Deutsche Bank: Global AI Capex Benefits the Dollar in the Long Run but Worsens Growth Divergence

Overview

Authored by George Saravelos, Global Head of FX Research at Deutsche Bank, this report provides an in-depth analysis of the global distribution, macroeconomic implications, and long-term societal consequences of the artificial intelligence (AI) capital expenditure cycle. It observes that while the AI boom originated in the U.S., its capex footprint is highly globalized, with substantial investment flowing into North Asian supply chains. In the short term, AI-driven demand is fueling inflation without significantly disrupting labor markets; over the longer horizon, productivity gains will accrue predominantly to users (wealthy nations), while ownership returns concentrate in the U.S. and China. The report warns that if AI fully replaces—rather than augments—labor, it risks triggering severe inequality and economic paradigm collapse, making national capacity to manage political-economic transitions decisive for relative success.

Core views

Global Flow and Allocation of AI Capital Expenditure The report stresses that the AI capex cycle is not confined to the U.S. In a hypothetical $100 billion AI capex scenario, ~60% 'leaks' offshore—with North Asia as the principal beneficiary. Specifically, Taiwan (TSMC) and South Korea (Samsung, SK Hynix) capture productive capital stock, while the U.S. captures the profit pool via high gross margins (e.g., NVIDIA’s >70%). Despite outward capital flows, ~$8 billion in offshore funds ultimately recirculates to the U.S. via foreign direct investment, dividend repatriation, and current account surplus recycling. Overall, the U.S. captures ~60% of total value across the cycle, with the remainder genuinely transferred to North Asia. Current State of Inflation and Labor Markets There is currently no robust evidence that AI has materially impacted U.S. or global labor markets; early studies failed to adequately control for coincident Fed tightening. However, clear demand-side inflation signals exist—including rising U.S. software CPI and import prices for capital goods. This capex-driven demand boom should lift the global neutral interest rate (r*) and pose hawkish risks for economies most exposed to AI capex—particularly North Asia and the U.S. Long-Term Implications: Growth Divergence and Social Unrest Once the capex cycle concludes, productivity dividends will accrue to AI users. Data show strong correlation between AI usage intensity and per-capita GDP: wealthy nations are intensive users, while poorer ones lag—potentially exacerbating global growth divergence, especially between high- and low-income countries. On income flows, the U.S. (alongside China) will be the principal beneficiary, expected to generate stable service export revenues equivalent to ~1% of U.S. GDP—supporting the dollar exchange rate. Yet the report invokes Marx to warn that if AI fully substitutes—not augments—labor, capitalism’s traditional labor-capital distribution mechanism collapses, leading to zero labor income, extreme demand-side traps, and social conflict. Such full automation resembles early 19th-century industrialization—but may prove more disruptive. Fiscal policy and institutional reform will then determine whether the transition proceeds orderly; nations capable of successfully managing this political-economic transformation will reap economic and monetary rewards.

Analysis framework

The report employs three analytical lenses: global fund-flow analysis (Sankey-diagram thinking), macro supply-demand frameworks, and historical comparative analysis. First, by dissecting the sources and destinations of AI capex, it quantifies value transfer and repatriation mechanisms between the U.S. and Asia—explaining why Asian FX has not appreciated sharply despite strong exports. Second, integrating inflation data and labor-market indicators, it distinguishes AI’s differential lags on demand-side (inflation) versus supply-side (employment) effects. Third, drawing on Marxist political economy, it contrasts 'augmentation' versus 'substitution' AI pathways to model macroeconomic collapse risk under extreme automation—and underscores the policy imperative for mitigation.

Methodology notes

• Industry/Industrial Analysis FrameworkUpstream-Midstream-Downstream Value Chain Transmission

  Global value chain transmission of AI capital expenditure

  The report traces how AI capex flows from U.S. hyperscalers to Asian hardware manufacturers (upstream), then back to U.S. profits (downstream), revealing unequal value distribution and fund-recycling mechanisms within the global supply chain.

• Macroeconomic framework

  Neutral interest rate (r*) and the capital expenditure cycle

  The report notes that AI-driven demand-side capex booms reduce global savings rates, thereby lifting the neutral interest rate (r*), which is critical for interpreting current and future monetary policy stances.

• Cycle and Business Conditions Framework

  Societal transition risks of technological revolutions

  Drawing analogies to historical industrialization, the report analyzes how AI’s shift from 'labor augmentation' to 'labor substitution' could trigger extreme inequality and social unrest—highlighting the decisive role of political-economic management capacity during technology cycles.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• NVIDIA (NVDA)

  Beneficiary

  Strengths

  Gross margins exceeding 70%, capturing the majority of AI capex-related profits

  Comparison

  Occupies a higher-margin segment of the value chain compared to Asian hardware manufacturers

• TSMC (Taiwan Semiconductor Manufacturing Company)

  Beneficiary

  Strengths

  Captures productive capital stock and receives substantial capex inflows

  Comparison

  Lower margins than U.S. peers but larger scale of capital accumulation

• Samsung/SK Hynix

  Beneficiary

  Strengths

  As Korean representatives, they capture part of AI capex

  Comparison

  Together with TSMC, form the core of North Asia’s productive capital base

Key data

• Offshore share of AI capex~60%Flows outside the U.S., primarily to North Asia
• U.S. share of total-cycle value capture~60%Achieved via profits, wages, and fund repatriation
• Projected impact of AI on U.S. current account~1% GDPDriven by higher service export revenues
• NVIDIA gross margin>70%Reflects U.S. dominance in the profit pool

Impact & implications

The report concludes that the AI capex cycle supports the U.S. dollar in the near term, as U.S. tech firms capture most profits and convert them into current-account surpluses. Yet Asian currencies have not appreciated sharply despite strong exports—partly due to energy crises and fund repatriation to the U.S. Over the longer horizon, AI may widen global growth divergence, with wealthier nations benefiting disproportionately through broader AI adoption. For policymakers, the key challenge lies in managing potential social disruption from automation via fiscal redistribution and institutional reform; otherwise, economic gains may be offset by inequality costs.

Risks

• Extreme inequality and social conflict arising from AI fully substituting labor
• Heightened uncertainty from intensifying U.S.-China tech competition
• Slower global AI adoption due to strategic autonomy concerns
• Uncertainty around monetization of closed-source versus open-source AI models

What to watch

• Evolving patterns of global AI adoption (driven by strategic autonomy considerations)
• Commercialization progress of closed-source versus open-source AI models
• Fiscal and institutional policies adopted by governments to manage automation-driven transitions
• Persistent impact of AI on the global neutral interest rate (r*)