News & Insights
Stop us if you’ve heard this one before. The whole investor world is piling into stocks with a hot AI story to tell, driving prices into the stratosphere. Then, out of nowhere, something comes along – a research piece, an offhand comment by someone at the center of the AI universe – that pours cold water over the bullish vibes. Suddenly, everyone is talking up “rotation” and “equal-weighted index” and running as fast as their little legs will take them away from all things AI.
The Rotations That Weren’t
Well, we have arrived at the midsummer-2025 version of this dance, thanks to the combined jab-cross effect of a recent study conducted by a team of researchers at the Massachusetts Institute of Technology and some out-loud musings by Sam Altman, the head of OpenAI. We’ll get to the gist of what these deliveries had to say in a moment; first, though, let’s do a quick review of bygone Cassandras signaling the apparent end of the AI narrative.
Just about one year ago it was a research piece by Goldman Sachs that dared to ask the question: what is all this $1 trillion in AI capital expenditure actually going to do? The point being made was that all that money going into sprawling data centers and voracious large language modules had yet to prove the existence of profitable use cases to justify the outlays. AI stocks pulled back sharply, CNBC pundits announced the dawn of the rotation into value stocks, and then…the rotation fizzled, nobody came up with a more compelling narrative for the market, and everyone went back to buying AI names.
The next head-fake came at the beginning of this year in the form of DeepSeek, the Chinese AI platform that appeared to have nearly the same whiz-bang capacity as their American competitors, but at a fraction of the cost. Once again, investors ran for the hills, pundits wrote the obituaries, and then the rotation faded away. Eventually, the market talked itself into the idea that, actually, DeepSeek was more of a validator for the AI story than a detractor.
MIT and Altman Deliver a One-Two
Now we come to the present moment, on the heels of yet another long run of outperformance by the so-called Magnificent Seven and their assorted hangers-on in the AI space. This week a group of researchers affiliated with the Massachusetts Institute of Technology published a study with the conclusion that 95 percent of organizations are getting “zero return” from their generative AI investments. In other words, the MIT study seemed to validate what that Goldman Sachs research article was saying a year ago – that all the jaw-dropping sums being poured into AI were failing to demonstrate how the technology was going to make money for its users.
With the MIT study fresh off the press, investors then went back and looked at some comments made late last week by Sam Altman of OpenAI (the enterprise that gave ChatGPT to the world). Altman remarked that a bubble in AI may well be at hand and that investors had the potential to lose a lot of money before the technology lived up to its full potential (which Altman, naturally, believes is eventually going to be world-changing). Rotate into value, call the pundits, rinse and repeat.
The Innovation – Adoption Gap
One fact which has almost always been true about new technologies is that there is a window of time – sometimes a very long window – between the original innovation and its widespread adoption. The light bulb, to cite one example, was invented in 1879. Forty years later, only half of all American households were connected to the electric grid. Another example: computers didn’t start landing on the desks of office workers in any meaningful number until the early-mid 1980s. The productivity gains from automating business processes didn’t show up in macroeconomic data until the early 2000s.
It is likely that a similar innovation – adoption gap will exist between the introduction of generative AI, which happened only two and a half years ago, and its ability to demonstrate a profitable impact on enterprise functions. So-called AI agentics – the ability to leverage personnel and related expenses with scalable AI operations – are already out there, but have yet to approach any kind of critical mass.
In the meantime, though, we see no signs of any kind of let-up in the flow of investment dollars into AI infrastructure. The companies that are making these investments and own the intellectual property behind them are going to continue to have a compelling story to tell. For that reason, we imagine that the current AI freak-out is going to end the way the previous ones have. Rinse and repeat.