Anthropic’s Chief Economist Warns AI Could Cut Labor’s Share of Income by 15 Points, Floats ‘Token Tax’
TREE NEWS reports: At a Harvard Kennedy School forum, Anthropic chief economist Peter McCrory and former Obama economic adviser Jason Furman laid out a stark framework for how artificial intelligence may reshape the macroeconomy by 2030. McCrory, who leads the AI lab’s economic research and its “Economic Index” tracking real-world Claude usage, argued that AI is not yet showing up in productivity data because of a J-curve effect — firms must make costly organizational and data investments before gains materialize — while also warning that capital will capture a sharply larger share of national income as automation deepens.
What Happened
McCrory said the range of tasks AI models can complete autonomously is doubling roughly every four to seven months, yet total factor productivity has not surged and U.S. unemployment sits near 4.1%. He attributed the gap to diffusion lags and corporate restructuring costs. Based on observed task-level time savings, he estimated AI could add about 1.8 percentage points to labor productivity growth — far above the Congressional Budget Office’s 1.7% long-run growth forecast and the Fed’s most optimistic 2.4% projection.
On distribution, McCrory was blunt: in extreme scenarios, labor’s share of income could fall by roughly 15 percentage points from its historical 60% level, meaning shareholders capture most AI-driven gains. He also floated a “token tax” — analogous to carbon or tobacco levies — as a Pigouvian tool to curb over-automation, alongside a shift toward consumption-based taxation as labor income shrinks.
Market Implications
- Equities: A higher-productivity, capital-heavy economy favors large AI infrastructure incumbents and firms with proprietary data moats. But the same dynamic implies weaker mass-consumption demand if labor income falls, a long-term risk for consumer discretionary and retail names.
- Bonds: If AI lifts potential growth toward 1990s levels, term premia and neutral rates could drift higher, pressuring long-duration bonds. Conversely, displacement-driven demand weakness would pull the other way — a genuine two-sided risk for fixed income.
- Crypto: A token tax concept is notable for digital-asset markets because it treats machine-to-machine economic activity as a taxable base, potentially legitimizing on-chain AI agent transactions while raising compliance questions for tokenized compute and inference markets.
- Commodities: Capital deepening implies heavy demand for power, cooling, and semiconductor supply chains — bullish for electricity, copper, and uranium; bearish for labor-intensive service inputs.
- Currencies: Economies that adopt AI fastest could see relative productivity gains and currency strength, while those reliant on labor-intensive exports face structural headwinds.
Key Takeaways for Investors
- Do not expect AI productivity gains to appear in macro data quickly; the J-curve means the payoff is back-loaded and lumpy.
- Position for a world where capital, not labor, captures the marginal dollar of AI-driven growth — favoring equity owners and infrastructure providers.
- Watch for policy signals around consumption taxes and automation levies; these would be regime-changing for both equities and digital assets.
- Treat McCrory’s 1.8-point productivity estimate as an upside scenario, not a base case — the gap between capability and adoption remains the key uncertainty.
