Warsh Is Right About the Boom — But Wrong About What It Proves
Secular stagnation was never simply the idea that we had run out of things worth investing in. It was a theory about structural forces pushing the natural rate of interest lower. Those forces have not disappeared. They are being offset.
Before we declare secular stagnation dead, it helps to remember what the hypothesis actually said.
At Jackson Hole, Fed Chair Kevin Warsh looked back at the post-financial-crisis era, when economists worried about secular stagnation and a global saving glut. He described a world with too much capital, too few compelling investment opportunities and a belief that “all the good stuff had been invented.”
Now, he argues, the world looks very different. An enormous investment boom—much of it associated with the AI buildout—is underway. At the G20 meeting in Asheville, Warsh called the moment a “global investment surge.”
On the investment boom, he has a point.
Warsh may be right that secular stagnation no longer describes today’s economy. But that is different from showing that the mechanism behind it has disappeared. What changed is the forces acting on the economy, not the mechanism itself.
What secular stagnation actually means
The modern secular-stagnation hypothesis, revived by Larry Summers and formalized with Gauti Eggertsson and Neil Mehrotra, starts with a simple idea.
At full employment, there is some real interest rate—economists call it r-star—that balances how much households and businesses want to save with how much firms and governments want to invest or borrow. If desired saving rises relative to desired investment, that equilibrium rate falls. And if it falls low enough, monetary policy can run into trouble. Nominal rates cannot fall far enough to generate the real rate needed to keep the economy at full employment. The result is weak demand, low inflation and a persistent tendency toward slow growth.
There was never just one cause.
The secular-stagnation literature pointed to slower population growth, weaker productivity growth, rising inequality, deleveraging, falling investment-goods prices and foreign capital inflows as forces that could push saving up relative to investment and pull r-star lower.
Fiscal policy mattered too. Less government borrowing reduces demand for funds and can push the natural rate down. More government borrowing can push it up.
So the theory was not: we ran out of inventions.
That is much closer to Robert Gordon’s productivity pessimism.
The secular-stagnation theory was: persistent structural forces can push the natural rate of interest very low.
Those forces are being offset
That distinction matters today.
If AI creates a new wave of profitable investment opportunities—data centers, chips, software, power generation, networking equipment—firms want to invest more at any given interest rate.
That pushes in the opposite direction.
If investment pushes r-star higher, that does not invalidate secular-stagnation theory. It is the same saving-investment mechanism operating in the opposite direction.
And AI is not the only force doing the work.
Government borrowing is also much larger than it was before the pandemic. The primary federal deficit averaged about 2% of GDP during 2015–19. Since 2023, it has averaged a little above 3%. Federal debt held by the public is now close to 99% of GDP, versus roughly 75% before the pandemic. That borrowing absorbs saving and puts upward pressure on equilibrium real rates.
So the cleaner interpretation is this:
Warsh is dismissing secular stagnation, but the theory itself is intact. Elevated government borrowing and the AI investment boom have helped pull the natural rate higher and move the economy away from the low-rate trap.
The question is whether that escape lasts.
Escaping secular stagnation is not the same thing as eliminating the forces that cause it.
The distinction is testable. For Warsh’s secular-growth story to represent a durable structural break, the natural rate should remain persistently higher even as today’s elevated government borrowing eventually recedes. And if AI is the reason, the investment boom eventually has to translate into durable, economy-wide productivity gains. If those forces fade and r-star falls back, the underlying secular-stagnation pressures were offset, not eliminated.
The investment boom is real
Warsh’s strongest evidence is investment itself.
Private nonresidential fixed investment averaged 13.69% of GDP during 2015–19. In the latest quarter, it reached 14.23%. That is a large move, but it is also recent.
On a trailing 12-quarter basis, business investment averages 13.85% of GDP, only modestly above the pre-pandemic average.
The digital investment increase looks more persistent. My narrow measure of information-processing equipment plus software rose from 3.88% of GDP before the pandemic to 4.34% over the latest 12 quarters, and nearly 5% in the latest quarter.
The boom is here but calling it secular requires something more: persistence.
R-star has moved—but not decisively
The evidence on the natural rate is also moving in Warsh’s direction.
The St. Louis Fed compares six different estimates of r-star. Their geometric mean reached its highest level since the financial crisis in 2025 before slipping to 1.43% in the fourth quarter.
That is a real point in Warsh’s favor.
But the estimates remain extraordinarily dispersed—from below 1% to above 3%.
And one widely used measure, Holston–Laubach–Williams, shows almost no regime break at all. HLW r-star averaged about 1.01% during 2015–19, about 0.98% since 2023, and roughly 1.01% today.
Inside the model, stronger trend growth is pushing r-star higher. Other persistent forces are pulling almost equally hard in the opposite direction.
The HLW model does not tell us how much of the change comes from AI, fiscal policy or any other specific force. What HLW does show is that the upward contribution from stronger trend growth is being almost entirely offset by other persistent influences in the model.
That is consistent with the broader story: many of the structural forces associated with low rates are still with us.
The real test of Warsh’s secular-growth story is productivity
This is where Warsh’s argument becomes a more direct challenge to Gordon than to Summers and his co-authors.
Gordon’s benchmark is demanding because his argument is about productivity regimes measured over decades, not a few strong years.
If AI really marks the beginning of a new secular-growth era, then the investment boom must eventually produce broad and persistent productivity gains.
There is encouraging evidence.
Nonfarm-business labor productivity has grown about 2.5% annually over the past three years. Private nonfarm business TFP has also improved relative to the weak post-2007 period.
But the longer record is less dramatic. Ten-year TFP growth remains below 1%, and TFP slowed from 1.5% in 2024 to 0.8% in 2025.
A few strong years do not make a new productivity regime.
The chain Warsh needs is straightforward:
AI investment → widespread adoption → sustained productivity gains → faster potential growth.
We clearly have the investment. The rest remains to be proven.
Why this matters
If AI delivers persistent productivity gains, the economy could sustain faster growth, stronger investment demand and a permanently higher natural rate.
That really would represent a structural change.
But if the AI buildout fails to generate durable productivity gains, the story changes quickly.
We cannot assume government borrowing will provide an ever-larger boost to the natural rate indefinitely. If fiscal policy eventually tightens—through lower spending, higher taxes or both—one of today’s biggest upward pressures on rates will weaken.
Then the old structural forces remain: slower population growth, aging, inequality, strong global demand for U.S. assets and weak labor-force growth. In that world, the natural rate could drift lower again.
We could end up much closer to where we started.
So Warsh may be right that the economy has escaped the post-financial-crisis low-rate trap.
But escaping the trap does not mean the forces that created it have disappeared. For now, they are being outweighed by powerful countervailing forces: government borrowing and a historic investment boom.
The investment boom is real. The productivity evidence is getting more interesting. Whether this becomes a new era of secular growth depends on whether AI can make those gains last.
