Since 18 months ago, a lot of hype has been made about artificial intelligence and how it could change business profits and worker productivity. Now, Wall Street seems to be losing faith.
Analysts are now taking a close look at the numbers that support the technology. They learned that the huge investments in capital expenditures being made by “hyperscalers”—big companies with big cloud-services businesses like Microsoft Corp. MSFT, +1.46%, Amazon.com AMZN, +0.23%, and Alphabet Inc. GOOGL, +1.16%—might not pay off for a long time, if ever.
Some analysts even questioned the main ideas behind the “AI revolution.” Jim Covello, global head of equity research at Goldman Sachs, said he didn’t think the technology would be as profitable or important as its supporters say it will be.
Many people think that AI will be the most important technological invention of their lifetime, but I don’t agree because the internet, cellphones, and laptops have changed our daily lives so much that we can do things that were not possible before, like make calls, do math, and shop from anywhere. Covello said this in a question-and-answer section of a Goldman Sachs report that has been talked about a lot on social media sites like X.
Based on what Covello said, companies will have spent about $1 trillion on AI-related capital over the next few years. He said that AI has to show that it can solve complicated problems, which is something it wasn’t designed to do, in order to justify this huge cost.
He also disagreed with the idea that the very high costs of running generative AI products will go down over time. For ten years, other chip designers have been trying to take market share from Nvidia Corp.’s (NVDA, +2.69%) GPU business, but they haven’t been able to. This goes against the idea that more competition should make the chips needed to train AI models and run applications much cheaper.
In addition, making AI work has become very expensive. These costs would have to drop a lot for most companies to be able to use AI in the ways that supporters of AI want them to.
Big investments in AI have also caused people to question whether they will pay off. There was a lot of investment in data centers by cloud-computing service providers. An analysis team at Barclays thought this might have more to do with “FOMO” (fear of missing out) than “Field of Dreams,” the movie famous for the line “if you build it, they will come.”
They pointed out a difference between what Wall Street thinks AI-related capital expenditures will cost and how much extra money these investments are supposed to bring in.
Analysts on Wall Street think that cloud service providers will spend an extra $60 billion a year on chips and data centers, but they will only make an extra $20 billion by 2026.
The Barclays team came to the conclusion that the biggest tech companies are spending too much on infrastructure, just like the companies that laid fiber-optic cable during the dot-com era.
Barclays’ numbers show that the extra capacity being built by projects that are already underway would be enough to power the current internet plus 12,000 new apps with the same number of users and input needs as ChatGPT. This suggests that hyperscalers’ plans to build new data centers are bigger than what is needed.
Finally, a group of strategists at Citigroup brought it back to the markets by warning that investors are feeling too good about the top AI stocks. They told their clients to buy high-flying semiconductor companies like Nvidia and Advanced Micro Devices AMD, +3.87% and make money.
What’s more, the Citi team said that shares of many of the biggest names in AI are now too expensive, even when compared to Wall Street’s already high estimates for free-cash flow growth over the next five years.
Companies like Tesla Inc., TSLA, +0.35% Nvidia, and Microsoft are in the bank’s “high AI exposure” basket and are pricing in growth that is higher than what Wall Street expects.
At this point, though, the Citi team also said that telling clients to completely avoid the AI theme wouldn’t work for most of them.
The Citi team wrote, “In practice, not owning AI or outright positioning against it would be hard for many on the buy-side.”
Of course, Wall Street hasn’t been the only place that doesn’t believe AI will have a big impact. Some academics, like Daron Acemoglu of MIT, aren’t sure that AI’s productivity gains will be as revolutionary for the world economy as many of its supporters say they will be.
Acemoglu found that AI will only increase productivity in the U.S. by about 0.5% over the next ten years, but it will also add 0.9% to GDP.
A study from the McKinsey Global Institute, on the other hand, found that AI and automation could add up to 3.4 percentage points to GDP growth in developed economies over the next ten years.
Investors still seem to believe in the theme, though. A lot of stocks with an AI theme have already recovered from a short drop earlier this year.
AI stocks like Nvidia Corp. and Broadcom Inc. AVGO, +0.66% went down for a short time at the end of last month. On June 24, Nvidia stock went into a correction after ending the day more than 10% below its recent high point. Both stocks have recently made up most of the ground they lost since the beginning of July.
The Invesco AI and Next Generation Software ETF IGPT has been doing very well. On Wednesday, it hit its highest level since late 2021, which shows that shares of AI-related software companies have mostly recovered from their early-year losses.