The Half-Percentage Point

In the AI Story

The calculation incorporates Cowen's career-long attention to what actually moves median welfare rather than what generates excitement. Productivity growth of 0.5% annually does not sound transformative in a quarterly earnings call, but compounded across decades it produces the difference between stagnation and flourishing. A society that grows at 2% per year doubles its output every 35 years; at 2.5%, every 28 years. The seven-year difference means an entire generation experiences either the comfort of their parents' living standards declining relative to their own, or the anxiety of falling behind despite working harder. The Great Stagnation documented what happens when growth slows from 2% to 1.5% for forty years. The half-percentage-point boost reverses that trajectory.

Cowen's modesty is analytical, not temperamental. He has observed that markets are not currently pricing in transformative AI scenarios: research on Treasury yields around major model releases found that long-term rates fell rather than rose, signaling market expectations of slower growth, not acceleration. This creates a puzzle—if AI is genuinely transformative, why don't markets price it in? Cowen's answer: markets are pricing the institutional constraints correctly. They anticipate that the technological potential will vastly exceed the institutional capture, and they are pricing the captured fraction, not the potential. The market's skepticism validates Cowen's bottleneck thesis and explains why his estimate is lower than the technology alone would suggest.

The compound-math realization that transforms modest into momentous is available to anyone with a spreadsheet, but Cowen observes that human intuition is catastrophically bad at exponential thinking. We think linearly. We underestimate compounding. A 0.5% growth difference feels negligible year-to-year, imperceptible quarter-to-quarter, and it is—until it isn't. The first decade accumulates to perhaps 5% additional GDP. The second decade adds another 11%. By year thirty, the baseline economy has grown 64% while the AI-boosted economy has grown 87%—a gap large enough to be the difference between a society that feels it is advancing and one that feels it is treading water.

The estimate's most important feature is what it excludes: the possibility that institutional adaptation accelerates, that the bottleneck narrows, that societies learn to reorganize around AI faster than historical precedent suggests. Cowen's half-point is the realistic scenario, not the optimistic one. The optimistic scenario—where universities restructure in years instead of decades, where corporations reorganize around judgment-density rather than defending existing hierarchies, where regulatory bodies build adaptive frameworks that improve with deployment—could deliver the full two-to-three-point boost. Whether that scenario is achievable depends on choices being made now, in the institutional design phase, while the technological trajectory is still visible but the organizational response is still forming.

Origin

Cowen's 0.5% estimate appears in his 2024-2025 public commentary as the AI moment crystallized, representing his attempt to integrate technological optimism with institutional realism. It builds on growth accounting—the economic method for decomposing GDP growth into capital, labor, and total factor productivity contributions—and applies the framework to AI as a general-purpose technology comparable to electricity or the internet. The estimate also incorporates Robert Solow's productivity paradox observation ('you can see the computer age everywhere except in the productivity statistics') and Paul David's explanation of why transformative technologies take decades to show up in aggregate data. Cowen is betting that AI's organizational integration will be faster than electricity's but slower than optimists hope, producing a growth boost that is real, significant over long horizons, and modest enough to be compatible with existing institutional capacity.

Key Ideas

Compound math transforms modest into momentous. Half a percentage point annually for thirty years is not a gentle improvement but a different civilization—16% larger economy, seven-year acceleration in doubling time.

The estimate prices institutional friction. Not the technology's potential (2-3%) but the realistic fraction human institutions can capture given structural decision-making lags and organizational inertia.

Markets currently underprice AI transformation. Treasury yields fell around major model releases, signaling market skepticism that societies will capture the technological potential—validating Cowen's conservative estimate.

The long run is where growth lives. Quarter-to-quarter, the effect is imperceptible; decade-to-decade, it is the difference between advancing and stagnating—requiring patience institutions struggle to maintain.