华尔街的AI狂热症

When Wall Street catches a whiff of a new innovation that could drive earnings growth, it tends to get a little overexcited. Sell-side analysts at investment banks rush to put out bullish research reports on related companies; economists and strategists start discussing the potential impacts of the new tech for growth, inflation, and productivity; and money managers begin to ponder the implications of it all for their portfolios. It’s a flurry of activity and interest that can quickly turn into a self-reinforcing hype cycle featuring highly optimistic forecasts.

The rise of artificial intelligence is the latest example of how the Street can become attached to a new technological advancement. The widely successful release of OpenAI’s chatbot ChatGPT in November sparked a wave of enthusiasm about generative AI’s potential to reshape the world, usher in an age of abundance, or even cure the incurable—all while boosting worker productivity and reducing corporate costs. Ever since, analysts have been watching for signs of AI’s impact on corporate earnings, particularly when it comes to big tech, and firms that show promise in drawing in new AI-related revenues (like Microsoft or Nvidia) have seen their stocks soar. Meanwhile, there’s been a steady stream of research reports and studies pouring out of Wall Street that push (and only occasionally downplay) the AI hype.

Fortune analyzed hundreds of pages of AI reports from investment banks, hedge funds, and investment research firms and found some striking similarities. In short, the Street is incredibly bullish on the long-term prospects of AI, with analysts arguing it will boost worker productivity and GDP on a scale similar to the birth of the internet — eventually. But that’s the key word: eventually. Wall Street’s research reports and studies reveal a split between experts who believe that the near-term AI hype is overdone, and those who argue it’s justified given the rapid adoption of the technology and potential for long-term succes.

The titles of the Street’s reports are typically quite revealing. The private equity giant Carlyle went with “Brave New World: AI and Its Downstream Implications” for its October report that called the birth of generative AI a “watershed in human history,” comparing it to electrification. Citi decided on “Unleashing AI: The AI Arms Race” for its 57-page September opus that emphasizes the potential for generative AI to democratize tech innovation and increase worker productivity. “We believe it is a game changer,” the report’s authors put it plainly.

Most investment banks and Wall Street research firms issued several major AI reports this year, as well as a smattering of AI-focused single stock notes. But a large portion of the ink spilled on this trend has been devoted to rebutting concerns that the recent AI enthusiasm is overblown or premature. In a July report titled “Generative AI: Hype, or Truly Transformative,” Goldman Sachs spoke with AI experts on opposite sides of the debate. In the piece, Gary Marcus, professor emeritus of psychology and neural science at New York University, argued generative AI is being “overhyped” and true artificial general intelligence is a long way from reality. But Sarah Guo, founder of the venture capital firm Conviction, made the case that it will usher in an era where technology becomes “more useful, accessible, and less expensive.”

Goldman followed up its July report with another, slightly more defensive, offering in September called “Why AI is Not a Bubble” that emphasized the potential of the technology and argued that AI-related stocks were not trading at the typical bubble prices. “We believe we are still in the relatively early stages of a new technology cycle that is likely to lead to further outperformance,” the Goldman team wrote of the potential for AI-linked stocks.

Some veteran Wall Street analysts even leaned into the AI bubble talk, arguing that bubbles are necessary, unavoidable, and even maybe a net positive that can help a truly revolutionary technology develop more rapidly.

“Bubbles ensure that more than sufficient amounts of capital are invested into the emerging technology to give it the greatest chance of breaking out. Without ‘excess’ capital chasing every bad idea, especially bad ideas that were actually great, we may have never enjoyed the full benefits of railroads, the automobile, or the Internet,” Gene Munster, managing partner at the investment firm DeepWater Asset Management, explained in an Oct. 17 report titled “The AI Bubble Playbook.”

Hype-fed “bubbles get an unfairly negative connotation,” Munster argued in the report, before detailing how investors can play the ups and downs of the AI hype cycle. The veteran tech analyst recommended investors look to hardware companies that create the technologies that allow AI to operate, like the semiconductor giant Nvidia — even after its over 180% year-to-date rise. But mainly, Munster highlighted cloud storage and software firms that will use AI to make data more accessible and usable, like Microsoft, MongoDB, and Databricks, as well as AI application software firms, like the gaming focused developer Unity Software or the data science and analytics firm Alteryx.

“AI might be frothing, but objective comparison to a bonafide bubble says we’re nowhere near roaring insanity,” he concluded.

Wedbush tech analyst Dan Ives also played on the bubble talk in a June report that detailed the “AI gold rush.” Ives was one of the first to compare the rise of AI to that of the internet during the early days of the dotcom bubble. On the AI timeline, Ives believes it’s 1995, not 1999 — meaning this bubble has a lot of room to inflate before it bursts, and investors should go along for the ride. To his point, the enthusiasm over AI can be seen in corporate earnings calls and presentations. Mentions of the term “AI” in corporate earnings reports, press releases, and other public filings have soared 363% since 2018, according to the market intelligence firm AlphaSense, which ironically uses AI to delve through financial data.

Still, as Fortune previously reported, some experts fear that investors on both Wall Street and Main Street are getting a bit too excited over AI’s potential. “We’re in a hype curve—a bubble,” Robert Marks, an electrical and computer engineering professor at Baylor University, said in a July interview. “And I think that people have to slow down and be a bit more sober in terms of their thinking.”

David Trainer, founder of investment research firm New Constructs, added that he believes investors’ fear of missing out is driving some stocks to unsustainable levels. “Investors just have to be really careful,” he said.

Despite all the warnings from experts, you’d be hard pressed to find a major investment bank or hedge fund that isn’t an AI believer — at least over the long term. If you ask the Street, AI is akin to the “advent of electricity” and one day may be something like a “digital god.” Here’s a look at some of the highlights from Wall Street’s latest research reports on its AI obsession.

Bank of America: “Reality justifies the hype”

Key quote: “It’s not just tech companies that will benefit from AI’s newest wave. Our view is that GenAI’s transformation potential will likely be similar to the transformation driven by past disruptive technologies like the telephone, automobile, personal computer and internet, which have historically reached mainstream adoption after 15-30 years.”

Key statistic: Corporate AI implementation could cut S&P 500 companies’ costs by ~$65 billion over the next 5 years and AI’s economic impact could reach $15.7 trillion by 2030, according to BofA’s estimates.

This quote and statistic were pulled from Bank of America Research’s October AI reports, “AI Evolution: Reality Justifies the Hype” and “AI Evolution: Stock & ETF Beneficiaries,” written by analysts Alkesh Shah and Andrew Moss.

Carlyle: AI is as revolutionary as electricity

Key Quote: “Generative AI has been analogized to the advent of electricity, and this comparison may be apt…We must also be mindful of the ‘hallucination problem’ with LLMs [Large Language Models]…human verification of their outputs will still be required in many cases, and their use in mission critical applications like aeronautics or defense could lay very far in the future.”

Key statistic: Between 60% and 70% of employee workloads could eventually be automated by generative AI applications, according to a McKinsey study.

This quote and statistic came from an October Carlyle report titled “Brave New World: AI and its Downstream Implications” by Jason Thomas, head of global research and investment strategy, and Michael Wand, managing director and co-head of the Carlyle Europe Technology Partners investment advisory team.

UBS: “A generational transformation”

Key quote: “While not a ‘digital god’, Generative AI's capabilities have advanced at a remarkable pace. The costs of developing and training models today typically run to many millions of dollars, but open-source models are gaining ground and adoption of the technology appears set to be swift.”

Key statistic: Some 40% of all working hours could be impacted by the rise of generative AI, according to data from Accenture.

This quote and stat came from a 131-page UBS report titled “Will Generative AI deliver a generational transformation?” that was written by a dozen UBS analysts led by Michael Briest and published in May.

Citi: AI will handle your marketing—and much more

Key quote: “As a concept, artificial intelligence (AI) is not new, and Generative AI represents the latest inflection point in the evolution of AI. However, what is distinctive about Generative AI is the tremendous potential it holds to transform work across industries and boost overall productivity.”

Key statistic: By 2025, 30% of outbound marketing messages from large organizations will be synthetically generated, up from roughly 2% last year, according to estimates from the industry research firm Gartner.

This quote and statistic is from a September 57-page Citi report, “Unleashing AI: The AI Arms Race”, which was written by a group of dozens analysts and industry experts.

Goldman Sachs: AI could boost global GDP 7%

Key quote: “Current valuations in the technology sector are not as stretched as in previous bubble periods and the 'early winners' that have enjoyed the strongest returns have unusually strong balance sheets and returns on investment.”

Key statistic: Productivity growth could rise by just under 1.5% annually over the next decade if AI undergoes widespread adoption, boosting global GDP by 7%, according to a Goldman Sachs study.

This quote and statistic comes from a September 5 report titled “Why AI is not a bubble” which was written by a group of experts and analysts, led by chief global equity strategist Peter Oppenheimer.

Morgan Stanley: Don’t believe the unemployment hype

Key quote: “For employment, I continue to be skeptical of the gloomy prognostications that a myriad of workers will lose their jobs to new technology. Economies evolve, and the advent of the lightbulb has not left a lingering horde of unemployed candlemakers.”

Key statistic: Monthly downloads of AI models on Hugging Face, the most popular repository for open-source AI models, have soared this year. Between January 2021 and August 2023, the top 10 models on Hugging Face have been downloaded 3.5 billion times.

Morgan Stanley has written various major AI reports including its “AI Guidebook,” but this quote and statistic are from the August reports “AI: The Long and Short of It” by chief global economist Seth Carpenter and the “AI & The Annual Beauty Contest” by equity strategist Edward Stanley.

Deutsche Bank: AI will “augment rather than replace” workers

Key quote: “Job losses have been one of the most immediate concerns about generative AI, despite historical evidence showing that over the long term technological revolutions have typically led to new, higher quality jobs, greater productivity and economic growth...However, emerging evidence over the summer suggested some fears may be unfounded, with AI likely to augment rather than replace white-collar jobs.”

Key statistic: The potential effects of AI-based automation on employment are much more drastic in high-income countries. Some 5.5% of total employment in high-income countries is exposed to automation effects, compared to just 0.4% in low-income countries, an August International Labour Organization （ILO） study shows.

This quote and statistic are from a September research note titled “AI Update: Six key themes you missed while you were at the beach” written by Deutsche Bank’s head of global economics and thematics research, Jim Reid, and two of research analysts, Adrian Cox and Galina Pozdnyakova.

Capital Economics: Like the dotcom bubble on the way up

Key quote: “AI will have a similar impact on equities to the one that the Internet had during the second half of the 1990s, when the “dotcom bubble” was inflating…Of course, bubbles eventually burst. But it could be years before that occurred in this case. Regardless, we don’t think that would sound the death knell for AI – any more than the bursting of the dotcom bubble, which was at least five years in the making, sounded one for the Internet.”

Key statistic: Productivity gains during the internet revolution hit 1.5% per year in the U.S., and that could be a “reasonable guide to what is achievable” during the AI revolution as well, according to Capital Economics.

This quote and statistic are from two September research reports titled “AI, Economies and Markets – How artificial intelligence will transform the global economy” and “How we see AI playing out in stock markets.”