Beware of the AI hangover

Artificial Intelligence (AI) related stocks have been on an absolute tear. The Magnificent Seven, which is one proxy for AI plays, has beaten the S&P 500 for six consecutive quarters. But the degree of outperformance is decelerating. This looks like the early stages of an AI hangover.



I believe AI stocks are overhyped short term but underhyped long term. Here’s why.


AI doubts emerge

The exuberance over AI can be exemplified by the market reaction to the earnings of two AI ecosystem stocks, Palantir (PLTR) and Cloudflare (NET), both of which reported double beats on earnings and revenue. Reports of segmented results were strong. In Palantir’s case, the company even raised guidance, but the stock fell because guidance failed to exceed the whisper numbers. In Cloudflare’s case, the market sold off because the company failed to raise guidance.


From a fundamental perspective, signs are appearing that the narrative is turning from the upside to the downside of AI adoption. The CNBC article, “AI engineers report burnout and rushed rollouts as ‘rat race’ to stay competitive hits tech industry”, tells the story of AI development for the sake of appeasing investors instead of actual productive development. The key points are:

  • Artificial intelligence engineers at top tech companies told CNBC that the pressure to roll out AI tools at breakneck speed has come to define their jobs.
  • They say that much of their work is assigned to appease investors rather than to solve problems for end users, and that they are often chasing OpenAI.
  • Burnout is an increasingly common theme as AI workers say their employers are pursuing projects without regard for the technology’s effect on climate change, surveillance and other potential real-world harms.

At the Berkshire Hathaway annual general meeting, Warren Buffett was asked about AI, his response focused mainly on the downside of AI:

Fairly recently, I saw an image in front of my screen. It was me, and it was my voice and wearing the kind of clothes I wear. My wife or my daughter wouldn’t have been able to detect any difference. And it was delivering a message that in no way it came from me.
When you think about the potential for scamming people… Scamming has always been part of the American scene. If I was interested in investing in scamming— it’s going to be the growth industry of all time.
Journalist Izabella Kaminska recently raised an important question of the usefulness of future Large Language Models. Current LLMs are trained on “mainly established authors who have a substantial body of pre-published work that can provide enough learning material to ensure it emulates their intelligence faithfully”, but what happens when LLMs begin to proliferate and try to train themselves on LLM output?


Technical warnings

From a technical perspective, momentum in AI stocks is flagging. I identified 29 stocks exhibiting strong price momentum in January (see At Least 29 Reasons to be Bullish), most of which were technology stocks. Most, like this chart of NVIDIA, are either consolidating sideways or pulled back like Palantir and Cloudflare.


By contrast, market leadership has shifted to cyclical names, like Costco, which broke out of a multi-year base and rallied to fresh highs.


The homebuilding stocks are exhibiting a similar pattern of market strength.


GE, which is an important cyclically sensitive industrial stock, has been on an absolute tear.


In other words, leadership has shifted from AI-related technology to value and cyclical stocks. But from a long-term perspective, the AI bull isn’t over. Here is the percentage of NASDAQ 100 stocks bullish on point and figure, which recently bottomed out at below 25% and rebounded. In the last 10 years, such episodes have always seen the NASDAQ 100 stage a relief rally. However, the initial rally was often fleeting and followed by a lower low.


Another source of warning of near-term NASDAQ weakness can be found in the relative volume of NASDAQ/NYSE volume, which is weakening.



Be patient

Even though AI stocks appear to be overhyped in the short run and prone to a hangover, investors are advised to be patient on the long-term potential of AI. A St. Louis Fed study compared the pace of AI adoption to other technologies. It concluded that “early evidence on the diffusion of AI seems to suggest a pattern similar to those of personal computers and cloud computing”. In other words, it will be a long cycle. Investors need to be patient about the pace of AI benefits.


In conclusion, I believe AI is overhyped in the short run and underhyped in the long run. Recent instances of earnings report reactions of stocks in the AI ecosystem indicate excessive frothiness and even AI bellwethers like NVDIA have begun to consolidate sideways. A St. Louis Fed study of the pace of technology adoption suggests that AI adoption will undergo a long cycle like the PC, so investors need to be patient.


Investors who want to monitor the turn may want to keep an eye on the semiconductor stocks as AI proxies. These stocks are correcting and consolidating sideways after breaching a rising trend line. Upside absolute and relative breakouts of resistance will define the re-emergence of the AI bull.



3 thoughts on “Beware of the AI hangover

  1. This reminds me of the old saying about market reaction to news. When a company reports great earnings and revenues but goes down is a warning sign. Another saying is “it doesn’t go straight up”.
    Considering how the market has gone from 666 in 2009 to present, what kind of bear is lurking in the future?

  2. One step further, monitor TSMC about AI adoption. By extension AMAT, LRCX, KLAC, ASML. TSMC updated recent booking as very robust last week. And another sign is the Arista Network report Thursday. Both stocks are at ATH Friday. People either buy Nvidia-owned Mellanox or they would opt for Arista for software-based networking. Some derivative plays are also at ATH: VRT, APH, etc. Overall hardware is still going strong. Among hyperscalers GOOGL and AMZN made ATH last week while MSFT is going up. Watch AAPL as it is a company who always signals the onset of mass adoption, with or without merit. Among third legs of the stool, enterprise software and consulting, it is too early. None is doing well: ACN, GLOB, EPAM, IBM, etc. So give it some time for corporations to figure out how the integrate. But it will come.

    Average users and small firms will immediately benefit from handling routines and chores, but that’s about it. Firms in various fields with specialized deep domain knowledge will benefit big time from the integration of AI algos. This is especially true in material science, biology, and chemistry. Check Google’s newest AlphaFold3, for example. You can try it for free. Be ready for amazing progress.

    For general purpose chatbot, there is not much to talk about. It is getting to be chaotic and dubious. Most LLMs have run out of original public data to train. What’s next? These models started to train on their own produced data. Imagine a simulated world in which no one knows where the truth is. The worst outcome is like Led Zeppelin drummer John Bonham dead from chocking on his own vomit. This is so very relevant in the creative industries, like music, motion pictures, illustration, etc.

    Whatever happens in the future is going to be eye popping and the world is going to be very chaotic.

  3. So they invest huge amounts of money in AI. But this needs to pay for itself eventually. Where will all the money come from? If it steals from someone else, that someone will suffer and this has an impact on spending.
    At a certain point it is like building really strong arms so that you can yank your shoelaces hard enough to fly. I don’t buy it.
    Unless someone is tossing more and more fiat money into the system, at some point the inherent inefficiencies of life take over.
    Covid was an excuse, what will they come up with next.
    Next crisis, watch the magician’s hands.

Comments are closed.