Putting the economic impact of GenAI into scale

Putting the economic impact of GenAI into scale

June 20, 2024

By Tal Roded

To place forecasts of Generative AI’s impact into scale, in this article we compare projections of Generative AI’s economic impact with that of other technologies from both modern and historical times. We consider impact along two dimensions - GDP and productivity - as these are the two most standard metrics provided by economic reports and provide easy to compare estimates. 


Technological advancement is one of the driving engines of economic growth. These advancements can often occur in waves, such as the Industrial Revolution and the Information Revolution. In other instances, economic booms and productivity boosts are driven by single, impactful technologies. Often cited examples are the modern computer, the Internet, electricity, the steam engine, and, going back even further, the printing press. The newest claimed-to-be revolutionary technology is Generative AI.

The potential for such transformative impact began with the release of GPT-3 in 2020 (and soon-to-follow ChatGPT in November 2022), which spurred the current AI-arms race among technology companies and brought discussion and usage of AI tools to the wider public. As GenAI products based on these innovations proliferate, so do writings and postulations on the economic impact of this technology. Expert claims of Generative AI’s effects run the gamut from a significant boost to the GDP of developed economies to a complete and total restructuring of the global labor market, with the possibility of today’s jobs being performed primarily by GenAI agents in the near future.

GenAI forecasts show lots of uncertainty, potential for huge impact

Despite the massive volume of discussion surrounding Generative AI’s potential effects on the economy, there have been surprisingly few specific, numeric estimates released on the size of its economic impact. Among the few forecasts released, the estimates also tend to be wide and variable. Accenture’s 2024 report and Goldman Sachs’ 2023 report have offered the lowest ranges for impact on global GDP, both with a median estimate of under $1 trillion annually over the next decade. At the top end is an oft-cited McKinsey 2023 report, foreseeing an uplift in GDP of anywhere from $2.6 to $4.4 trillion annually - a 2x to 4x larger impact than the low-end estimates.

Figure 1: Forecasts of Generative AI Impact on Global GDP [source1 source2 source3 source4]. Note: Accenture’s forecast is limited to 22 countries, which may partially account for its lower impact estimate. Goldman Sachs did not provide a range to its estimate. Accenture’s forecast is for impact by 2038, EY’s and Goldman Sachs’ by 2033, McKinsey’s for 2022-2040.

Generative AI’s GDP boost could be equal to the world’s largest economies

The highest end of these forecasts, McKinsey’s 2023 report, would represent an increase in global GDP larger than the current UK economy - the world’s sixth-largest economy. Even on the low end of the estimates, an increase of $1 trillion annually in added value (not just revenue) to global GDP would be a notable boost, representing an increase in the size of the Netherlands’ or Saudia Arabia’s economy, annually.

Figure 2: Top 20 Economies by 2023 nominal GDP, IMF (blue) with Mean Generative AI GDP Impact Forecasts (pink) [source]

Productivity boosts from Generative AI would also be significant

Looking at the forecasted impacts on productivity, we see similarly substantial effects that would represent a global shakeup in economies worldwide. Global productivity growth among OECD nations in the years prior to COVID fluctuated between 0.5-1.6%. In Goldman Sachs’ and McKinsey’s predictions for GenAI productivity impact, there is potential for anywhere from 0.1% to 2.9% annual productivity growth, with the mean forecasts representing a doubling of annual productivity growth due to Generative AI.

Figure 3: Comparing Forecasts of Generative AI Impact on Productivity [source1 source2]. Note: Goldman Sachs’ forecast is by 2033, McKinsey’s for 2022-2040.

Suffice it to say, Generative AI is predicted to have a massive economic impact. To put this technology’s effect into perspective, it is helpful to compare GenAI forecasts to those of other technologies, including those also currently being forecasted and those that have been measured.

2010s AI economic impact forecasts outscales that of Generative AI

Before the release of GPT-3 and ChatGPT, there was a slate of forecasts on artificial intelligence that considered more traditional use cases of AI. Comparing these “pre-GPT” estimates of AI’s impact to the new GenAI forecasts may provide the clearest picture of how big of an effect the new Generative AI technology could have over the existing technological state. 

Figures 4a and 4b: Forecasts of AI on Global GDP and Productivity [source1 source2 source3 source4]. Note: McKinsey and PwC forecasts are for impact by 2030, Analysis Group forecast is for 2026, Accenture forecast is by 2025.

Interestingly, we see that AI forecasts of the 2010s are much higher than the current Generative AI impact predictions. In terms of GDP, GenAI falls in line with Analysis Group’s forecast and is at the same level of magnitude as the other forecasts. For productivity impact, GenAI does not come close to the AI forecasts, appearing indifferentiable from 0% when plotted on the same scale as those estimates. This hints to us that the forecasts for GenAI’s value are predicated on the cost savings potential of the technology rather than the productivity-boosting channel.

Other technology forecasts are more comparable with GenAI’s projections

There are plenty of other useful benchmark technologies with economic impact forecasts such as the Internet of Things (IoT), cloud technology, and autonomous vehicles. Like Generative AI, many of these technologies are very much still in their early stages of use but are projected to have big impacts. GenAI’s GDP impact appears to be right in the middle of a range of current-day technological advancements, with the mean of the GenAI forecasts landing squarely between advanced robotics and next-generation genomics. Though the range of many of these forecasts is wide, it seems that if forecasters are guilty of being over-optimistic about GenAI’s potential, they are equally guilty of such optimism with other technologies.

Figure 5: Forecasts of Technologies and GenAI on Global GDP and Productivity [source1 source2]. Forecasts were produced in 2013 and are for economic impact by 2025.

Generative AI also falls in the middle of historical technologies’ impacts

Now that we have compared forecasts of Generative AI's economic impact to other modern technologies, there is one more way we can place GenAI into scale - looking at the actual, measured impact of past technologies. Although it is difficult to ascertain the specific impact of a specific technology over time, several research papers have used statistical methods to glean individual effects. These include such cornerstone technologies as the steam engine and factory robotics. Putting GenAI forecasts side-by-side with the realized value of past technologies serves as a further benchmark for how aligned with reality such forecasts may be.

Figure 6: Measured Impacts of Historical Technologies and Mean Forecast Range of GenAI’s Impact on Productivity [source1 source2 source3].

Here again, we see that the GenAI forecasts are well in the range of the measured impacts of 20th-century and early 21st-century technologies. On the high end of the forecasts, GenAI may provide more than double the productivity boost of the Information and Communication Technologies (ICT) and Information Technology (IT) advancements of the early 2000s. Conversely, the low end of the productivity forecasts places GenAI’s impact just below these technologies.


If there is only one takeaway from the comparison of Generative AI and other technology forecasts, it is that there remains great uncertainty around GenAI’s impact. Forecasts range from the scenario that GenAI may transform the global economy, adding annual value equivalent to the world’s sixth-largest GDP, all the way to the scenario that it is just a blip on the radar compared to previously existing AI technology. 


Note: It remains very, very early in the lifecycle of Generative AI adoption and innovation. Any forecast, no matter how good the data or methodology, requires certain assumptions about the state of the future and the development of a technology. Thus all forecasts, and particularly those covering such a recent development as Generative AI, should be considered with a healthy dose of skepticism.

For further information on the potential impact of Generative AI and artificial intelligence technology more generally, see the following papers: https://www.nber.org/system/files/working_papers/w31161/w31161.pdf


Data visualizations in this post were made in R using the tidyverse, readxl, ggthemes, and RColorBrewer packages by Tal Roded. Data was collected and cleaned manually by Tal Roded.

‍Feedback provided by Neil Thompson and Peter Slattery