Estimate Time8 min

Investing in AI: What you should know

Key takeaways

  • A new development in artificial intelligence (AI) that claims to provide high-performance outputs at a lower cost than current AI models has many wondering what the future may hold for this much-talked-about technology and the companies who depend on it.
  • While it’s unclear exactly what effect this may have in the long term, it may be possible to estimate the overall economic impact of AI by examining the historical relationship between technological innovation and productivity.
  • A recent Fidelity Institutional analysis found that a major AI-driven acceleration in productivity in the next decade is unlikely; however, even a modest boost in the short term could be powerful though much is still unclear.1

Recent news about an innovative development in artificial intelligence (AI) technology has introduced new uncertainty into the market—and new potential.

Although the accuracy of these reports has yet to be verified, a new AI model claims to be able to provide performance that is competitive with current industry leaders such as Open AI, Google, and Meta—but at a fraction of the cost and with far fewer chips.

Though this news led to some initial volatility, markets steadied themselves as the potential opportunities presented by this new innovation became more apparent. For example, the ability to achieve higher performance with a smaller number of chips may, in the long run, potentially benefit those more established companies, which could make use of their access to a larger number of chips to offer more powerful services. Still, the full impact of the new technology may take time to develop.

"As more information comes out, the story may evolve further," says Naveen Malwal, institutional portfolio manager for Strategic Advisers, LLC. "There are varying opinions about the earnings outlook for the leading AI companies at the moment, but it's important to consider the potential value these companies still hold and the long-term outlook they may maintain."

"Long-term investors need to consider whether headlines such as these will meaningfully impact the pace of economic growth in the US or globally or the profit outlook for the market," says Malwal. "That sort of thing is rarely upended by news about a single breakthrough at one company."

Malwal says that such volatility demonstrates the benefits of diversification. "When something occurs that affects a subset of the market, such as AI stocks or domestic stocks in general, there may be other parts—bonds or international stocks—that don’t feel as much of an impact," says Malwal. "Having a diverse portfolio can potentially help mitigate the impact of market volatility and allow you to stay focused on your long-term goals, although it does not guarantee immunity from emotional reactions to market downturns."

Insights from Fidelity Wealth Management

Get our exclusive Fidelity perspective with Insights from Fidelity Wealth ManagementSM


AI: The long view

Though the impact of this new innovation will take time to discern, history may provide some insight into AI’s economic prospects in the long run. According to Irina Tytell, an economist and team leader on Fidelity’s Asset Allocation Research Team (AART), it might be possible to estimate the effect that AI could potentially have on labor productivity and profits in the coming years. By gauging historical capital investment patterns and examining the adoption of important technological innovations of the past, we may be able to develop a more informed perspective on what to expect from AI in the next 5, 10, or 20 years—and better appreciate how much uncertainty still surrounds it.

What is AI?

When discussing AI, it’s crucial that we be clear about exactly what it is we’re talking about.

Artificial intelligence (AI) is a general term that encompasses technologies capable of accomplishing tasks that typically would have required human reasoning to achieve. Within AI, “machine learning” refers to processes that operate without explicit programming—in essence, these technologies make decisions on their own, based on learnings they have synthesized from large sets of data. “Generative AI,” perhaps best known through tools like ChatGPT, is an application of artificial intelligence, using patterns found within those large datasets to create original content, such as text, images, or video, on the fly.

AI technology may show great potential in a variety of sectors, with a wide-ranging set of possible applications, such as powering chatbots and voice-activated personal assistants; writing and summarizing content or computer code; and improving the speed and accuracy of medical diagnoses. Given the potential efficiencies and significant improvements in productivity these applications may represent, it’s no surprise that many companies—and investors—have taken an interest.

But how realistic are these expectations, and when might the impact of AI be seen in the broader economy? According to a December 2024 survey, only 6.3% of businesses reported using AI tools, and a number of important headwinds may still need to be addressed before AI can reach widespread adoption.2

What might slow the adoption of AI technologies?

While new technologies always face challenges, 4 in particular stand out when discussing AI.

Financial costs

"In terms of digital technology, AI may be the most expensive one we’ve ever had," says Tytell. To date, the cost of standing up an AI platform has been substantial. OpenAI’s GPT-4 is estimated to have cost over $78 million to train.3 Tytell also notes that data costs, which have largely not been a factor to date, could also rise in the future. "These models have already consumed most of the free data that’s available," she says. "If they need access to more data, where is that going to come from and who is going to pay for it?"

Environmental costs

It is estimated that training OpenAI’s GPT-3 took 1,287 MWh, or about as much electricity as 120 US households consume in a year, and generated 552 metric tons of carbon emissions.4 This does not include the additional power needed to facilitate use of the model on an ongoing basis. While new advances in AI may alleviate this somewhat going forward, it’s possible that more efficient models may lead to an overall increase in power demand. This could still constrain adoption—and draw the attention of regulators.

The impact on workers

"AI is a labor-saving technology," says Tytell. “By definition, such technologies tend to displace certain workers. On the other hand, they can strengthen the prospects of occupations that are complementary to the technology and perhaps create new jobs, as well." In fact, according to one analysis of US Census data, more than 60% of jobs today did not exist before World War II.5 "New jobs are what power the labor market in the long run," says Tytell.

Fear, uncertainty, and doubt

There are many questions about the accuracy and reliability of AI outputs, as well as legal uncertainties with regard to copyright (for generative AI models), discrimination (when used in hiring), data privacy, and cybersecurity. In short, there’s still a lot we don’t know. "In a high-stakes environment where accuracy is really important, it’s likely that adoption will be limited until these issues are improved sufficiently," says Tytell.

Furthermore, Tytell cautions, it’s simply not easy to roll out a new technology—it takes time and effort. "The need to invest capital both in order to make use of the technology and to educate workers on how to use it most effectively is significant," says Tytell. "Because of that, it could take a while to see the productivity gains."

Looking back at the development of other technologies that required changes in worker skills and business practices, we can see that adoption was a slow and methodical process—in some cases taking as long as 20 years to achieve 50% penetration.

This chart shows the adoption rate of past technologies from 1830 to the present. It examines the rise over time of railway miles, steel production, telephones, motor vehicles, radios, electricity production, airway miles, personal computers, internet users, industrial robots, and cell phones. Generally, the adoption of these technologies was a multi-decade process.
Railway miles per 1,000 people, steel short tons per employee, telephone lines per household, electricity 10 megawatts per household, automobiles per household, radios per household, airway miles per 1,000 people, televisions per household, households with computers, cell phones per person, internet users per household, robots per 1,000 manufacturing employees. Sources: "Historical Statistics of the United States, Earliest Times to the Present: Millennial Edition," edited by Susan B. Carter et al., New York: Cambridge University Press, 2006. Diego Comin and Bart Hobijn, "An Exploration of Technology Diffusion," The American Economic Review, Vol. 100, No. 5 (December 2010), pp. 2031–2059. World Bank, U.S. Census Bureau, Bureau of Labor Statistics, Haver Analytics, Macrobond, Fidelity Investments (AART).

How might AI impact the economy?

Tytell is the author of a recent Fidelity Institutional analysis that found that a major AI-driven acceleration in productivity in the next decade is unlikely. Sizable productivity gains are more likely to be seen when and if AI technologies approach 50% adoption, perhaps 10 to 15 years out from the present.

Using 3 approaches (aggregating experiences of past technologies, extrapolating technological adoption patterns by sector, and deriving estimates from capital spending), the analysis estimates a potential productivity boost of 0.2%–0.3% over the next decade, with gains of 0.5%–0.9% in the lead-up to widespread adoption.6

This chart shows the estimates of AI productivity calculated using three different methods. In the 10 years after reaching 5% adoption, productivity is estimated to be between 0.2 and 0.3%; in the 10 years before reaching 50% adoption, between 0.5 and 0.9%. The potential for AI to increase in adoption from 5% to 50% is a possibility that could occur over time.
Source: Fidelity Investments (AART), as of 5/31/24.

"Set against the 1.4% productivity rate of the last decade, these estimates are encouraging. Even a modest productivity boost would help counter demographic headwinds and support stronger economic growth," says Tytell. “However, with any new technology, there needs to be significant capital investment to get it going. There’s a very clear association between capital investment in the present and productivity in the future, and I don’t think we’ve seen enough of a pickup in the former to make us optimistic about the latter. It’s happening, but not yet in a big way.”

Much remains to be seen

It’s easy to get swept up in the excitement of a new technology, but patient analysis shows that a more measured approach may be the best way to take advantage of technological developments without getting too far ahead of things.

"My impression is that very often, as people, we tend to overstate short-term consequences and understate long-term consequences," says Tytell. “And I think AI is no exception to that. It is a very powerful new technology that has the potential to enhance productivity. But people may be expecting too much too soon. We may need to be a bit more patient. The potential benefits may be substantial, but they may also be further in the future than some anticipate.”

Start a conversation

Already working 1-on-1 with us?
Schedule an appointmentLog In Required

More to explore

1. Tytell, Irina. "Artificial intelligence: An X-factor in a new investment regime," Fidelity Institutional, July 2024.
2. Sources: Census Bureau’s biweekly Business Trends and Outlook Survey, Fidelity Investments (AART). December 2024.
3. Nestor Maslej, Loredana Fattorini, Raymond Perrault, Vanessa Parli, Anka Reuel, Erik Brynjolfsson, John Etchemendy, Katrina Ligett, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Russell Wald, and Jack Clark. (April 2024). “The AI Index 2024 Annual Report,” AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, Stanford, CA.
4. D. Patterson, J. Gonzalez, et al. (2021). "Carbon Emissions and Large Neural Network Training."
5. David Autor, Caroline Chin, Anna Salomons, and Bryan Seegmiller. (2024). "New Frontiers: The Origins and Content of New Work, 1940–2018." The Quarterly Journal of Economics, Volume 139, Issue 3, August 2024, Pages 1399–1465; David Autor. (2022)."The Labor Market Impacts of Technological Change: From Unbridled Enthusiasm to Qualified Optimism to Vast Uncertainty." NBER working paper 30074.
6. Tytell, Irina. "Artificial intelligence: An X-factor in a new investment regime," Fidelity Institutional, July 2024.

Keep in mind that investing involves risk. The value of your investment will fluctuate over time, and you may gain or lose money.

Past performance is no guarantee of future results.

This information is intended to be educational and is not tailored to the investment needs of any specific investor.

Stock markets are volatile and can fluctuate significantly in response to company, industry, political, regulatory, market, or economic developments. Investing in stock involves risks, including the loss of principal.

Views expressed are as of the date indicated, based on the information available at that time, and may change based on market or other conditions. Unless otherwise noted, the opinions provided are those of the speaker or author and not necessarily those of Fidelity Investments or its affiliates. Fidelity does not assume any duty to update any of the information.

Investment decisions should be based on an individual’s own goals, time horizon, and tolerance for risk.

Fidelity Brokerage Services LLC, Member NYSE, SIPC, 900 Salem Street, Smithfield, RI 02917

1186399.1.2