The AI Question Even the Experts Can't Answer | The Spillover

Video thumbnail: The AI Question Even the Experts Can't Answer | The Spillover
Aug 18, 202644m 43s video lengthCouncil on Foreign Relations

The Signal

Predicting an economic winner in the AI race is a fool’s errand, as the outcome depends on a multi-input contest rather than a single metric. While the U.S. currently leads in private capital expenditure and development revenue, China shows surprising strength in research publications, power infrastructure, and deployment, suggesting the U.S. lead is narrower and more contingent than often assumed.

The Case

Infrastructure and Capital

  • The U.S. maintains a clear lead in private capex with an estimated $700–800 billion spend, compared to China’s $150 billion in private investment plus $60 billion annually from the government.6:43
  • China may effectively narrow this gap through distillation, a technique where actors query and imitate frontier models to train at a fraction of the cost, making raw capital spend a misleading proxy for actual output.7:09
  • Infrastructure gives China a structural edge, as the nation produces twice as much electricity as the U.S. with a more diversified renewable buildout, providing a more reliable foundation for energy-hungry data centers.9:35

Talent and Deployment

  • Research leadership is contested; while the U.S. excels in revenue, China produced over 35% of top-tier AI research publications by 2022 and currently hosts a significant concentration of global talent.3:51
  • Deployment is a distinct bottleneck; the U.S. ranks only 24th in the Microsoft AI diffusion report, suggesting that domestic model-building dominance has not yet translated into broad-based economic usage.14:47
  • European adoption remains sluggish, with 60% of surveyed firms reporting no or infrequent AI use, reinforcing the idea that advanced economies are not automatically leading the adoption curve.15:24

The Organization Gap

  • Individual-level productivity—such as a coder doubling their output—rarely translates to full organizational gains, as internal approval layers and workflow bottlenecks often prevent meaningful speedups.22:23
  • Real-world success likely depends on radical structural shifts within firms rather than simply adding AI tools, with different regions showing distinct norms: the U.S. accepts higher job volatility to chase efficiency, while Europe prefers preserving existing labor relationships.19:26

The 1 Minute Signal Take

The race is less a competition of model quality and more a test of which country can integrate AI into its labor and energy workflows while managing public anxiety. Because deployment and infrastructure lag behind development, the U.S. advantage in frontier model revenue could be eroded by cheaper, distillated alternatives from international rivals.

Pro Analysis

Why It Matters

This analysis moves the conversation from speculative 'AI winner' scenarios to a granular focus on geopolitical inputs. It identifies that the economic value of AI is not solely derived from frontier model performance, but from the messy, organizational-level reality of adoption and the strategic control of power infrastructure.

Strategic Implications

The reliance on 'chokepoints' like ASML or TSMC as permanent leverage is likely a losing strategy. As other nations move to build workarounds, the relative power of current gatekeepers will diminish. For firms, the lesson is that AI integration requires a complete redesign of internal 'sign-off' processes to unlock the productivity gains currently trapped by bureaucratic speed limits.

Evidence & Hype Audit

The content relies heavily on NBER working papers and Microsoft diffusion data, which are credible. However, it explicitly flags some of its own projections (like the 2028 talent estimate) as speculative. It remains refreshingly honest about the 'epistemic humility' required to analyze a field moving this rapidly.

Counterarguments

One could argue that the 'distillation' threat is overstated if frontier models continue to improve in capability at a pace that keeps them perpetually ahead of the best distillable versions. Additionally, the focus on adoption rankings may ignore the depth and strategic importance of the use cases being deployed in the U.S. compared to other regions.

What To Do Next

  • Conduct a 'bottleneck audit' within your firm: identify which approval layers are negating AI-driven speedups.
  • Monitor energy price trends, as they will dictate the future cost-benefit of local data center expansion.
  • Evaluate your reliance on singular 'chokepoint' suppliers and prepare for alternative procurement routes.
  • Shift focus from 'generative capabilities' to 'agentic integration' to see where real ROI is being captured.
  • Track cross-country adoption metrics as a leading indicator of regional economic health.
Time saved:41m 14s

Share this

Tags

Written by: 1 Minute Signal Editorial Team