Kevin Warsh and the New Age of Investment: What History Says About High Rates, Productivity and Stock Markets
High rates do not decide the equity outcome by themselves. History suggests that the decisive variables are why rates are rising, how quickly they move and whether productivity and earnings can keep the return on new investment above the cost of capital.
At Jackson Hole in August, Federal Reserve Chair Kevin Warsh said, “We've come to a hinge point in history.”
He was describing a world that looks very different from the era of secular stagnation and excess savings. Capital is no longer waiting for a compelling use. It is pouring into data centers, semiconductors, power generation, transmission networks, software and artificial-intelligence infrastructure. https://www.federalreserve.gov/newsevents/speech/warsh20260828a.htm
The historical question is straightforward:
Can unusually strong investment and productivity allow stock markets to absorb higher interest rates—and what separates those regimes from inflationary or overinvestment episodes that eventually break?
The answer is conditional. High interest rates are not automatically bearish, just as falling interest rates are not automatically bullish. What matters is why rates are moving, how quickly they move and whether productivity, earnings and the expected return on marginal capital can continue to outrun the full cost of financing.
How My September View Evolved
On August 18, my base case was that the Fed would hold in September. That call was wrong. The same analysis, however, separated the policy rate from the long end and argued that AI capital deepening could support output even as hiring weakened. By August 29, after testing layoffs, claims, hours and vacancies, I revised the labor thesis: weak hiring was not yet broad contraction—“low hire, low fire”—and would not necessarily stop a hike. On September 16, the FOMC raised rates 25 basis points unanimously. The sequence was 1.an incorrect near-term call, 2. an evidence-led revision before the meeting, and 3. an outcome that made the deeper investment, productivity and long-rate question more important.
From Technological Revolution to Investment Revolution
Technological revolutions do not begin as clean productivity statistics. They begin with physical systems, new business models and large commitments of capital.
Railroads required track, locomotives, terminals, finance and a reorganization of where firms produced and sold. Railroads and American manufacturing produced substantial manufacturing-productivity gains, including gains from reallocating activity across locations.
Electrification required generators, grids, motors and redesigned factories. It was not enough to replace a steam engine with an electric motor. Firms had to reorganize production around the flexibility electricity made possible. Research using U.S. manufacturing data from 1890 to 1940 finds that electrification raised labor productivity, with complementary changes in capital intensity and industrial organization.
These episodes establish a pattern, not a market backtest. Nineteenth- and early-twentieth-century monetary systems, financial markets and government-bond benchmarks were too different to compare directly with today's 10-year Treasury yield.
The useful lesson is narrower: a general-purpose technology often becomes an investment revolution before its full productivity effects are measurable. Infrastructure comes first. Organizational change follows. Diffusion takes time. Some projects earn extraordinary returns; others finance capacity that never earns its cost.
That is why technological importance and investment success must be analyzed separately.
A Rising-Rate Investment Expansion
The strongest modern U.S. comparison begins after World War II.
The Federal Reserve's monthly 10-year constant-maturity Treasury series started in April 1953. The monthly average yield was about 2.83% then. By December 1969, it was about 7.65%.
That was not a short rate shock. It was a long upward repricing across an economy expanding its industrial capacity, household infrastructure, transportation networks and corporate investment.
Productivity provided a powerful offset. The Bureau of Labor Statistics estimates that nonfarm-business labor productivity grew at a 2.7% annualized rate from the first quarter of 1947 through the fourth quarter of 1973.
Equities did not rise every year, but they rose through long stretches of the yield increase. NYU Stern's historical series shows positive S&P 500 total returns in 12 of the 17 calendar years from 1953 through 1969. It also records losses of about 10.0% in 1966 and 8.2% in 1969.
The lesson is not that productivity makes valuation irrelevant.
It is this:
Rising rates did not prevent equities from rising for long stretches—but strong productivity did not immunize stocks from valuation shocks or corrections.
When productive opportunities are abundant, companies can accept a higher financing hurdle because expected cash flows and returns on new capital are also rising. A higher discount rate compresses what investors should pay for a given stream of earnings, but the earnings stream itself may be growing fast enough to offset that pressure.
This is the first investment-age regime:
Expected returns rise faster than the cost of capital.
In that regime, higher rates are restrictive at the margin without necessarily ending the expansion.
The 1970s Counterexample
The 1970s show why the level of nominal rates cannot identify the regime.
BLS data show nonfarm-business productivity growth slowing from 2.7% in the 1947 Q1–1973 Q4 period to 1.4% from 1973 Q4 to 1980 Q1. At the same time, the economy absorbed repeated oil shocks, rising inflation expectations and policy errors that allowed price instability to become entrenched. The Federal Reserve's own history of the Great Inflation treats it as a multi-causal episode, not the result of one shock.
That combination was fundamentally different from a productive investment expansion.
Nominal rates rose because inflation and inflation uncertainty rose. Productivity slowed. The real return available from many investments became harder to estimate. Policy eventually had to become much more restrictive. Equity valuations faced both a higher discount rate and a weaker foundation for real earnings growth.
The distinction is essential:
Productive capital formation with improving real returns can coexist with higher rates.
Inflation with weak productivity and unstable policy credibility makes high rates destructive.
A 7% yield generated by stronger real growth does not carry the same information as a 7% yield generated by unanchored inflation. The number may be identical. The economic regime is not.
The Internet Boom: Real Technology, Real Excess
The late 1990s provide a more recent warning against false choices.
From the end of 1998 to the end of 1999, the 10-year Treasury yield rose from roughly 4.65% to 6.45%. Over the same period, the Nasdaq Composite climbed from 2,192.69 to 4,069.31—about 85.6%. It continued to 5,048.62 on March 10, 2000, when the 10-year yield was still near 6.4%.
High and rising rates did not mechanically stop the melt-up. Investors expected the return from internet and communications investment to remain far above the risk-free rate. For a time, rising productivity, capital spending and market prices reinforced one another.
The technology was real. So was the overinvestment. When marginal project returns, financing conditions and expectations turned, the assets that transformed the economy did not disappear—but much of the capital built around them was impaired. That is the central warning in The AI Bubble Clock: productive investment and financial excess can coexist.
The New Age of Investment
Today's evidence is strong enough to identify an investment boom. It is not yet strong enough to prove an AI-driven productivity revolution.
Warsh said the four-quarter change in investment in equipment and intangibles was around 9%, the fastest since 2021. He estimated that more than half of this year's capex growth could likely be attributed to the AI buildout. He also pointed to strong large-company profits, elevated margins, tight credit spreads and heavy issuance.
Those observations describe a financial system still willing to fund the expansion. They do not settle where the returns will ultimately land. Warsh explicitly asked whether AI would create a sustained economy-wide rise in productivity, how capital-intensive the next generation of models would be and how the surplus would be divided among AI labs, chipmakers, energy producers, cloud providers, businesses, workers and consumers.
The Fed's September decision reinforced the near-term picture. On September 16, the FOMC voted 12–0 to raise its target range by 25 basis points to 3.75%–4.00%. Its statement said economic activity was expanding at a solid pace, productivity growth was strong and capital investment was robust. https://www.federalreserve.gov/newsevents/pressreleases/monetary20260916a.htm
The long end remained demanding. The official Treasury curve put the 10-year yield at 5.01% on September 18 and 4.96% on September 21.
Productivity has improved too. The September 3 BLS revision showed second-quarter nonfarm-business productivity 2.2% above a year earlier. From the fourth quarter of 2019 through the second quarter of 2026, it grew at a 2.1% annualized rate, compared with 1.5% in the previous business cycle.
But the detail prevents a premature victory declaration. Manufacturing productivity has grown only 0.5% annualized in the current cycle. Aggregate productivity is affected by many forces—capital deepening, business formation and exit, labor reallocation, pandemic-era restructuring, sector mix and measurement, as well as technology.
The defensible conclusion is therefore limited:
Productivity has improved while AI investment and adoption have accelerated, but aggregate data do not yet establish how much of the improvement is caused by AI.
The Real Test: Return on Marginal Capital Versus Cost of Capital
The market should not judge the investment age by total capex alone.
The important variable is the expected return on the next dollar of capital.
This framework explains why a 5% Treasury yield is a hurdle, not a stop sign. As the first GMS Market Guide explained in What Does a 5% 10-Year Treasury Yield Mean for Stocks?, the Treasury yield raises the discount rate, competes with equities for capital and lifts corporate financing costs.
This analysis asks the next question: why can some economies and companies clear that hurdle for years?
They clear it when new capital produces sufficiently strong revenue, cash flow, productivity or strategic value. They fail when project economics depend on permanently cheap financing, heroic utilization assumptions or demand that exists only because the financing ecosystem keeps supplying it.
Every investment boom eventually faces the same test:
Can the return on the next dollar of capital continue to exceed its cost?
Three Rate Paths, Different Equity Outcomes
The direction of interest rates is not enough. Investors need four variables: direction, speed, cause and earnings response.
The same logic can be expressed through a simple price-to-earnings framework.
Investment-age expansion: Earnings per share rise strongly, the required return rises modestly and the P/E multiple compresses slightly. The stock price can still rise.
Rate shock: Earnings stagnate while the required return rises sharply. The P/E multiple falls and price pressure intensifies.
Benign disinflation: Earnings remain stable or rise while the required return falls. The P/E multiple can expand.
Recessionary rate decline: Earnings fall sharply even as the required return declines. Multiple support may not offset the earnings loss.
Falling rates are not automatically bullish. Rising rates are not automatically bearish.
Investors should also separate a nominal yield into its economic components. A higher real yield may reflect stronger expected real growth, a tighter policy stance or heavier demand for capital. Higher inflation compensation is usually less constructive. A rising term premium can tighten financial conditions even without a change in the expected policy-rate path.
That is why the speed and cause of the move matter as much as the level.
What the Market Should Watch Now
The constructive investment-age case requires more than rising capex and confident management guidance.
First, productivity improvement must broaden and persist. One strong cycle-to-date aggregate does not establish a permanent regime.
Second, earnings must keep validating the investment. Revenue, operating income and free cash flow need to grow fast enough to absorb rising depreciation, power costs, lease commitments and financing expenses.
Third, credit must remain selective rather than restrictive. Tight spreads and open issuance markets support the boom today. A widening that begins in capital-dependent AI intermediaries, project finance or private credit could reveal stress before it reaches the largest hyperscalers.
Fourth, investors must distinguish average returns from marginal returns. Existing cloud platforms can be highly profitable while the next data center earns less. The return on embedded assets may remain excellent even as the economics of new capacity deteriorate.
Finally, valuation must retain some relationship to realized cash generation. A productive economy can support higher earnings. It cannot make the price paid for those earnings irrelevant.
What Would Break the New Investment-Age Thesis?
The thesis should be rejected or materially downgraded if several of the following appear together:
Nonfarm-business productivity fades back toward the previous cycle's weaker trend.
Earnings revisions turn down while AI capex and long-duration commitments remain elevated.
Cloud and AI-related revenue growth slows faster than depreciation and operating costs.
Return on invested capital deteriorates across the companies leading the buildout.
The economics of marginal projects weaken, with more delays, cancellations or renegotiations.
Credit spreads widen and refinancing becomes meaningfully more expensive or less available.
Financing structures rely more heavily on fragile leverage, collateral assumptions or guarantees.
Real yields remain high while expected corporate returns fall.
Physical capacity begins to outrun durable, independently financed demand.
No single indicator is decisive. The dangerous combination is falling marginal returns, sticky capital costs and commitments that cannot adjust quickly.
Conclusion: The Hurdle Is There, but the Return Still Decides
The strongest historical investment regimes were not defined by cheap capital alone. They were defined by productive opportunities capable of earning returns above the cost of that capital.
The postwar expansion showed that rising rates and rising equities can coexist when productivity and earnings are strong. The 1970s showed that high nominal rates become far more destructive when inflation rises and productivity weakens. The internet boom showed that real technology can transform the economy while investors simultaneously finance too much capacity at the wrong price.
AI may be opening another investment age. The scale of capital formation is already visible. The productivity payoff is promising but not yet proven, and the distribution of returns remains uncertain.
If AI investment continues to lift output, earnings and the return on marginal capital, high rates may be restrictive without being fatal. If marginal returns fall while real yields, credit spreads and long-duration commitments stay high, the boom will become a source of financial stress.
The 10-year Treasury yield does not decide that outcome by itself.
The next dollar of investment does.
Related Analysis
The AI Bubble Clock: Can the AI Boom Keep Running With 5% Treasury Yields?
How High Can U.S. Treasury Yields Go Before America Has to Push Back?
Warsh Has Drawn His Line on Inflation — What Would Stop a September Rate Hike?
Why the Fed Is Likely to Hold in September and Why the 10-Year Treasury Could Still Rise