder what other obstacles might be preventing this correction, besides the tech companies.
Perplexity is a good tool to research these things.
Here are some basic data:
- of the current GDP growth one part is driven by the tech spending and the other part is driven by consumer spending because of the money that is earned in the stock market growth. So if the stock market declines ( will happen if AI spending stops ) there would be big consequences
- the job market is bad and salaries are not increasing. At the current level it is not yet a big problem but if the data may get worse in 2026 it could be a problem for the economy
- Consumption is misaligned with income, supported by the wealth effect of the stock market growth (vulnerable to equity shocks)
- Tariffs and supply chains: Tariff risk is the big unknown for 2026. Companies with exposure to imports should hedge.
Basically, a recession is not expected in 2026 if anything goes well, it may happen in 2027-2028 if job market continues to shrink.
But if there is a huge correction of the stock market it could possibly lead versus a recession by itself in relation with the other data
Thanks for writing this, it clarifies alot, and it's super important to get both perspectives laid out so clearly instead of all the hype.
Yes! And so rare to find, doesn’t get as many clicks as doom
Appreciate the balanced, data-driven framing that separates real AI risk from hype and helps readers think beyond extremes
Thanks, Petar!
I love this new format Joel! I do believe that there are unrealistic valuations at the moment not supported by strong fundamentals or real world impact. In 2026, I think we are going to see a correction in that investors start asking “what has happened to my money”, but AI is becoming too foundational to the modern economy to collapse.
Would love to see more debates.
Glad to hear it!! I'm glad it resonated, I'll need to think of a few more.
My view is that ‘AI wrappers” die off while the underlying infrastructure, enterprise adoption, and productivity gains keep compounding in the background.
That seems most likely For me too
My take is that the jagged frontier is going to get even more stark, and you're going to witness real agentification across orgs in major companies. Some in the same industry will unlock real scale; otherwise will simply lag. A rotation is on the horizon.
That will be really interesting to see, Raj!
Locked in as vested interest is a high hurdle or 👇
Agree. Once capital, careers, and infrastructure are locked in, vested interests alone make a full reversal extremely unlikely.
Where do you both see the greatest risk: overvalued wrappers or overstretched infra bets?
Overvalued wrappers, to me, but I see more and more of the big AI companies starting to add true value, which is encouraging.
As I see it, the current infrastucture bets are projected for years ahead, currently they need to invest on more infrastructure because basically all their computing power is currently used.
If you predict a possible 5-10X increase in AI usage per year in the next 2 years, they in fact need this new infrastructure.
But the big spending in infrastructure is projected for the next 5-7 years ( OpenAI ), they can still cut that spending in future if it will not be needed anymore.
Great question. The fragility is mostly in overvalued wrappers with weak moats. Infra bets can be overstretched, but they’re long-cycle assets.
I think it was a good approach to define what you mean by an AI bubble crash because I think a lot of people say it meaning different things.
Exactly!
Very thoughtfully laid out arguments! Thank you, Laura and Joel, for sharing.
I think that the current AI situation looks very similar to what was happening in the biotech industry back in 2021/2022. With advanced technologies like gene therapy and gene editing becoming better understood and approved by regulators, there was almost this "gold rush" to set up another gene or cell therapy company.
But that sort of R&D is expensive, never mind the low probability of clinical success, so the funding and valuations for these quickly went down after the initial spike of excitement. But this was actually a "blessing in disguise" – companies that didn't have strong scientific and financial foundations went under, but those with good evidence and fiscal discipline actually weathered the storm.
So with AI, it might play out similarly: companies with weaker AI market positioning will likely disappear, but that gives more opportunity for stronger players to reassess and adjust their current strategies to make sure they come out on top.
Great analogy.
That’s exactly the pattern I expect: not destruction of the category, but selection pressure. Weak foundations get exposed, strong ones compound.
Thanks for taking the time to read the article :)
Alena thank you for sharing! Such an insightful take! That’s what I think will happen also, weaker companies will disappear
The bubble feels real because adoption is being mistaken for competence. Using AI reduces surface effort, but it does not reduce the need for judgment. In fact, it raises the cost of bad judgment because errors now scale faster.
Exactly. Adoption without competence creates the illusion of progress. Thanks for reading :)