AI, Productivity and the Solow Growth Channel: Why Higher A Matters Now

In the Solow growth framework, artificial intelligence enters the economy most directly through technological progress: the (A) in the aggregate production function. The standard Solow production function is:

Where:

• (Y) = total output (GDP)

• (A) = total factor productivity (technology)

• (K) = capital stock

• (L) = labor input

• (\alpha) = capital’s share of income

The defining insight of the Solow model is that long-run increases in living standards cannot come solely from adding more labor or more capital. Because of diminishing marginal returns, sustained growth in output per worker ultimately depends on improvements in (A): technological progress.

This is precisely where AI matters economically. AI is not merely another software tool. It is potentially a general-purpose technology capable of raising the efficiency of both labor and capital simultaneously. In Solow terminology, AI increases total factor productivity.

The production function can be rewritten in per-worker terms by dividing through by labor:

Where:

• (y = Y/L) = output per worker (labor productivity)

• (k = K/L) = capital per worker

This equation is the critical bridge between AI and macroeconomic growth. Any sustained increase in (A) raises (y), meaning workers become capable of producing more output with the same amount of labor input.

The conceptual curve in Chart 1 illustrates the transmission mechanism from AI adoption to economic growth through the Solow framework. At low levels of adoption, AI functions primarily as a worker-assistance technology: summarization, drafting, coding assistance and research acceleration. In this phase, gains are incremental and largely task-based. Productivity improvements exist, but they are fragmented and difficult to detect in aggregate statistics.

As adoption deepens, firms begin redesigning workflows around AI. This stage introduces capital deepening, process automation and stronger human-machine complementarity. In the Solow model, this represents an increase in both effective capital intensity and total factor productivity.

The mathematical transmission mechanism can be expressed as:

In words:

1. AI raises technological efficiency ((A))

2. Higher (A) increases output per worker (Y/L)

3. Higher output per worker raises aggregate GDP (Y)

This transmission is particularly important in an economy where labor-force growth is slowing structurally.

The growth accounting decomposition further clarifies AI’s macroeconomic role:

Where:

• (g_Y) = GDP growth

• (g_A) = growth in total factor productivity

• (g_K) = growth in capital input

• (g_L) = growth in labor input

This equation demonstrates why AI is increasingly central to economic growth. If labor-force expansion slows—as demographic projections suggest—then future GDP growth must come disproportionately from productivity growth (g_A) rather than employment growth (g_L).

The projected acceleration in productivity growth from approximately 1.4% historically to 2.1% through 2030 is therefore economically significant. In a mature economy, a persistent increase in productivity growth changes the entire trajectory of real GDP, corporate profitability, wages and fiscal capacity.

Historically, major technological revolutions follow a recognizable pattern. Early investment appears excessive relative to near-term revenue generation, yet over time the infrastructure enables widespread productivity gains. The internet cycle followed this exact sequence. During the late 1990s, firms overbuilt telecommunications and digital infrastructure well ahead of realized demand. After the dot-com collapse, however, the infrastructure remained in place and enabled decades of productivity-enhancing digital expansion.

AI appears to be following a similar trajectory. Hyperscalers are deploying unprecedented levels of capital expenditure into data centers, semiconductors and computing infrastructure. Bloomberg estimates approximately $3.7 trillion of AI-related infrastructure spending over the next five years. Much like the railroad expansion of the nineteenth century or the internet buildout of the 1990s, this investment may initially appear speculative. Yet if the infrastructure accelerates diffusion of AI across the economy, it could generate a long-duration productivity cycle.

Importantly, the current macroeconomic contribution of AI is already material. Information-processing equipment and intellectual-property investment recently accounted for nearly half of annual U.S. GDP growth. That means AI investment is already affecting output through capital accumulation even before the full productivity benefits of adoption have materialized.

The occupational evidence reinforces the distinction between technological capability and realized productivity. The chart comparing theoretical AI capability with observed occupational usage reveals a large adoption gap. Occupations such as management, business and finance, computer and mathematics, legal services and education exhibit high theoretical exposure to AI, but observed utilization remains substantially below potential.

Economically, this gap represents unrealized total factor productivity. The technology exists, but firms have not yet fully reorganized production around it. Historically, this lag is normal. Productivity gains from general-purpose technologies often emerge years after the initial invention because organizations require complementary investments in software, training, managerial systems and workflow redesign before productivity becomes measurable.

This dynamic is especially visible in Europe.

The European evidence demonstrates that AI adoption raises labor productivity by roughly 4% on average, but the gains are highly uneven across firms. Medium and large firms experience substantially stronger productivity effects than smaller firms.

This pattern is economically intuitive. AI productivity gains are magnified by complementary intangible capital:

A = 𝒇(AI, software, data, training, management) 

AI alone does not automatically raise productivity. Firms must combine it with workforce training, software integration, organizational redesign and data infrastructure. The CEPR evidence shows that additional investment in software and workforce training dramatically amplifies AI’s productivity effects, especially for larger firms capable of absorbing integration costs.

This is classic Solow economics. Technology diffusion increases aggregate productivity only when firms successfully integrate innovation into the production process.

The worker-level evidence in the United States provides a microeconomic foundation for these macroeconomic effects. Occupations with higher AI usage rates also report larger time savings. Computer and math workers, managers and business professionals exhibit particularly meaningful gains.

Conceptually, labor productivity can be represented as:

Labor Productivity = Output / Hours Worked

If AI enables workers to produce the same quantity of output in fewer hours, labor productivity rises mechanically. Federal Reserve Bank of St. Louis estimates suggest workers using generative AI save approximately 5.4% of working time on average. Across the economy, this translates into an estimated aggregate productivity increase of approximately 1.1%.

The implication is important: AI does not necessarily need to replace workers to raise GDP. Productivity gains emerge even if employment remains stable, provided workers become more efficient.

Industry-level evidence shows a similar pattern. Information services, professional services and finance exhibit the highest AI utilization and largest time savings, while lower-digital-intensity sectors lag behind.

This suggests that the first major AI productivity wave is concentrated in knowledge-intensive, office-using industries. These sectors possess several characteristics favorable to AI integration:

• high information density

• repeatable cognitive tasks

• scalable digital workflows

• strong complementarities between labor and software

As a result, AI is likely to reshape the composition of work more than aggregate employment levels themselves.

This distinction matters because much public discussion around AI focuses excessively on job destruction. Historically, technological revolutions have tended to displace tasks rather than eliminate labor altogether. The internet displaced some occupations but created entirely new sectors and forms of work. Roughly 20% of current jobs did not exist in 1999.

The same dynamic may emerge with AI. Research suggests occupations facing higher automation risk often also possess higher adaptive capacity. Workers in these occupations frequently have transferable skills that allow them to shift toward higher-value activities as repetitive tasks become automated.

The demographic backdrop strengthens the macroeconomic case for AI-driven productivity growth. Long-run GDP growth can be decomposed approximately as:

If labor-force growth slows because of demographic aging and weaker working-age population expansion, then productivity growth becomes increasingly important.

CBRE projects annual employment growth of only 0.5% through the end of the decade. Under those conditions, maintaining 2%+ GDP growth requires a substantial increase in productivity growth. AI therefore arrives at a critical economic moment. It potentially allows a stagnant labor force to generate stronger output growth through higher efficiency.

This is the essence of the Solow interpretation of AI. The economic importance of AI does not primarily depend on whether it replaces workers. Its importance depends on whether it persistently raises (A), increases output per worker and expands potential GDP.

The evidence today suggests the process is already underway, though unevenly. Capital expenditure is accelerating, workers are reporting measurable time savings, larger firms are realizing stronger productivity gains and high-exposure occupations are beginning to reorganize around AI-enabled workflows. Yet the most important phase likely remains ahead.

Historically, productivity revolutions become visible only after firms fully integrate the technology into the production function itself. The decisive transition occurs when AI moves from being an individual productivity tool to becoming embedded within organizational systems, capital allocation and economic structure.

That is the point at which technological progress (A) rises persistently enough to alter the trajectory of output per worker and long-run economic growth. In the Solow framework, that is when AI ceases to be merely a technological innovation and becomes a macroeconomic growth regime.

Sources & References

Anthropic. (2026). Labor market impacts of AI: A new measure and early evidence. https://www.anthropic.com/research/labor-market-impacts 

Bick, A., Blandin, A., Deming, D.J., 2025; The Rapid Adoption of Generative AI, Federal Reserve Bank of St. Louis Working Paper 2024-027. https://doi.org/10.20955/wp.2024.027 

CBRE. (2026). AI’s Impact on the Economy, Employment & Productivity. https://www.cbre.com/insights/reports/ais-impact-on-the-economy-employment-and-productivity 

CEPR. (2026). How AI is affecting productivity and jobs in Europe. https://cepr.org/voxeu/columns/how-ai-affecting-productivity-and-jobs-europe 

Federal Reserve Bank of St. Louis. The Impact of Generative AI on Work Productivity. https://www.stlouisfed.org/on-the-economy/2025/feb/impact-generative-ai-work-productivity