Forecasts
| Date | Author | Annual Excess Growth | Labor Share | Employment | Total Compute | Consumer Surplus |
|---|---|---|---|---|---|---|
| March 2023 | Briggs and Kodnani (2023) (Goldman Sachs) | +1.5% | (not discussed) | 300M full-time jobs globally exposed; 7% US employment substituted, 63% complemented, 30% unaffected under baseline | US AI investment may approach 1% of GDP by 2030; training compute doubling every 6 months | (not discussed) |
| May 2023 | Baily, Brynjolfsson, and Korinek (2023) | +2.8% | Three forces impacting distribution: shift away from wages towards capital, increased returns to valued skills, and foreign competition effects | Up to 49% of workforce could have half+ of tasks performed by AI, but historically job destruction offset by creation | Compute used to train cutting-edge AI doubling every 6 months | (not discussed) |
| June 2023 | McKinsey Global Institute (2023) | +0.1-0.6% | (not discussed) | (not discussed) | (not discussed) | (not discussed) |
| March 2024 | Korinek and Suh (2024) | +18% | Wages collapse by ~85% once automation reaches critical thresholds in baseline AGI scenario | Full employment maintained but wages fall dramatically rather than unemployment rising | Compute is key metric for task complexity; frontier AI compute doubling every ~6 months vs Moore’s Law ~2 years | (not discussed) |
| April 2024 | Acemoglu (2024) | +0.07% | AI predicted to widen gap between capital and labor income | Minimal employment disruption by 2030; only 4.6% of all tasks impacted within 10 years | (not discussed) | Discusses TFP for welfare; notes potential “negative social value” AI tasks |
| May 2024 | Aldasoro et al. (2024) (BIS) | +2.5% | (not discussed) | Three competing forces: productivity increases demand, new task creation (both positive), vs worker displacement (negative); net effect uncertain | Computational resources doubling every 6 months; larger models need more data | (not discussed) |
| June 2024 | Aghion and Bunel (2024) | +0.68-1.3% | (not discussed) | Automation has positive effect on labor demand at firm and industry level based on French firm data | (not discussed) | (not discussed) |
| July 2024 | Tytell (2024) | +0.2-0.9% | (not discussed) | (not discussed) | (not discussed) | (not discussed) |
| Dec 2024 | Filippucci, Gal, and Schief (2024) (OECD) | +0.25-0.6% | AI could substitute for high-skilled labor and narrow wage gaps (ambiguous impacts) | Potential labor market disruptions noted but no specific unemployment predictions | (not discussed) | (not discussed) |
| March 2025 | Bergeaud et al. (2025) (ECB) | +0.29% | Uses labor share in methodology but no predictions about changes | Most firms don’t intend to reduce headcount; 5% of jobs face high automation risk, 13%+ high augmentation potential | €33B invested in EU AI companies vs €120B+ in US | (not discussed) |
| March 2025 | Erdil et al. (2025) (Epoch) | +30% | (not discussed) | (not discussed) | Investment in global compute supply may exceed 10% of world GDP, ~50-fold increase; optimal 2025 investment ~$25 trillion | (not discussed) |
| April 2025 | Misch and Zhang (2025) (IMF) | +0.2% | (not discussed) | (not discussed) | (not discussed) | (not discussed) |
| June 2025 | Filippucci et al. (2025) (OECD) | +0.4-1.3% | (not discussed) | (not discussed) | (not discussed) | (not discussed) |
| Sept 2025 | Arnon (2025) (Penn Wharton) | +0.15% | Assumes exposed GDP share equals labor income share; 40% of labor income faces automation exposure | Employment in completely automatable jobs fell 0.75% since 2021; growth in highly exposed occupations slowed since 2022 | (not discussed) | (not discussed) |
Detailed Paper Summaries
Briggs & Kodnani (2023) - Goldman Sachs
Main Findings: Generative AI could raise global GDP by 7% and US labor productivity growth by 1.5pp/year over a 10-year period following widespread adoption.
Extended Quotes
“Generative AI could raise annual US labor productivity growth by just under 1½ percentage points over a 10-year period following widespread business adoption.”
“Generative AI could eventually increase annual global GDP by 7 percent, equal to an almost $7 trillion increase in annual global GDP over a 10-year period.”
“Widespread adoption of generative AI could raise overall labor productivity growth by around 1.5pp/year, roughly the same-sized boost that followed the emergence of prior transformative technologies like the electric motor and personal computer.”
“Roughly two-thirds of current jobs are exposed to some degree of AI automation, and generative AI could substitute up to one-fourth of current work.”
“If AI-enabled task automation raises the level of aggregate productivity by 15% over roughly 10 years, generative AI would unlock around $4½ trillion in annual value for the US economy (in 2024 dollars).”
Key Assumptions
- Examined 13 of 39 work activities across 900+ occupations using O*NET database
- AI capable of completing tasks up to difficulty level 4 on 7-point scale
- Jobs with ≥50% exposed tasks likely substituted; 10-49% exposure likely complemented
- Excludes outdoor/physical labor jobs (no robotics considered)
- Historical pattern: productivity booms start ~20 years after breakthrough, when ~50% of businesses adopt
Baily, Brynjolfsson & Korinek (2023)
Main Findings: In Level+Growth scenario, AI adds 18% productivity boost over 10 years plus 1pp sustained growth rate increase, leading to near-doubling of output over 20 years.
Extended Quotes
“A recent report by Goldman Sachs suggests that generative AI could raise global GDP by 7%, a truly significant effect for any single technology.”
“For instance, if generative AI makes cognitive workers on average 30% more productive and cognitive work makes up about 60% of all value added in the economy (as measured by the wage bill attributable to cognitive tasks), this amounts to a 18% increase in aggregate productivity and output.”
“The baseline follows the current projection of the Congressional Budget Office (CBO) of 1.5% productivity growth, giving rise to a total of 33% productivity growth over 20 years.”
“The third projection labeled ‘Level+Growth’ additionally includes a one percentage point boost in the rate of growth over the baseline rate, resulting from the additional innovation triggered by generative AI. Through the power of compounding, the effects grow bigger over time, leading to a near doubling of output after 20 years, far greater than the baseline projection.”
“The second channel is the acceleration of innovation and thus future productivity growth… If cognitive workers are more efficient, they will accelerate technological progress and thereby boost the rate of productivity growth—in perpetuity.”
Key Assumptions
- Cognitive worker productivity gains: 30% average improvement
- Cognitive work share: ~60% of economy (by wage bill)
- Two channels: Level effect (18% boost over 10 years) + Growth effect (1pp sustained increase)
- Timeline: 10 years for gains to materialize, compounding over 20 years
Korinek & Suh (2024)
Main Findings: In baseline AGI scenario with full automation within 20 years, growth increases ~10x to 18% annually, but wages collapse by ~85%.
Extended Quotes
“In their simulations, reaching full AI automation within 20 years boosts growth roughly tenfold to about 18 percent a year.”
“In a faster five-year scenario, output doubles every four years.”
“Once the automation index surpasses the threshold, the economy enters a second region, where the scarcity of labor is alleviated…wages decline starkly to equal the marginal product of capital.”
“The takeoff in output and the collapse in wages in the two AGI scenarios are both driven by the same force: the substitution of scarce labor by comparatively more abundant machines.”
“Compute in frontier AI systems has doubled roughly every six months over the past decade.”
Key Assumptions
- Task-based framework with tasks varying in computational complexity
- Three scenarios: Business-as-usual (unbounded tasks), Baseline AGI (20 years to full automation), Aggressive AGI (5 years)
- Automation index I growing exponentially; tasks automatable if complexity < I
- CES production function (σ<1) aggregating tasks
- Full employment maintained but wages collapse rather than unemployment rising
Acemoglu (2024)
Main Findings: TFP effects within 10 years should be no more than 0.71% total (0.07% annually), potentially lower at 0.53% when accounting for hard-to-learn tasks.
Extended Quotes
“This paper evaluates claims about the large macroeconomic implications of new advances in AI… Using existing estimates on exposure to AI and productivity improvements at the task level, these macroeconomic effects appear nontrivial but modest—no more than a 0.71% increase in total factor productivity over 10 years.”
“This calculation implies that total factor productivity (TFP) effects within the next 10 years should be no more than 0.71% in total—or approximately a 0.07% increase in TFP growth annually.”
“The paper then argues that even these estimates could be exaggerated, because early evidence is from easy-to-learn tasks, whereas some of the future effects will come from hard-to-learn tasks, where there are many context-dependent factors affecting decision-making. Consequently, predicted TFP gains over the next 10 years are even more modest and are predicted to be less than 0.53%.”
“I don’t think we should belittle 0.5 percent in 10 years. That’s better than zero…But it’s just disappointing relative to the promises that people in the industry and in tech journalism are making.”
Key Assumptions
- Task-based model applying Hulten’s theorem
- 20% of US tasks exposed to AI (Eloundou et al. 2023)
- 23% of exposed tasks profitably automated within 10 years (Svanberg et al. 2024)
- Final calculation: 0.23 × 0.20 = 4.6% of all tasks impacted
- Average labor cost savings: 27% per automated task
- Differentiates easy-to-learn vs hard-to-learn tasks
BIS (2024) - Aldasoro et al.
Main Findings: AI raises annual productivity growth by 1.5pp for a decade, resulting in GDP almost 30% higher after 10 years.
Extended Quotes
“we assume that AI raises annual productivity growth by 1.5 percentage points for the next decade, in line with plausible estimates in the literature … Growth is fastest in the first 10 years – i.e. the period in which AI directly raises industry-level TFP – at which point GDP is almost 30% higher than it would have been without”
“Estimates range from 0.5 to 1.5 percentage points over the next decade”
“AI, by spurring productivity growth, significantly raises aggregate output, consumption and investment in both the short and long run”
“Output grows by twice as much for the same increase in aggregate productivity when AI affects sectors producing consumption rather than investment goods, thanks to second round effects through sectoral linkages.”
“computational resources employed by AI systems…has been doubling every six months”
Key Assumptions
- Multi-sector general equilibrium model with input-output linkages
- AI modeled as permanent productivity increase differing by sector
- 1.5pp productivity growth assumption based on Baily et al. (2023) and Goldman Sachs (2023)
- Inflation effects depend on whether productivity gains are anticipated
- Historical technology patterns from electricity and smartphones inform adoption curves
Aghion & Bunel (2024)
Main Findings: AI should increase aggregate productivity growth by 0.68-1.3pp per year over the next decade, using two complementary approaches.
Extended Quotes
“Based on the first approach, we estimate that the AI revolution should increase aggregate productivity growth by between 0.8 and 1.3pp per year over the next decade.”
“If productivity gains from AI are comparable to those of the electricity wave of the 1920s in Europe, then productivity growth would increase by 1.3 percentage points per year starting in 2024.”
“If using the digital technology wave of the late 1990s and early 2000s in the United States as a point of comparison, the increase in productivity growth would be around 0.8 percentage points per year.”
“Using their own reading of the recent empirical literature regarding each component [of the task-based formula], Aghion and Bunel obtain a median estimate of 0.68pp additional annual total factor productivity (TFP) growth.”
“The productivity of employees who had access to an AI assistant increased by 14% in the first month of use and stabilized at approximately 25% after three months.”
Key Assumptions
- Two approaches: Historical comparison (0.8-1.3pp) and task-based framework (0.68pp median)
- Electricity revolution (1920s Europe) as upper bound; digital wave (1990s-2000s US) as lower bound
- Task-based approach revises Acemoglu’s parameters based on different empirical literature reading
- Long-run multiplier of ~1.5 for converting TFP gains to labor productivity gains
- Implementation lags: productivity from electricity took ~30 years to materialize
Filippucci, Gal & Schief (2024) - OECD
Main Findings: AI could contribute 0.25-0.6pp to annual TFP growth (0.4-0.9pp to labor productivity) in the US over the next decade.
Extended Quotes
“AI could contribute between 0.25 and 0.6 percentage points to annual total factor productivity growth in the US over the next decade.”
“The results suggest that AI could contribute significantly to aggregate productivity growth over the next decade, contributing between 0.25 to 0.6 percentage points to annual Total Factor Productivity (TFP) growth in the United States (or 0.4 to 0.9 percentage points to annual labour productivity growth) in our main scenarios.”
“These estimates imply a substantial improvement in the context of the weak productivity growth across the OECD over the past decades, which has been in the range of 1-1.5% per year.”
“The aggregate productivity gain from AI is the sum of three effects: 1) a direct effect of increasing productivity at the sectoral level; 2) an input-output multiplier effect as productivity gains in one sector also benefit other sectors through reduced costs of intermediate inputs; and 3) a negative reallocation effect in the spirit of Baumol’s growth disease that arises if the sectors with limited productivity growth increase as a share of GDP.”
Key Assumptions
- Novel micro-to-macro framework combining micro-level performance gains, AI exposure, and adoption rates
- Assumes 30% cost savings from AI at task level
- Multi-sector general equilibrium model with input-output linkages
- Five scenarios with varying elasticity and reallocation assumptions
- Current adoption <5% of US firms (as of early 2024)
Bergeaud (2024) - ECB
Main Findings: Euro area could see TFP boost of 0.3pp per year over next ten years (2.9pp cumulative after adjustments).
Extended Quotes
“For the euro area, this would reduce the productivity gains to around 3.1 pp over 10 years if either one is considered, and 2.9 pp if both are”
“The baseline estimate suggests a productivity increase of around 0.35 pp per year on average for the euro area, which equals 3.5 pp over ten years for the whole economy due to AI”
“One widely accepted methodology estimates that the euro area could see a boost to total factor productivity (TFP) of around 0.3 percentage points per year over the next ten years”
“Compare that with the past decade, when annual TFP growth averaged just 0.5%”
“Most firms state that they do not intend to reduce their headcount through the adoption of AI”
Key Assumptions
- Uses Acemoglu (2024) task-based framework
- Analyzed 16,937 tasks across 220 professions from O*NET database using GPT-4
- Adjustments for: IMF AI Preparedness Index and UN Productive Capacities Index
- Cross-country results: 1.5pp (Ireland) to 3.3pp (Belgium) over 10 years
- Only 75% of large euro area firms use AI; <25% of employees regularly use it
Erdil et al. (2025) - Epoch GATE Model
Main Findings: Model predicts 30% annual GWP growth by 2035 under default parameters, though lead author expresses personal skepticism.
Extended Quotes
“It predicts trillion-dollar infrastructure investments, 30% annual growth, and full automation in decades.”
“even when AIs have only automated 30% of all economically useful tasks, economic growth rates may exceed 20%.”
“at the point that 40% of tasks are automated, GATE still predicts 12% GWP growth rates, on par with peak GDP growth rates observed in East Asian economies during the 20th century”
“GATE predicts that AI automation leads to significantly accelerated economic growth, with rates elevated by 2-20 times compared to the recent historical average of ~3% per year.”
“the model consistently projects that the global economy can marshal enough effective compute to automate most tasks within two decades.”
Ege Erdil’s personal view: > “i don’t personally predict 30% mean annual gdp growth in the US over the next 10 years … the model does with some reasonable parameter values, but my timelines for that kind of growth are longer, and there’s model uncertainty and so on.”
Key Assumptions
- Three core components: compute-based AI development, AI automation via scaling laws, semi-endogenous growth
- Social planner maximizing lifetime utility of representative household
- Raw compute required to train neural networks halves every 9-16 months
- Investment in global compute supply may exceed 10% of world GDP (~50-fold increase)
- Optimal 2025 AI investment: ~$25 trillion
- Model designed for “qualitative insights” not precise quantitative predictions
Misch et al. (2025) - IMF
Main Findings: Medium-term productivity gains for Europe likely modest at around 1% cumulatively over five years (0.2% annually).
Extended Quotes
“The medium-term productivity gains for Europe as a whole are likely to be modest, at around 1 percent cumulatively over five years.”
“While economically still moderate, these gains are still larger than estimates by Acemoglu (2024) for the US, and they vary widely across scenarios and countries and are substantially larger in countries with higher incomes.”
“There is substantial cross-country variation, ranging from around 0.5 percent in Romania to close to 1 percent in Luxembourg, with higher-income countries having larger gains driven by the higher prevalence of white-collar services including financial services, which tend to be more exposed to AI.”
“In the preferred scenario, the gains in Luxembourg could be 2 percent cumulatively, almost twice the European average, and more than 4 times larger than those in Romania.”
“National and EU regulations around occupation-level requirements, AI safety, and data privacy combined could reduce Europe’s productivity gains by over 30 percent if AI exposure were 50 percent lower in tasks, occupations and sectors affected by regulation.”
Key Assumptions
- Uses Acemoglu (2024) task-based framework adapted for Europe
- Three key parameters: 19.9% AI exposure, 23% adoption rate, 27% labor cost savings
- 5-year horizon (vs Acemoglu’s 10 years) because framework doesn’t capture long-term transformational effects
- Regulation modeled as halving AI capabilities (middle ground assumption)
- Multiple scenarios based on pooled evidence on automatable tasks
Filippucci, Gal, Laengle & Schief (2025) - OECD G7
Main Findings: Annual labor productivity growth of 0.4-1.3pp in high-exposure countries (US, UK), up to 50% smaller in other G7 economies.
Extended Quotes
“Annual aggregate labour productivity growth due to AI range between 0.4-1.3 percentage points in countries with high AI exposure – due to stronger specialisation in highly AI-exposed knowledge intensive services such as finance and ICT services – and more widespread adoption (e.g. United States and United Kingdom).”
“Projected gains in several other G7 economies are up to 50% smaller, reflecting differences in sectoral composition and assumptions about the relative pace of AI adoption.”
“AI could contribute between 0.25 and 0.6 percentage points to annual aggregate TFP growth in the United States over the next decade.” [0.4-0.9pp for labor productivity]
“These estimates imply a substantial improvement in the context of the weak productivity growth across the OECD over the past decades, which has been in the range of 1-1.5% per year.”
Key Assumptions
- Extends methodology from 2024 “Miracle or Myth” paper to all G7 economies
- Micro-to-macro framework with input-output linkages
- AI exposure defined as proportion of sector tasks AI can enhance
- Three scenarios: baseline, optimistic, pessimistic
- Current adoption limited: <5% of US firms
- Accounts for general equilibrium effects including negative reallocation (Baumol disease)
Arnon (2025) - Penn Wharton Budget Model
Main Findings: AI will increase productivity and GDP by 1.5% by 2035 (0.15% annually), with peak annual contribution of 0.2pp in 2032.
Extended Quotes
“We estimate that AI will increase productivity and GDP by 1.5% by 2035, nearly 3% by 2055, and 3.7% by 2075.”
“We estimate that 40 percent of current GDP could be substantially affected by generative AI. Occupations around the 80th percentile of earnings are the most exposed.”
“AI’s impact on total factor productivity depends on two factors: the share of economic activity impacted by AI tools and the cost savings from adopting AI tools.”
“AI’s boost to productivity growth is strongest in the early 2030s, with a peak annual contribution of 0.2 percentage points in 2032.”
“Considerable caution is required when interpreting these projections. Currently, our analysis does not account for AI-driven changes in product quality or the emergence of new products.”
Key Assumptions
- Jobs “exposed” if ≥50% of tasks automatable; 42% of jobs meet this threshold
- Only 23% of exposed tasks eventually automated profitably (Svanberg et al. 2024)
- Labor cost savings: ~25% currently, growing to 40% over coming decades
- ~10% of current GDP likely impacted, potentially 15% within two decades
- Adoption follows historical patterns: 40-50% of workers using by end of first decade
- Excludes product quality changes, new products, new task creation
Bibliography
Citation
@online{cunningham2025,
author = {Cunningham, Tom},
title = {Forecasts of {AI} \& {Economic} {Growth}},
date = {2025-11-06},
url = {tecunningham.github.io/posts/2025-10-19-forecasts-of-AI-growth-extended.html},
langid = {en}
}