Council of Economic Advisers Frames AI as Defining Force
Commentary by Vijay Dewan
In its Annual Report to Congress, the Council of Economic Advisers ("CEA") provided a sweeping analysis of the implications and economic trajectory of artificial intelligence and the United States' competitive position relative to China and Europe.
In Chapter 5, The Revolution of Artificial Intelligence, the CEA concluded:
- The range of impact on GDP is enormous, but the floor is rising. The CEA said that GDP impact estimates from recent studies ranged from roughly 1 percent to 45 percent over 10 years. They note the low end is already being ruled out: AI-related investment alone added an annualized 1.3% to GDP in the first half of 2025, comparable to railroad investment during the Industrial Revolution. Mid-range projections from Goldman Sachs, McKinsey, and Oxford Economics cluster around a 2–7% GDP lift over the next decade, with more aggressive models pointing significantly higher.
- The investment surge is unlike anything in the tech era. U.S. private AI investment reached $109 billion in 2024, accounting for roughly 75% of global venture funding in generative AI. The CEA highlights a defining paradox: training costs are growing at 2.5x per year — Grok 4 alone cost nearly $500 million to build — while inference costs are collapsing, falling as much as 900-fold annually. That combination of capital intensity and commoditizing output mirrors the economics of railroads and electricity, historically the signature of a true platform technology.
- Adoption is early but accelerating fast. The share of U.S. firms using AI in production grew from under 4% in 2023 to roughly 10% by late 2025; paid AI subscriptions jumped from 7% to 45% of companies over the same period. The CEA points to one capability metric as the clearest leading indicator: the length of tasks AI can complete autonomously has doubled every seven months for six years, suggesting agentic AI is approaching practical deployment at scale.
- The geopolitical dimension is the wildcard. The CEA framed AI as a potential "second Great Divergence" — a structural break in relative economic performance between nations that embrace the technology and those that lag. The U.S. currently controls roughly 74% of global AI compute capacity, and almost all Chinese AI models run on American-designed hardware. The EU, by contrast, invested less than $50 billion cumulatively versus over $470 billion in the U.S. The policy response — deregulation, accelerated data center permitting, the One Big Beautiful Bill's 100% bonus depreciation for IT infrastructure, and foreign investment commitments including $40 billion in AI chips from the EU — was explicitly designed to extend that lead. CEA said the implication is straightforward: the gap between AI leaders and laggards, both at the country and sector level, is likely to widen before it narrows.
Commentary
In Chapter 5, the CEA frames AI as a general-purpose technology rather than an industry, drawing the right comparison to steam, electricity, and the Internet rather than to any single sector's productivity boom. The invocation of Jevons' Paradox is a useful corrective to the reflexive assumption that labor-saving technology must reduce employment; the historical record on steam, electricity, and computing supports the CEA's view. But the range of GDP impact estimates - from 1 to 45 percent over ten years - reflects an even broader range of assumptions. The high-end figures assume AI can substitute for most human labor which actually describes a different technology than the "narrow" AI the chapter says it is analyzing.
The CEA points out that the U.S. lead is structural rather than incidental. Cumulative private investment of roughly $470 billion against $50 billion across the EU is not a gap that industrial policy alone can close. The investment surge the chapter celebrates is also a concentration risk. The CEA cites railroad investment as a favorable comparison, but the nineteenth-century rail buildout produced both enduring infrastructure and a speculative collapse, a reminder that scale of investment is not, by itself, evidence that the investment will pay off.
The productivity case made by the CEA, meanwhile, remains unproven. The CEA says the leading indicators tell us that capital is being deployed and benchmark scores are rising. They do not yet tell us whether deployed capital is producing economic output to match its cost. The CEA is right that we should not expect productivity gains to show up in the data yet, but that means neither the optimistic nor the pessimistic case can currently be proved. The honest position is that the macroeconomic verdict is still some years away.