Artificial Intelligence: The Next Wave

Futurist and Stanford professor Roy Amara noted, in what has become known as Amara’s Law, that “we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

Since OpenAI introduced ChatGPT on November 30, 2022, the attention paid to Artificial Intelligence (AI) by corporations, academia, media and governments has expanded at a rapid pace. Companies that either have, or are perceived to have, ties to AI have seen their share prices soar, with investors seeming to believe that AI will be the next great investment opportunity. AI offers the hope of a new technological revolution, whereby software will be able to handle vast amounts of data, solving complex problems in a more efficient manner than has been possible. If this comes to fruition, it creates the potential for AI to complete complex interactive tasks that, until now, have been solely completed by humans.

In response to the hype about AI, significant efforts have been made to forecast its developmental trajectory and identify its most promising applications. However, the nascent technology's inherent complexity, along with the broader economic and social dynamics surrounding such a potentially transformative technology, make this forecast particularly challenging. As we consider the range of potential outcomes, we would probably be well served to keep Amara’s Law in mind.

AI impacts on the economy

McKinsey’s Technology Trends Outlook 2024 estimates that the use cases for generative AI (gen AI) have the potential to produce annual value of $2.6 trillion to $4.4 trillion. To put this in context, Japan’s annual GDP as of April 2024 was $4.1 trillion, according to the International Monetary Fund.

AI has the potential to impact the economy across multiple dimensions. For example, not only could it significantly affect labor market productivity, but it could also influence the allocation of capital expenditures as substantial investments will be required to produce the computer hardware, data centers and energy infrastructure necessary to support its widespread adoption.

Labor productivity

Much of the narrative surrounding the impact of AI is centered on its potential to improve labor productivity by changing how employees do their work, and enhancing the ability of workers by automating some of their current tasks. A main driver of AI’s ability to enhance productivity comes from its ability to understand natural language rather than being dependent upon specialized programming language. This allows AI applications to have shorter training cycles and much greater ease of adoption.

While the potential for AI to make a significant impact on productivity seems to be real, we should keep in mind that this is not the only dimension of productivity where technological change may have an impact. It can also bring about a wider array of consumer choices, create brand new markets for goods and services, change the nature of labor demand, and lead to the reconfiguration of organizations and their processes related to production and service.

Hardware investments

Graphics processing units (GPUs) are required to run the vast numbers of calculations demanded by AI, and investment in these GPUs has been front and center in the narrative around AI. The advancements made in the last decade have led to a leap in their performance and energy efficiency. While the GPUs were originally designed for graphic imaging purposes, their ability to handle vast amounts of data and complex calculations make them perfect for powering AI technology. The development of additional tools and computing frameworks has made GPUs much more widely accessible for general-purpose computing, allowing them to be utilized in ways well beyond their original graphics applications.

Energy infrastructure and data centers

We have already seen significant capital investments to build and develop the data centers that will be necessary to run AI systems and programs. These infrastructure projects have contributed to the investment narrative centered on AI-driven energy usage. Currently, data centers are responsible for approximately 3% of U.S. power demand. Goldman Sachs projects that U.S. data center power demand (excluding cryptocurrency) will grow by 160% from 2023 to 2030, resulting in data centers expanding their share of U.S. power demand to approximately 8% by 2030. They estimate that this growth in demand will require an additional 47 GW of power generation capacity.

Demographic trends, interest rates, geopolitical shocks and the overall price level will also be important economic variables that could support or dampen AI-driven economic expansion. For example, increased labor productivity driven by AI may be able to help alleviate the shortfall of workers in economies facing demographic challenges as a result of aging populations. But challenging demographics can also be deflationary if they offset some or all of the potential growth that AI could stimulate.

AI’s impact on investment markets

The AI narrative has been credited with driving a stock market rally led by gen AI. This rally was seen primarily in the so-called “Magnificent Seven” mega-cap technology stocks. These seven stocks were up 76% in 2023 and an additional 33% in the first half of 2024. Their returns were responsible for 62% and 63% of the total return of the S&P 500 Index in 2023 and the first half of 2024, respectively. Given these spectacular returns, it’s fair to wonder if there is a bubble developing in these stocks, similar to the internet bubble of the late 1990s.

One significant difference between today’s AI companies and the earlier dot-com boom is the generally superior valuations and estimated earnings growth of the current market leaders. In addition, at this point, markets don’t seem to be valuing companies on ephemeral criteria, such as “eyeballs,” which was the case during the internet bubble. In fact, while current tech companies comprise a larger combined weight in the S&P 500, their valuations have not soared as high as their brethren from 2000, despite having higher near-term earnings expectations.

While the valuations of these companies do not seem to be extreme, especially when compared to previous market bubbles, there are other risks that investors should keep in mind. One potential risk could be the concentration of capital investment in the AI space among a small number of companies. Should the pace of monetization of these investments lead to pressure for companies to reduce their capital expenditures, both the chip manufacturers and other suppliers of AI infrastructure will be negatively impacted. Other risks could arise if advances in computing power and chip efficiency lead to lower demand from customers, or if the adoption curve for AI applications isn’t as steep as seems to be expected.

While some investors may be tempted to chase the recent returns of AI-associated stocks, they should keep in mind the difficulty of predicting which companies will win during times of significant technological change.

Valuations for market leaders lower in 2024 than in 2000

Dot-com boom beneficiaries

AI adoption beneficiaries

Source: Goldman Sachs and Corient, as at 6/30/2024. This information is not intended to be an endorsement of the companies mentioned or a recommendation, offer, or solicitation to buy, sell or hold any securities. Equity investing is subject to market volatility, liquidity risk, and complete loss of principal. CAGR is the Compound Annual Growth Rate. NTM P/E is the price-to-earnings ratio, where price is the share price of the reference stock and earnings are measured as the average Wall Street analyst estimate for earnings per share for the stock over the following 12 months.

Amazon: A short case study

Amazon offers an example of how difficult it can be to predict the future. In the late 1990s, Amazon was an online book seller that also sold a broad array of consumer goods. In mid-December 1999, Amazon reached a market capitalization of $36 billion. In 2001, after the internet bubble burst, its market capitalization fell to $2.2 billion, down 94% from its previous high. In 2024, Amazon is now a global leader in online retail as well as a web services provider and entertainment company, with a market capitalization above $2 trillion.

The spectacular growth of Amazon was due, in many ways, to their ability to use a new technology to build a new way of doing things. In the early days of the internet, it was not a pure “internet company,” but a retailer that utilized the internet to change the way customers shopped, first for books, then a significantly wider range of consumer goods. It was only much later that Amazon became a web services company, which opened a new avenue of growth.

In a similar fashion, there may be companies that today would not be considered an “AI company,” but may leverage AI technologies to grow their business and serve their customers in new ways, much like Amazon was able to do.

There’s no guarantee AI will produce changes on the same scale as the internet, but if it does, we shouldn’t be surprised if picking the long-term winners proves to be just as difficult as predicting Amazon’s rise to the behemoth it is today from its much more modest beginnings in the 1990s.

For AI to have a maximum impact for investors, companies will need to convert AI technology into a competitive business advantage that ultimately drives revenue and earnings growth, which tend to influence long-term market returns.

Where do we go from here?

While the hype surrounding AI seems to have exploded into public consciousness over the last few years, AI applications have been in our lives for some time now. AI has already been a part of the technology in our smartphones, as a part of marketing and sales tools, customer service chatbots and the autonomous driving features in automobiles. It was only with the release of ChatGPT in 2022, which allowed individuals direct access to AI applications, that AI moved to the forefront of public consciousness. 

We are still in the early stages of AI adoption and the exact path forward has yet to be determined. In thinking about how the evolution and development of AI might proceed, it’s worth casting our eyes back to the internet mania of the late 1990s. At the time, there were forecasters who believed that the internet would, rather quickly, render the office environment a relic of the past and lower the value of real estate investments in office buildings, as large numbers of workers would begin working from home. They were right, and all it took for their predictions to come true was more than 20 years and a worldwide pandemic that nearly brought the global economy to a halt. It can take time, and often some sort of exogenous shock, for technological changes to become fully adopted across society.

The potential for AI to produce broad, structural changes for individuals, companies and the economy certainly seems real. Exactly how these changes occur will only be certain with hindsight. In the meantime, within a well-diversified portfolio spanning a wide range of industries, investors can gain exposure to companies – both large and small, public and private – that may benefit from AI. Along the way, there will be risks that will only reveal themselves with time, and blind alleys of growth that end up going nowhere.

Long-term success in technology investing will likely accrue to those who are patient, strategic and mindful of Amara’s Law.

ABOUT THE AUTHOR