AI spending initially depresses firms' revenues but delivers long-term growth once experience accumulates; coupling AI with R&D strongly amplifies revenue gains.

Main Finding

AI investment exhibits a U-shaped effect on firm revenue growth across 26 European countries (2012–2023): early-stage AI adoption tends to reduce revenue growth (likely due to upfront costs and implementation inefficiencies), but with experience and time the effect becomes positive and sizable. Crucially, AI yields substantially larger revenue gains when combined with R&D/innovation activities—standalone AI spending is generally insufficient for sustained entrepreneurial success.

Key Points

• U-shaped relationship between AI investment and revenue growth:
  • Short-term negative or dampening effect after initial AI investment.
  • Long-term positive returns as firms accumulate experience and complementary capabilities.
• Complementarity with innovation/R&D:
  • Firms that integrate AI investments with R&D/innovation strategies realize much stronger revenue growth.
  • Highlights that strategic alignment and complementary assets (skills, processes, products) matter.
• Robustness and causal concerns:
  • Results hold under alternative AI proxies.
  • Instrumental-variable (2SLS) regressions and robustness tests were used to mitigate endogeneity concerns.
• Theoretical framing:
  • Interpreted through resource-based theory: AI is a resource whose value depends on firm-specific capabilities and complementarities.

Data & Methods

• Data:
  • Unbalanced panel of 1,479 firms across 26 European countries.
  • Time span: 2012–2023.
  • Outcome: firm revenue growth; key regressor: firm-level AI investment proxy (and alternative proxies for robustness).
• Estimation:
  • Primary estimation technique: Cameron et al. (2011) multi-way clustering (CGM) to account for heteroscedasticity and cross-sectional dependence.
  • Specification tests and robustness checks include alternative AI measures.
  • Instrumental-variable approach (2SLS) applied to address potential endogeneity between AI investment and revenue growth.
• Interpretation:
  • Identification strategy leverages panel structure and IVs to separate short-run disruption effects from long-run gains, while controlling for observed/unobserved confounders to the extent possible.

Implications for AI Economics

• For firm strategy:
  • Expect non-linear returns: managers should plan for an initial investment/learning phase and not judge AI investments by short-run revenue outcomes alone.
  • Invest jointly in AI and complementary capabilities (R&D, human capital, process redesign) to unlock scalable revenue benefits.
  • Resource allocation should include budgets for change management, training, and experimentation to shorten the negative initial phase.
• For policymakers:
  • Short-run disruption suggests a role for transitional support (training subsidies, implementation grants, tax credits) to help firms—especially SMEs—overcome upfront costs and capability gaps.
  • R&D and AI policy should be coordinated (e.g., AI-linked R&D incentives) to promote complementarities that accelerate productive adoption.
  • Evaluation of AI policy impact should use medium-to-long horizons rather than only short-term firm outcomes.
• For researchers:
  • Empirical work should account for dynamic, non-linear treatment effects of AI and for complementarities with organizational capabilities.
  • Further research directions: heterogeneity by industry, firm size, and country institutional context; mechanisms (labor reallocation, process automation, product innovation); measurement improvements for AI adoption and experience.
• Broader economic implications:
  • Aggregate productivity and entrepreneurship gains from AI will depend on the pace at which firms build required complementary assets; uneven capability diffusion could create short-to-medium-term disparities in firm performance and market structure.