Jamie McGeever
ORLANDO, Fla. (Reuters) – Artificial intelligence (AI) may finally be losing its luster.
If so, the recent shift away from big tech companies and towards small caps could rapidly accelerate, leaving the question of whether outperforming laggards are propping up the broader market or whether the AI sell-off will push the benchmark index into the red.
If the second-quarter earnings season shows another upbeat result and outlook, it could reinvigorate the bull market for all things related to artificial intelligence and microchips.
But the “Magnificent Seven” stocks, which account for a third of the S&P 500’s market capitalization and roughly two-thirds of the index’s overall gains this year, are at big risk if they fail to live up to rising expectations.
The US-China trade war could escalate with President Trump in office, offering a glimpse into how this could play out, if at all, to the benefit of mid-market US-based companies.
Last Friday, the Magnificent Seven ETF, which includes semiconductor chip maker Nvidia (NASDAQ:), fell 4.4%, its biggest drop since its inception in April 2023. The small-cap index posted its biggest one-day risk-adjusted gain on record and the third-biggest outperformance against the Nasdaq, according to Bank of America analysis.
US inflation data released on July 10th was surprisingly weak, with the Russell 2000 Index up 10%, but both the “Mag Seven” ETF and the NYSE FANG Index, which includes the Mag Seven stocks, down more than 5%. The S&P 500 Index is also currently down.
Like the emperor’s new clothes, the question is now being asked whether AI really lives up to expectations.
Daron Acemoglu, a professor of economics at the Massachusetts Institute of Technology, wrote an article in May titled “Don’t Believe the AI Hype,” a concise follow-up to a broader research paper he wrote earlier that month titled “Simple Macroeconomics of AI.”
Acemoglu argues that the impact of AI technology, at least in its current form, on “total factor productivity” over the next decade is estimated to be a relatively small 0.53% – a negligible 0.05% per year.
His prediction that AI will boost productivity and GDP growth by about 0.5% and 1%, respectively, over the next decade is significantly lower than a similar forecast by economists at Goldman Sachs of about 9% and 6%.
High costs but low profits
Acemoglu’s thoughts and findings were published in a Goldman Sachs report released June 25th called “Gen AI: Too Much Spending, Too Little Benefit?”, which analyzes the pros and cons of AI.
Jim Covello, the investment bank’s head of global equity research, is much more skeptical than his colleagues on the economics team.
Covello predicts that investment in expanding AI infrastructure, including data centers, utilities and applications, will exceed $1 trillion over the next few years. In his view, the key question is: what trillion-dollar problem will AI solve?
“Replacing low-wage jobs with very expensive technology is fundamentally the opposite of any previous technological transition I’ve witnessed in the 30 years I’ve followed the tech industry closely,” he says.
Comparisons to the early days of the internet are misplaced. Even in its early days, the internet was a low-cost technology solution that allowed e-commerce to literally replace expensive existing structures like brick-and-mortar buildings.
Covello gives the example of integrating GPS into smartphones: The technology for widespread deployment of this didn’t exist in the early 2000s, but the “roadmap” — no pun intended — existed. Roadmaps for what other technologies could eventually achieve also existed from the beginning.
Is that the case with AI today? “Eighteen months after generative AI was introduced to the world, we have yet to see a single truly transformative, let alone cost-effective, application,” he argues.
Not a game changer
Covello is one of the few outspoken voices on Wall Street to criticize the AI boom, echoed this week by Unlimited Funds CEO and former Bridgewater executive Bob Elliott.
Even in the most optimistic scenario, Elliott said the benefits to S&P 500 companies from increased AI spending and economy-wide productivity gains would be “modest.”
This scenario assumes that AI spending by S&P 500 companies will increase by $1.3 trillion by 2032, and revenue growth will rise from 4% to about 6.5%. In total, he estimates that S&P 500 revenues would be about $650 billion higher by 2032 than today, or about 25% higher in nominal terms.
Even if we ignore the difficulty of predicting earnings eight years into the future, the current market capitalization of the S&P 500 would increase by about $10 trillion, or 25%.
“This is a fairly small impact, not game-changing… (and) likely already priced in… probably fully priced in during the AI ’boom’ last summer,” Elliott wrote in a post on X this week.
Investors may be gradually coming around to this view: A July Bank of America survey of fund managers found that 43% of respondents believe AI is a bubble, up 5 percentage points from May, while 45% do not, down from more than 50% in May.
But nothing changes sentiment quite like price, and it will probably take a bigger reversal in these crowded trades to convince investors that AI’s well has run dry.
(The opinions expressed here are those of the author, a Reuters columnist.)
(Editing by Jamie McGeever and Paul Simao)
