This post is opinion only. See full disclaimer below.
The case of the spectacular rise in AI stocks which have now been the market leaders and reigning stock market narrative for some time certainly requires a great deal of macro-political analysis by any investor regardless of their particular style of investing. Also educationally this run up is extraordinarily interesting and important as a study of many key fundamental areas for markets.
Does the AI narrative have all the hallmarks of a classic bubble? Absolutely. As I have written of elsewhere regarding bubbles the market narrative has in this case also been used to create tremendous excitement by the retail investors and to generate a river of money flows not just from the retail investors but also the private sector and governments. As they say about Wall Street, and if you know actual NYC you’d know this is the case with the actual street–at one end is a river and the other end a graveyard–and in this regard, AI has indeed generated a huge flood of money into the markets. (Those of you who want a more detailed advanced account of the key factors in bubbles see my essay on this subject and on the Efficient Market Hypothesis).
As happened with the tech bubble it’s not that none of the AI companies may continue to grow and stay powerhouse companies, but that the sector as a whole may not be able to support the valuations currently predicted. For every Oracle and Amazon in the earlier tech bubble there were dozens of companies that ultimately failed. And this is nothing new–study the usual course for any new revolutionary technology and you will in many cases see how this all often plays out. The car was an incredibly revolutionary invention indeed but only a few car companies ultimately lasted. Also before Amazon and Oracle got their present valuations they had a few issues when the tech bubble crashed in the early 2000s. Go back and look for yourself at the charts from that period for those stocks and how far they sank when the bubble crashed during that period—and longer term these were the winners. The fact that some of the current charts look like rocket ships heading to the moon at least requires pondering.
Along with this is the fact that as happens with every bubble the idea that the source of the bubble totally changes the investment landscape and that this time is different, as they always say, is dominant. What if this time isn’t different after all? I will have more to say about this in later posts, but I just want to lay out a few key issues here.
There are also two more bubble characteristics in crowd psychology at present that are not unimportant: First before bubbles fail there are always a few prominent investors who begin sounding the alarm: Michael Burry the famous “big short trader” who correctly predicted the 2008 housing crash recently did just that arguing that AI is similar to the earlier tech and housing bubbles and includes the structural risks of accounting tricks, capital spending excesses, and investor complacency among other problems. And you are starting to hear more and more murmurings lik Burry’s from Wall street that it might be a bubble. But as always happens others quickly step forward to refute their claims and reassure the crowd at least for a while.
Second at the top of bubbles everyone seems to be in the markets and knows the narrative. This is certainly the case at this point. AI has become the favorite conversation starter at cocktail parties and AI stocks all the rage across many social strata. As more and more people get into the bubble, future flows slow down and eventually the river of capital runs dry.
However despite all this I want to also suggest the counterfactual to the idea that AI is a bubble and what might actually be different this time, believe it or not and some key issues critics like Burry might be missing. It is unfortunately conceptually not always the case that what might seem like a bubble actually is one. If you go back to either the tech bubble or the housing bubble retrospectively and even to some of us at the time, the counterfactuals to the bubble thesis were not very strong. In the case of AI the counterfactuals are much more powerful. Could it be a bubble? Sure. But the arguments the other way are also quite powerful. If only life, markets, and reality were simpler than they are in fact.
The most obvious counterfactual is AI’s potential role as a productivity multiplier. I don’t mean in a simple sense its ability to totally replace human labor itself. I know that is often the narrative, but I believe in that regard it actually has many weaknesses. If your goal is simply to make it that when people call your national company or any bureaucracy about a problem they can’t readily get a real human being and you can save money by cutting staff I think thats actually longer term a negative. After all and I’m painting with large brush strokes so please don’t come back and tell me its more complicated, which I know, but if you fire all the workers whose going to buy your products and its a race to the bottom. This was something Henry Ford understood so many years ago when he paid his workers a living wage—people got to be able to afford the cars.
No, what I mean is that as a tool for real people properly used it allows not just more work but higher level work to get done. If I’m considering a difficult problem it can allow me to do days of work in mere hours—not give me a simplistic answer I don’t need which it can also do but just speed up the tasks of gathering the material itself. Right now it does so in an incredibly hit or miss fashion. Often it gives opposite answers about not unimportant things in succession. But if you know enough you end up correcting AI a lot, but then it builds from that corrected starting point and ultimately lets you go further. Right now that more complex discussion doesn’t add to the knowledge base of AI beyond the immediate discussion–so it gets temporarily smarter from the interchange and then reverts back with the next interlocutor to its simplistic and even sometimes wrong initial answer. But once it’s able to keep that development it’s potential could be quite amazing. However, that means that whoever is using it still has to know a lot and be able to think critically. It requires longer term for actual use a more not less educated populous—otherwise it’s just de facto an illusory bubble.
The second in some ways more important counterfactual is that when technologies or industries become crucial to national security and politics and power gets fully involved the potential bubble trajectory becomes much more complex. No doubt the superpowers security apparatuses have had something like AI for a while now but in a more circumscribed and specific form, if its true that the government has technology decades ahead of the public. But the limitation on computing power has always been scale. In order for AI to succeed and for a superpower country to use it in full it must amp up its scale incredible amounts. Those huge data centers need to be built. For those of you who missed the memo the US and China and indirectly many countries are in a spectacular race to see who can harness the full power of AI first. A lot is at stake. That is a unique situation akin to the period after the Cold War when the government heavily invested in what might be termed the military industrial complex. That investment led to a huge step forward economically for the US in that earlier period. Sure part of that economic advance was the undisputed hegemonic power of the US as victor after WW2, but also a lot of it was that government-industry commitment to that Cold War battle. Maybe I’m wrong and correct me if I am–but I don’t recall ever hearing of a military industrial bubble collapsing during that long period. So its not about just valuation but politics and power.
The third crucial counterfactual to the bubble hypothesis is connected to this one. AI will take huge amounts of energy. One of the criticisms of the clean energy carbon control initiatives of various sorts was that they would give too much power to the government over the whole economy and specifically the energy sector as energy was for all extents and purposes rationed. Energy after all is the key to every societies wealth and success.(For more on this see my earlier posts on energy.) AI creates this huge control expansion not by conserving energy but in the opposite direction entirely by needing its exponential increase to power all those data centers. But here’s the deeper point: AI is not just about AI, it will require tremendous change on many real levels of the economy. As such its hard to know analytically the limits of the narrative—where reality ends and illusory bubble begins. Most of the time with narrative hype you know right away or at least fairly soon it’s illusory. Perhaps that’s the case with AI also but its more difficult to discern than in less complex cases. There could be substance to the narrative below its obvious in some cases bubble-like millenarian promises.
I am not discussing more purely technical issues here like despite the fact that you hear this occasionally even incorrectly from AI technologists themselves that energy does not like computing power follow Moore’s law which is a potential problem. Or the fact that the human material to train AI has limitations in extent and as suggested in a recent important higher level analytical paper training AI on AI if you will does not seem to work. And the problem of queueing for energy sources already occurring in data centers in the UK or countless other not unimportant issues for AI and markets so theres a lot of dimensions to this topic indeed.
Nor is it simply time will tell. That may be the case, but any serious market participant needs to develop a workable framework that recognizes both the bubble and counter-bubble alternatives and follow all this closely indeed. It could be 1998 or 2006. Or not.
To ultimately analyze the bubble power situation of AI we will need to add to the above points the all important link between narrative power, group psychology, and trader action of the sort found in my new GPTA model discussed elsewhere, but the above discussion at least raises a few key issues surrounding AI and the markets.
Disclaimer:
The content on this blog is provided for informational, educational, and entertainment purposes only. It represents the author’s personal opinions and should not be construed as financial, legal, or tax advice.
By using this blog you agree that:
–You assume full responsibility for any outcomes resulting from the use of this content.
–You are solely responsible for all investment decisions and any gains or losses you incur.
–Past performance is not indicative of future results.
Any discussion of specific securities, sectors, or strategies is offered strictly for educational purposes and does not constitute a recommendation to buy, sell, or hold any investment. You should always:
Consult a qualified professional before making any financial decisions.
Consider multiple viewpoints and counterarguments.
Conduct your own research