How much could AI boost US stocks?

Published on14 JUN 2023
Artificial Intelligence

As investors debate the merits of generative artificial intelligence, a key question is what it could mean for the stock market. Take Nvidia for example: last month the semiconductor firm said it expects $11 billion of sales in the current quarter — fully 53% above the $7 billion consensus estimate amid AI-inspired demand for its advanced chips. It’s now the fifth-largest company in S&P 500 Index of U.S. stocks. The leap in Nvidia’s outlook was a galvanizing development for investors, who are assessing how much influence generative AI will have on revenue growth and profitability of other companies, according to Goldman Sachs Research.

Generative AI may boost corporate earnings, but the magnitude and timing are uncertain, Goldman Sachs Research equity strategist Ryan Hammond writes in the team’s report. Using our economists’ estimate that AI adoption could boost productivity growth by 1.5 percentage points per year over a 10-year period, Goldman Sachs analysts estimate that the S&P 500’s compound annual growth rate in EPS over the next 20 years would be 5.4%, compared with 4.9% that their dividend discount model currently assumes. That would point to an S&P 500 fair value that’s 9% higher than today.

Semiconductor companies are the most direct beneficiaries of generative AI because of the massive computing power necessary to drive large-language model algorithms, according to Goldman Sachs Research. In addition, our technology equity analysts estimate the total addressable market for generative AI enterprise software is around $150 billion, assuming a 30% adoption rate. This pool of potential revenues represents just 1% of current aggregate S&P 500 sales and will likely take years to access. However, “increased economy-wide output could translate into increased revenues and earnings for S&P 500 companies, even beyond those firms directly involved in the development of AI,” our analysts write.

Given that it’s hard to predict the timing and ability of companies to generate profits from AI, our analysts say it’s unlikely to be fully priced in by investors in the near term. Using a range of productivity scenarios, upside to the S&P 500’s fair value could be as small as +5% and as large as +14%. The boost to U.S. equities could also be amplified by an increase in profit margins.

At the same time, policy responses like higher tax rates could cut into an AI-assisted increase in earnings, according to Goldman Sachs Research. Our economists expect some jobs to be lost due to AI adoption. Meanwhile, U.S. corporate profits as a share of GDP are at an elevated level relative to history, while wages as a share of GDP remain near a historical low. And even though our economists also predict — and history suggests — that many workers who are displaced by AI automation will eventually become reemployed, this process could take time. “Policymakers may be sufficiently concerned about the economic and social impact of the interim displacement of workers to take action to support those workers,” according to Goldman Sachs Research.

Higher interest rates could also counteract much of the potential increase in fair value for the S&P 500 Index. “A productivity boom that leads to lower prices could be disinflationary and put downward pressure on interest rates,” our analysts write. However, large government spending programs, such as universal basic income, could also put upward pressure on interest rates. Goldman Sachs economists note that AI could increase investment demand and in turn raise estimates of the neutral rate (the interest rate that neither stimulates not restricts the economy), which is a key input for monetary policymakers. In our analysts’ dividend discount model, a 30-basis-point increase in interest rates would fully offset the upside from AI adoption in the central scenario, all else equal.

Equity prices, meanwhile, tend to closely track dynamics in the business cycle — investors’ long-term bullishness on AI contrasts with near-term bearishness due to recession risks. “Deteriorating economic growth would likely more than offset the potential long-term boost from AI adoption, especially given the extreme uncertainty around AI implementation,” our analysts write. AI adoption would need to translate into a 1.5 percentage point increase in long-term EPS growth estimates to offset the typical 15 percentage point cut to near-term EPS growth estimates around recession.

Furthermore, increases in productivity haven’t always boosted the S&P 500. Five-year productivity growth can explain just 1% of the variation in five-year S&P 500 returns, according to Goldman Sachs analysts. Our economists note the U.S. equity market appeared to price the impact of electricity in the 1900s after the productivity boom was realized. By contrast, at the end of the century, following the widespread adoption by businesses of the PC along with early-stage use of the Internet, the S&P 500 index priced the innovations’ impact as the productivity boom was realized, returning 26% annually between 1994 and 1999, near the peak in productivity growth.

And while AI-driven advances have promise, our analysts point out that the dot-com boom shows the risk from high investor expectations. During the late 1990s, fast-growing technology firms saw their valuations collapse as they failed to meet optimistic investor expectations, even as their sales grew.  “Stocks with high perceived growth potential are rewarded with high valuations, but investors flee and the multiple collapses at the slightest sign that the growth rate may not be sustainable,” our analysts write.

That doesn’t appear to be the case at the moment, at least not at the broader index level, according to Goldman Sachs Research. The S&P Index’s equity risk premium and long-term EPS growth expectations are roughly in line with historical averages, suggesting investor optimism on AI adoption is not at extreme levels. However some individual stocks, such as the largest AI beneficiaries, have valuations that resemble dot-com beneficiaries during the Internet boom.

Explore More Insights