Having struggled on Wall Street for over 30 years, Jim Covello, the head of stock research at Goldman Sachs, knows how painful it is to short a bubble that is still expanding as there always seems to be a way for the market to continue rising month after month.

This has happened before, during the dotcom bubble of the late 90s and in recent crypto craze. Covello, head of stock research at Goldman Sachs, says it's likely to happen with AI too, but shorting companies like Nvidia (NVDA) now could be dangerous, even foolish.

However, Covello has no doubt that the reckoning is coming. He says it may not be this year or even next year, but it will happen someday. In his view, the billions of dollars that companies are pouring into AI will not spark the next economic revolution, nor will they come close to matching the benefits brought about by smartphones and the internet. When this becomes clear, all stocks that soared because of this narrative will fall. Covello says,

"Most of the tech transitions throughout history, especially the ones that are transformative, have been ones where we've replaced very expensive solutions with far cheaper ones. But now we're replacing work with extremely expensive technology, which is essentially the opposite of history."

Covello is becoming a leading voice among a growing number of skeptics who are questioning a key argument that has supported the continued rally of the S&P 500 index (SPX) since the end of 2022, that the astonishing power of large language models will usher in the world's next industrial revolution, in which more and more work will be handed over to intelligent machines, boosting corporate profits, and efficiency, and hastening growth.

Many Wall Street figures believe it's possible, and JP Morgan CEO Jamie Dimon has said he's convinced AI will bring about extraordinary transformation that could be as significant as those brought about by printing, steam, and electricity. But Michael Arone, Chief Investment Strategist at State Street Global Advisors, says AI brings about "persistent and unprecedented productivity miracles." Even within his own firm, Covello points out, Goldman Sachs senior global economist Joseph Briggs estimates AI will eventually automate a quarter of all tasks and drive economic growth.

This speculation has fueled a real boom, with the world's largest tech companies pouring huge sums into the area in hopes of extending their dominance into the new landscape. For hardware providers such as Nvidia, Broadcom, and AMD, which supply AI models, this is welcome news. Even utilities are seeing a boost in sales as demand for data centers soars, requiring huge amounts of electricity.

However, skeptics argue that the problem is that commercial expectations for the technology could be greatly overestimated, and a rethinking of investment by tech giants would pose a risk of a market pullback.

Don't be the "last fool."

David Bahnsen, founder and chief investment officer of the Bahnsen Group, has long been preparing for just such a situation, avoiding Nvidia and other tech giants because he anticipates a potential "catastrophe." He says,

"The way we make money is by not owning those kinds of stocks when the last fool bought Cisco stock in March 2000. After the dotcom bubble burst, Cisco stock plummeted. If you didn't sell these stocks early, you lost a lot of money."

Although there is little sign of such a situation yet. Although technology stocks fell on Wednesday on concerns that chipmakers would be further embroiled in a trade war, they still remain close to historical highs.

Since hitting bottom in October 2022, almost half of the rise in the S&P 500 index has been concentrated in six stocks: Apple (AAPL), Microsoft (MSFT), Nvidia, Alphabet (GOOGL), Amazon (AMZN), and Meta (META).

Nvidia's market cap has already increased by nearly $2 trillion this year and is still one of the most popular stocks on Wall Street. Of the 64 analysts who track the chipmaker, 64 still recommend buying with only one rating it as a sell, even though it is up nearly 140% this year.

Despite heavy investment from these companies in AI, however, returns have been relatively small so far.

Microsoft, Alphabet, Amazon and Meta have invested over $150 billion in capital expenditures over the past four quarters, with most of it going towards computing power to train their own large language models and serve customers.

Microsoft has been integrating its product line with OpenAI's technology, and in April, the company announced that Azure and other cloud service sales grew by 31% in the third quarter, with AI contributing 7 percentage points, but did not provide an amount.

Amazon is expected to generate over $600 billion in revenue this year, with the company only stating that its AI business's "revenue scale has reached billions of dollars". CFO Ruth Porat acknowledged on Alphabet's first-quarter earnings call that "AI is playing an increasingly important role in driving Google Cloud revenue."

For investors like Adam Gold, CIO of Katam Hill, he still holds a heavy position in Nvidia and believes it is too early to focus on certain figures. He pointed out that companies like Meta, while not directly charging users, have achieved sales growth by improving advertising placement and user engagement through the use of AI.

However, for some of the cloud computing giants' clients, the benefits are not obvious. According to a survey by Lucidworks based in San Francisco, less than half of companies investing in AI have yet to see significant returns.

Covello, who has been following the technology industry since he joined Goldman Sachs in 2000, and has won the top analyst awards in the industry year after year, was promoted to Goldman Sachs US Stock Research Director in 2015.

He expects that investment in AI infrastructure will reach about $1 trillion in the coming years, but for AI to generate enough returns, it must be able to help companies tackle increasingly complex tasks.

In his view, AI has shown the potential to improve efficiency in certain jobs, such as programming, but is far from proving that its cost is reasonable.

He said that if there are no significant use cases in the next year and a half, the stock market will reverse. However, he thinks that the current AI narrative has not yet reached this stage and may continue to drive investors to buy stocks like Nvidia. He said:

"One of the most important lessons I've learned in the past 30 years is that bubbles may take a long time to burst."

Edited by Jeffrey