Nobel laureate economist Paul Krugman argues on his Substack that the surge in AI‑related spending is not merely a speculative future risk but a tangible phenomenon already reshaping U.S. investment patterns. He says the main channel is higher interest rates: competition for limited financial resources pushes up rates and displaces other private investment.
Why the situation is different now
Crowding out is typically discussed in the context of government deficits: when an economy is near full employment, the private sector may already be using most available credit at prevailing rates, so government borrowing forces up interest rates and reduces private investment. Krugman stresses that the current episode differs because the private sector itself — notably AI‑related investment — is absorbing resources.
He points to data showing that since the second quarter of 2023, technology investment (information processing equipment and software) has risen by roughly one percentage point of GDP, while other private investment has fallen by about 0.85 percentage points. Krugman argues this shift helps explain the rise in long‑term interest rates, with the AI boom as a primary driver.
Data centers and construction
The clearest example, Krugman says, is data‑center construction. Using Census Bureau construction data and referencing a chart by Steve Rattner, the analysis shows data‑center construction spending grew by about $55 billion (annualized) from December 2023 through July 2026, while all other private construction fell by roughly $100 billion. Rattner’s original figures differ slightly (+$51 billion and −$120 billion), and updated August data show further growth in data‑center construction (+$60.7 billion) and a somewhat smaller decline in other private construction (−$89.0 billion). The core point remains: data centers are drawing resources away from housing, office and factory building.
Other bottlenecks and channels
Krugman notes the composition of AI spending matters: much of the money goes to equipment (notably rapidly depreciating GPUs) and software rather than buildings. Because such equipment has a short effective life, borrowing costs matter less compared with long‑lived structures. He also highlights other constraints: data centers' heavy electricity demand and stronger demand for foreign‑made semiconductors have driven up memory‑chip prices (the so‑called “RAM apocalypse”), affecting the auto industry and computer manufacturers.
Why this is not simply a repeat of the 1990s tech boom
According to Krugman, the 1990s technology boom did not show the same crowding‑out pattern: non‑technology investment also rose. His calculations for 1995 through 1999 indicate technology investment increased by about 0.9 percentage points of GDP while other private investment rose by roughly 1.2 percentage points. He attributes that difference largely to large foreign capital inflows at the time, which effectively financed the boom without displacing other investments; current balance‑of‑payments data do not show a comparable inflow.
Corporate examples and the debate over waste
Krugman rejects the claim that private firms necessarily make the right investment choices, arguing we should not assume reallocations are automatically beneficial for the country. He also points to the role of tax policy — notably Donald Trump’s tax legislation (One Big Beautiful Bill) — in effectively subsidizing some of these investments. He highlights a double standard: a failed government program spending $80 billion would prompt lengthy congressional scrutiny, while large failed private investments often attract little official attention.
As an example, Krugman cites Meta’s metaverse spending: he notes the Reality Labs division lost roughly $80 billion since 2020. Because of that failure, Meta in early 2026 cut about one‑tenth of Reality Labs' workforce, closed several VR studios, and shifted strategic focus toward AI.
Financial risks and conclusion
Krugman also raises financial stability concerns. The Wall Street Journal has reported that rising rates are already straining commercial real estate deals, and some observers fear the AI boom could create larger negative spillovers into private credit markets and less‑regulated parts of the financial system than commonly appreciated.
Krugman’s conclusion is stark: unlike hypothetical risks such as technological unemployment or a bursting bubble, crowding out driven by AI investment is happening now and at large scale.
Brief biographical note
Paul Krugman was born in 1953 in Albany, New York. He earned his PhD at MIT, taught at Princeton, and later joined the Graduate Center of the City University of New York (CUNY). He was awarded the Nobel Memorial Prize in Economic Sciences in 2008. Krugman was a long‑time columnist at The New York Times and currently publishes on his Substack.
