Joseph Davis expects AI technology to resemble electricity, benefiting the users more than the providers

At Vanguard, we anticipate 3% US GDP growth in 2027, an estimate that is noticeably higher than other professional forecasts, implying continued strong support for risk assets. Such growth will not represent an incremental improvement, but rather a fundamental shift in the economy’s growth trajectory. Following a data-rich study of current AI capabilities and how the technology compares to other game-changing technologies of the past, we foresee an economic sea change.

In fact, AI is already boosting economic activity, even if it may be another year or two before we know for sure whether it will be as transformative an economic force as the personal computer. For our forecast to prove true, AI will need to move past its current automation phase, where it simply replaces human tasks, through an augmentation phase, where it makes workers better at their jobs, and eventually enables products, services, and industries that we have not yet envisioned. Today’s focus is on automation, but it is the realisation of these last two phases that will determine whether AI ever becomes a general-purpose technology.

Before electricity became economically viable, few people imagined electric streetcars, movie theaters, or household appliances. A similar expectation that AI will mature into a general-purpose technology makes us sanguine about the prospects for the labour market. Concerns about job displacement are understandable, but doomsayers who envision a dystopia of sidelined prime-age workers tend not to think about the jobs that have yet to be created.

Nor do they give enough attention to the higher incomes (and spending) that would come from workers boosting their productivity tenfold. Just ask accountants how their lives changed after computer software transformed their field. Their output increased dramatically. Disruption need not only mean automation. Moreover, fully realised AI would unlock productivity gains that would compensate for—and likely exceed—the headwinds of aging populations, declining fertility, and reduced immigration.

Still, the journey from AI investment to widespread productivity enhancements will unfold over years, not quarters. (We can look to 1997, during the internet buildout, for a historical parallel.) The current investment phase has at least a year or two to run, despite its eye-watering scale to date. Deep-pocketed AI hyperscalers appear capable of following through on their historic capital commitments, and companies are deploying AI tools in earnest.

Markets, however, have raced ahead of economic reality. Equity valuations, particularly for large-cap US technology companies, already assume much of AI’s potential upside. For the next year or two, strong earnings growth from AI-enabling investment can perhaps justify these valuations and may well drive markets even higher. But that’s a near-term phenomenon. Over longer horizons, the mathematics of investing tends to shift, especially during periods of rapid technological change.

History offers a clear lesson here. The companies building transformative technologies rarely capture the greatest long-term value. Instead, those benefits accrue to the users. Electricity created more wealth for manufacturers that could run assembly lines around the clock than for power utilities. The automobile enriched suburban developers and retailers more than the automakers themselves.

AI may well reproduce this pattern. The current buildout phase—dominated by hyperscalers, chip makers, and foundation-model developers—will give way to a consumption phase where end users across industries derive the greatest benefits. Such businesses are currently trading at value-oriented multiples, and many are outside the United States in service-oriented economies with aging populations, where more productive workforces will be a blessing.

What types of companies may benefit? Health-care providers will have ample opportunity to automate administrative tasks and enhance diagnostic accuracy. Financial services firms will be in a position to deliver even more personalised advice at an even lower cost. Business services companies could augment human expertise with AI-powered analysis. Such companies are starting to explore where they can automate tasks, and they will reap the rewards if AI eventually augments workers’ skills and fulfills its promise.

Of course, we cannot say with certainty that AI will positively transform the economy. But there will be clues: younger workers entering the labour force with AI-enhanced skills; startup creation accelerating outside the technology sector; and more frequent, genuine discoveries (such as a breakthrough in medicine) from AI-assisted research. As these patterns emerge, we will likely be seeing AI’s economic transformation in its infancy. It will look much like the path that electricity and the PC took.

The opportunity that is beginning to emerge lies in recognising that markets may be correctly assessing AI’s economic potential while incorrectly pricing where the benefits will accrue over the full cycle. Value-oriented US equities, non-US developed markets, and high-quality fixed income all offer compelling risk-return profiles—defensive if AI falters, opportunistic if it succeeds—over the coming five to ten years.

The point is not to abandon technology exposure or try to time the market. It is to recognise that in an AI-transformed world, we will have moved from the current phase, where AI builders dominate, to one where AI users should increasingly command attention. It happens every time a great technology transforms the world. For long-term investors, this future transition represents both a risk to manage in growth-heavy portfolios and an opportunity to seize ahead of the AI revolution’s next phase.

*Joseph H. Davis is Global Chief Economist at Vanguard and Global Head of the Investment Strategy Group. Copyright: Project Syndicate, 2026, published here with permission.