How AI Can Boost Productivity and Jump-Start Economic Growth

Our analysis quantifies the potential economic benefits of artificial intelligence. We see it as a potentially transformative force.

Artificial intelligence (AI) sparks strong opinions. Optimists call it a revolution that will reshape society. Skeptics warn it could be another overhyped bubble. In our view, AI , especially generative AI has the potential to deliver meaningful, long-lasting economic impact. The road ahead remains uncertain but several powerful drivers could accelerate its progress.

In this article, we take an economist’s lens to the central debate: How might AI affect the broader economy? Key questions include:

• How large and how soon could the productivity gains from AI be?

• Which jobs might be affected, and how could policymakers in the U.S. and Europe respond?

• Will AI push inflation higher or help keep prices in check?

• How quickly will companies actually adopt the technology?

Beyond the macroeconomic picture, we also explore investment implications. While AI-related stocks have seen strong gains, we do not view the space as a bubble. Although a handful of AI-linked companies now represent a sizable portion of U.S. (and increasingly European) market indices, we believe this concentration reflects genuine earnings potential rather than excessive risk. Overall, we see a broad range of opportunities across the AI value chain for investors on both sides of the Atlantic.

Productivity and Growth: Timing and Scale

Productivity growth is the key metric when evaluating AI’s economic potential. Higher productivity allows economies to expand faster, raise living standards and keep inflationary pressures in check.

The U.S. has not enjoyed sustained productivity gains since the 1990s tech boom. Europe has faced even slower productivity growth in recent decades. A return to stronger productivity could open a new chapter of robust economic expansion on both continents.

History offers useful context, let us compare AI to major past breakthroughs like the steam engine, electricity and the personal computer. Each of these innovations eventually lifted productivity but the benefits took time to materialize.

The steam engine required over 60 years to show clear economy-wide gains. Later technologies delivered results more quickly. If this pattern continues, we could begin to see measurable AI-driven productivity improvements in official U.S. and European economic data by the late 2020s. While the personal computer took about 15 years to boost productivity meaningfully, AI could achieve similar effects in roughly half that time.

AI is unlikely to match the economy-wide transformation of the steam engine or electricity. However, we believe it has the potential to be at least as impactful as the internet and personal computer combined, creating substantial value over the next two decades.

Quantifying Potential Job Displacement

To estimate AI’s effects, we adapted a framework from the International Monetary Fund (IMF). Our conclusions suggest AI’s impact could exceed the relatively modest productivity assumptions currently built into forecasts by bodies like the U.S. Congressional Budget Office or the European Commission.

The IMF has identified jobs most exposed to automation by AI. We assume that roughly half of the vulnerable positions in the U.S. and major European economies could be automated over the next 20 years. This could generate a cumulative productivity increase of around 17.5%, adding approximately $7 trillion to U.S. GDP beyond current baseline projections. Similar dynamics could play out across the EU, though outcomes will vary by country and sector.

It is worth noting that technology only lifts overall productivity when it meaningfully reduces the labor or capital required to produce goods and services. Ride-sharing apps like Uber, for instance, did not change underlying inputs (one driver, one car). Autonomous vehicles, by contrast, could.

White-collar roles in professional services, such as data analysis, technical writing and routine administrative tasks, appear more exposed than hands-on jobs like childcare or construction trades. Education levels also matter: Workers with university degrees are significantly more likely to be in AI-vulnerable positions than those with only secondary education.

Globally, advanced economies face higher exposure than many emerging markets. The IMF estimates that up to 30% of jobs in the U.S. could be affected, compared with lower figures in countries like India. Within Europe, northern and western countries with more service-oriented economies may see greater shifts than southern or eastern ones.

All such estimates come with considerable uncertainty. Implementation costs, including expensive AI infrastructure, could slow rollout if not economically justified. Projections vary widely: Goldman Sachs sees up to a 15% GDP boost from AI over the next decade, while our own view is more moderate at 8–9%. Other economists, such as MIT’s Daron Acemoglu, are more cautious, forecasting only 1–1.5% gains.

Economies often evolve in unexpected ways. More than 60% of today’s job titles in the U.S. did not exist in 1940, and many new roles have emerged from previous technological waves. We expect AI to follow a similar pattern, creating new opportunities even as it displaces some existing work.

Policymakers will play a crucial role. Investments in retraining programs, lifelong learning initiatives, and support for workers in transition, already priorities in the EU’s digital strategy, will be essential to ease the transition and maximize AI’s benefits.

Barriers to Corporate Adoption

AI’s ultimate economic influence will depend on how readily companies integrate the technology into daily operations. At present, adoption remains in its early stages. While many firms in the U.S. and Europe are exploring AI, only a small percentage, around 4% , have fully implemented it at scale.

We believe adoption rates will need to reach 50% or higher before AI begins to move the needle on economy-wide productivity. Several challenges could slow progress: shortages of advanced chips, regulatory hurdles (including the EU AI Act), limited energy supplies for data centers and the difficulty of identifying high-impact use cases.

Despite these headwinds, AI offers diverse applications across industries, from optimizing supply chains and customer service to accelerating drug discovery and financial analysis. The breadth of these possibilities supports the case for meaningful macroeconomic effects.

Investing in the AI Infrastructure Buildout

Many AI applications will unfold gradually but the need for supporting infrastructure is already urgent. This includes semiconductors, large-scale data centers and expanded electricity generation and transmission capacity. Training and running advanced AI models consumes significant power, for example, a single query to a large language model can require several times more electricity than a traditional web search.

Demand is rising not only for chips but also for data center construction, specialized engineering, copper wiring, nuclear and renewable energy sources, advanced cooling systems and supporting electrical components.

In the near term, this buildout could exert some upward pressure on prices in specific sectors (such as energy). However, we expect the overall inflationary impact on the broader economy to remain modest. Our estimates suggest cumulative AI-related capital spending over the next five years will total around $400 billion in the U.S. equivalent to just 0.3% of projected GDP, far smaller than the fiscal stimulus seen during the COVID-19 period.

Over the longer term, we anticipate AI will prove more disinflationary than inflationary as productivity gains take hold later in the decade.

Importantly, today’s AI investments are largely funded by highly profitable companies with strong balance sheets and low debt levels, a healthier foundation than during the late-1990s internet boom. Major cloud providers (hyperscalers) are self-financing much of the spending through robust cash flows, reducing reliance on high interest rates or excessive borrowing.

What This Means for Your Portfolio

AI has already driven excitement, investment and earnings growth. For long-term investors, we recommend focusing on the infrastructure layer: semiconductors, data centers, power generation and related enabling technologies. Many companies in these areas, including smaller, less-visible players, have already revised sales forecasts upward in response to rising demand.

While some clients worry that AI stocks are overvalued or that the market is too concentrated, we see important differences from past bubbles. Valuations of leading AI companies, though elevated, remain well below levels seen at the peak of the dot-com era. Moreover, today’s top performers contribute a larger share of actual earnings relative to their market weight than was the case in 2000.

In our view, current valuations are supported by the genuine productivity and earnings potential that AI could unlock. Rather than a speculative bubble, we see the early stages of a genuine technological revolution.

We can help tailor an AI-related investment approach that aligns with your goals.