Further discussion on whether there's an AI bubble; First day in Kyiv, Ukraine
In yesterday's e-mail, I shared a compelling essay about whether the AI-infrastructure boom is a bubble.
I also highlighted a debate on the topic I had with a friend, who's even more of an early tech adopter than I am and deeply incorporates AI into his life.
This led many readers to send me their own insightful comments, some of which I'd like to share...
Walt W., who worked in the tech sector from the 1960s through the 1990s, wrote:
Today, I have a cellphone that contains more computing power than I ever imagined, and I probably use 10% of its functionality. I use Excel to analyze my investment holdings and Stansberry recommendations in terms of performance. Among my family and peers, I am considered state-of-the-art. I know better – I am probably a caveman in terms of today's technology and AI. I do use Gemini for my searches and am impressed by the quality of the responses.
So, to your question – given my circle, no one I know is going to be like your friend, including me. A few others and I will be somewhat behind you, and the rest absolutely won't care as long as they can send emails or texts and access their music and pictures and post on social media. Will that require AI? In my opinion, not for them, other than searches.
Now, if the education community educates the young crowd to be like your friend, maybe it will be different. But I have no faith in our education system in that regard, so I think it will not be in my lifetime that AI will integrate into our daily lives like the Internet took 8 to 10 years or so to do, as you described.
Sounds like we are going to go down the overbuilt road long before the wholescale adoption road, so preparing for the bursting bubble sounds like the right strategy to me.
Chuck P. agrees:
The Internet is simply fundamental to how we do things. Could we function without it? Certainly, but not as comfortably as we do currently. On the other hand, can we function comfortably without AI? Also certainly. I personally have no use for it. But, as you correctly (I think) point out from your own experience, if AI went away, you would simply go back to doing what you did before without much hassle... I think more people will be like you.
We truly have no idea at present how (or how much) AI will fundamentally change society. I think it's very likely that, in time, AI may very well become foundational to much of what we do (for good or ill). I just don't happen to believe it will be the raging success that so many people seem to think it will.
Mark R. is more bullish on AI:
The complete contrast to your entire position is best expressed by Jordi Visser. I have been following his podcasts for two years or so now. Here's a link to his latest one – I recommend that you give it a listen or ask AI to summarize it. I agree with him that we are at an inflection point, which is best illustrated in his video at 16:38. The inflection point is taken from the essay by Leopold Aschenbrenner, "Situational Awareness: The Decade Ahead."
Automated AI agents are now escalating at a rapid rate and in a short period of time (6 to 12 months) will represent more usage of the entire Internet than humans. Coupled with a digital economy and tokenization, with transactions taking place on blockchains using crypto rails, is where all of this is heading. What are the constraints? Compute, memory, and power. This is why the hyperscalers, who are cash-flow rich, are investing in these areas. They are doing so because of the demand.
Your point of view has merit and the analogy makes sense, but it is all predicated on throttled demand for AI and open sourced models such as OpenAI and Anthropic not being able to continue their pace of revenue growth. I agree with Jordi that demand is outstripping supply and will continue to do so over the next 12 to 24 months.
Mark then uploaded my e-mail and Visser's podcast into Grok and asked it: "Who is right, Whitney or Jordi? Who is more correct on the bubble? What are the risks of both views? What is the most likely outcome?" Here's a link to Grok's reply, which concluded:
Visser is more accurate that the broad "AI is in a bubble about to burst" framing does not match observed scarcity, revenue growth, or valuation metrics, and that agent-driven demand changes the historical pattern. Tilson correctly flags that the financing structure can still produce solvency or refinancing stress for some participants before demand fully absorbs the capacity, and that the absolute scale is unprecedented. The more probable near-term path is continued high investment with periodic volatility, selective equity or credit pain, and ongoing scarcity in key bottlenecks (power, specialized chips), rather than a synchronized housing-style collapse. Both views will be tested most clearly by [remaining performance obligation]-to-revenue conversion, utilization, and actual agent/enterprise scaling over the next 18-36 months.
Billy D. is also quite bullish:
Per your latest email, this quote from the essay stuck out: "For committed compute merely to equal revenue in 2027 – not to be comfortably covered, simply to reach parity, before a single dollar is spent on wages, research, sales, or tax – revenue would have to compound at 217% annually off the 2025 base... [No] company at this scale of revenue has ever done [this] for a multi-year stretch. The obligation is accelerating at more than double the rate of the best case for the cash flow meant to cover it..."
If he's using 2025 as a revenue base, this actually makes the case that AI will scale accordingly. Anthropic's revenue in 2025 was just $4.6 billion. By July 2026, it was already at a $65 billion run rate (15x). Anthropic was also profitable in the last two quarters. So, at least Anthropic appears to be way ahead of the required compounding rate to justify the infrastructure spend.
OpenAI is not doing well, but still seems on pace to 5x revenue this year ($13 billion to $70 billion-plus), also ahead of the required pace.
Billy continues:
It remains an open question how much efficiency can be wrung out or go to open source models, but at least the two frontier AI labs seem well ahead of the required pace. Not sure what Gemini or xAI are at this year, so we will have to see.
Valuations also aren't as high in the 2000 dot-com bust. The Magnificent Seven actually seems quite reasonably priced. In terms of housing parallels, while private credit may have some trouble, it's not like banks are levered up 30-plus to 1 backing these data centers. Nvidia does have a big financial backstop though.
I do wonder whether the stocks levered to training the frontier may be a bubble, as the multitude of frontier models consolidates from like a dozen to 2 or 3. However, inference seems like it's only going to accelerate going forward.
Bull case to be found with the SemiAnalysis guys here.
Thank you, dear readers, for sharing your thoughts and insights!
Regarding Billy's statement that OpenAI is on pace to make 5 times revenue this year ($70 billion-plus), the Financial Times just reported that its annualized revenues are actually far less than expected.
My friend Chris Irons, in his Quoth the Raven Substack, thinks this is one indicator that "The AI Rally Trapdoor Just Swung Open Violently" (subscription required). He notes that:
$50 billion in annualized revenue is still an enormous number, and the difference between an inferred revenue estimate and an actual reported run rate isn't necessarily the same as missing formal guidance. But that distinction doesn't eliminate the broader problem: expectations surrounding this industry have become so enormous that even extraordinary growth may not justify the amount of capital already committed to it.
He shares my view that:
... nowhere is that problem more dangerous than at OpenAI, which sits at the center of what may be the most extraordinary financial circlejerk in modern market history, neatly displayed in this chart from Zero Hedge, who noted total that OpenAI has $1.5 trillion in commitments ahead of them:
Chris concludes:
That is the trapdoor I'm worried about. The entire AI ecosystem has spent years celebrating one another's investments, contracts, partnerships and projected revenues as though money circulating between the same handful of companies represents an inexhaustible source of genuine economic demand. Eventually, somebody outside that circle has to generate enough economic value to pay for the whole thing, and if that value doesn't arrive quickly enough, the financial structure can unravel long before the technology reaches its full potential.
Less than two weeks ago, I wrote that the AI emperor had no clothes and wondered who would be the first to stand up and say it. Now the financial statements are coming out, revenue expectations are being revised, the IPO market is pushing back and the credit markets have already been sounding the alarm.
The technology may very well be revolutionary, but that doesn't mean the price Wall Street has assigned to the revolution makes any sense.
And if I'm right about where we're heading over the next several months, the most important question won't be how much money everybody thought they were going to make from AI, but how much money everybody owes everybody else when the music stops.
Boy, does this look like a bubble:
Best regards,
Whitney
P.S. I welcome your feedback – send me an e-mail by clicking here.
P.P.S. Due to the increasingly frequent Russian missile attacks on Ukraine's capital, Kyiv, my volunteer team and I dropped off all but five of the 57 ambulances in Lviv, which is much further west – only a two-hour drive from the border with Poland. Here's a picture from the ceremony:
Most of the 64 volunteer drivers went home from Lviv, but two dozen of us continued on to Rivne and then Kyiv.
The next day, we drove 45 minutes to the suburb of Bucha, which is well known as the site of some of the Russians' worst atrocities. In the early days of the full-scale invasion in February 2022, Russians quickly advanced to the outskirts of Kyiv in an attempt to end the war quickly. As part of this, they occupied Bucha for 33 days, during which they tortured and killed more than 500 civilians (Wikipedia article here and video here).
The bodies of 119 of them were dumped in a mass grave behind the Church of St. Andrew the First-Called, which is now the site of a memorial. We then visited the local military cemetery, which, sadly, is now present in every city in Ukraine. Lastly, we stopped by the tank graveyard, where the Ukrainians dragged the rusting hulks of the Russian tanks they destroyed. Here are pictures:
