“…reminds us that inequality sends a signal of what society lacks most, in America’s case, entrepreneurship and risk taking.” - Lawrence Lindsey, CEO, The Lindsey Group, former Director of the National Economic Council
“Unintended Consequences is far smarter and more thought-provoking than most economics written for the general public” - Greg Mankiw, Harvard University, Former Chairman of the Council of Economic Advisors
“…reminds us that inequality sends a signal of what society lacks most, in America’s case, entrepreneurship and risk taking.” - Lawrence Lindsey, CEO, The Lindsey Group, former Director of the National Economic Council
“…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
“…a must-read for serious students of economic policy.” - Glenn Hubbard, Dean, Columbia Business School, and former Chairman of the Council of Economic Advisers
“A full-throated defense of economic dynamism.” - The Wall Street Journal
“…a fresh argument for the productive value of inequality.” - David Autor, Professor of Economics, Massachusetts Institute of Technology
“…serious thinking for serious thinkers. …a thought-provoking blueprint for growing middle- and working-class incomes.” - Mitt Romney, former Governor of Massachusetts
“…a very valuable contribution.” - Larry Summers, former Secretary of the Treasury and director of the National Economic Council, president emeritus, Harvard University
“Unintended Consequences represents the most cogent and persuasive analysis of the Financial Crisis to date.” - Andrei Shleifer, 1999 John Bates Clark Medal Winner
“…a very valuable contribution.” - Larry Summers, former Secretary of the Treasury and director of the National Economic Council, president emeritus, Harvard University
“…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
AI Summary.The Economist estimates artificial intelligence needs roughly $3tn in revenue by 2030 for the planned data-centre buildout to earn a conventional return, a figure other analysts have also reached. That equals almost a tenth of the roughly $35tn in total worker compensation across rich OECD countries.
Economist StaffThe Economist
Date Posted:
Is Database:
Database
The Economist calculates that the US data center buildout needs ~$3T in revenue by 2030 to offer a competitive ROI. That amounts to ~18% of US annual labor compensation inclusive of health care or ~10% of labor compensation across the OECD.
How much revenue does the AI buildout need to pay off?
Core argument: The Economist calculates that AI needs roughly $3trn in revenue by 2030 for the planned data-centre buildout to deliver a conventional rate of return.
The Economist calculates that AI needs around $3trn in revenue by 2030 for the planned data-centre buildout to offer a conventional rate of return. Other analysts, including Stijn Van Nieuwerburgh of Columbia Business School and a duo at the Stanford Institute for Economic Policy Research, have reached similar figures. To put that number into perspective, all compensation for all workers—wages, plus perks like health insurance—adds up to around $16.5trn in America, or $35trn across the OECD, a club mostly of rich countries. So AI would need to lay claim to almost a tenth of today’s rich-world wages for the data-centre boom to pencil out. Of course, that could happen in a more benign fashion if the entire economy grows swiftly—and AI evangelists do also expect dramatic growth from here.
Takeaways by Macro Roundup® AI
The Economist calculates that AI needs roughly $3trn in revenue by 2030 for the planned data-centre buildout to deliver a conventional rate of return.
The $3trn figure amounts to almost a tenth of the roughly $35trn in total worker compensation across the OECD, which compares with about $16.5trn in America alone.
Stijn Van Nieuwerburgh of Columbia Business School and a pair of researchers at the Stanford Institute for Economic Policy Research have reached similar revenue figures, while AI evangelists expect dramatic economic growth from here.
Related Articles:
Change Is Accelerating Because It’s All Interconnected— The authors estimate that artificial intelligence equity valuations discount roughly 3.5% annual productivity growth and trend nominal output growth well above the recent past, while consensus forecasts and broader markets imply growth closer to recent experience. On that basis, either AI stocks or broader macro markets are meaningfully mispriced.
No Signs of AI Payoff Outside the Tech Sector— Revenue per worker and profit margins are rising for the seven largest technology companies but flat for the rest of the S&P 500, with no signs of artificial intelligence boosting productivity growth. The author argues that as long as spending on artificial intelligence stays confined to a small part of the economy, its macroeconomic effects will remain very limited.
AI Summary.Occupations with higher work-from-home adoption are associated with slower employment growth for young college graduates relative to middle-aged workers, an accounting exercise attributing 56%–88% of the decline in young graduates' full-time-equivalent employment rate to remote work. Measures of artificial-intelligence exposure and adoption show statistically insignificant effects.
Ernie TedeschiStripe Economics
Date Posted:
Is Database:
Database
Employment is down among less AI-exposed Americans 22-26 without a 4-year college degree, as well as young college graduates, complicating an AI-only explanation. Stripe’s analysis indicates that work from home better predicts young graduates’ employment losses relative to older workers than does AI.
Is remote work, not AI, behind weaker young-graduate hiring?
Core argument: Each additional 10 percentage points of occupational work-from-home adoption is associated with employment rates for young college graduates growing 3.9 to 6.6 percentage points more slowly than for middle-aged workers, a statistically significant relationship.
The left panel shows the relationship between the occupational WFH [Work from home] share (x-axis) and growth in our young-grad–mid-career employment spread (y-axis). The relationship is negative and statistically significant. Between 2023 and the last 12 months, for every extra 10pp of WFH adoption, employment rates for young college grads grew between 3.9 to 6.6pp more slowly than those for middle-aged workers. The middle panel replaces WFH adoption with AI exposure as a separate explanatory variable. As jobs become more exposed to AI, the effect on the new-grad employment spread is also negative but much shallower than with WFH and statistically indistinguishable from zero. The same is true of industry AI adoption, shown in the right panel: industries that self-reported a higher likelihood of using AI in 2026 H1 actually saw, if anything, slightly faster young-grad employment growth relative to mid-career, though again, this effect is statistically insignificant. Our model is suggestive rather than truly causal.
Takeaways by Macro Roundup® AI
Each additional 10 percentage points of occupational work-from-home adoption is associated with employment rates for young college graduates growing 3.9 to 6.6 percentage points more slowly than for middle-aged workers, a statistically significant relationship.
In an accounting rather than causal exercise, work-from-home accounts for 0.69 to 1.08 percentage points of the 1.23-percentage-point decline in the young-grad full-time-equivalent employment-population ratio between 2023 and 2026, or 56% to 88% of the total.
AI exposure shows a negative but much shallower relationship with the young-grad employment spread that is statistically indistinguishable from zero, and industries self-reporting higher likely AI use in 2026 H1 saw, if anything, slightly faster young-grad employment growth relative to mid-career.
Related Articles:
August 2026: Adoption Rises to 22.4%— AI adoption has risen to 22.4%, with labor market effects visible but narrow: employment drags appear in marketing, graphic design, customer service, and some tech occupations, offset by 209k in added data center-exposed construction jobs. Corporate layoffs attributed to AI totaled 113k year-to-date, including 11k in a single month.
Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence— Employment for workers aged 22–25 in the occupations most exposed to artificial intelligence has declined roughly 18pp relative to less-exposed peers, with no comparable decline in other age groups. The authors find this operates mainly through reduced hiring, concentrated where AI substitutes for human tasks rather than complements them.
Looking for the Ladder— The article argues that the hiring downturn in occupations most exposed to generative AI reflects interest-rate sensitivity rather than AI itself, since roughly 38% of workers in the most AI-exposed quintile work in information, finance and insurance, and professional and technical services, versus under 2% in the least-exposed quintile.
AI Summary.The author argues the AI boom's economic significance lies more in stock market gains than in data center construction, and that the inability to monetize AI investments could end the boom. Simulations using the Federal Reserve's FRB/US model suggest that an end to the buildout plus a roughly 30% market decline would slow growth to about 1.5%, short of recession.
Jason FurmanEconvitals
Date Posted:
Is Database:
Database
Is Important:
Important
Furman estimates that the data-center buildout, which began in mid-2024, had raised US GDP by only ~0.3% by 2026Q3, as the equipment is largely imported and the US was already near full employment. Counting AI-driven stock market gains, the effect rises to ~0.6%. If the buildout ended and stocks fell ~30%, growth would slow to ~1.5%.
How would an end to the AI boom affect growth?
Core argument: If the AI data center buildout ended and the stock market fell about 30%, the FRB/US model predicts growth would slow to roughly 1½ percent, short of even a shallow recession.
The data center buildout is set to be among the largest investment expansions in peacetime American history. How different is our economy today because of it? The answer, based on a simulation using the large-scale macroeconomic model used by the Federal Reserve (FRB/US), is not much. The buildout by itself has raised GDP by only about 0.3%, because most of the investment is imported, and imports do not add to U.S. GDP or lower the unemployment rate. The stock market boom that came with it matters more—together they have raised GDP by about 0.6% and lowered the unemployment rate by about 0.3 percentage point.This analysis suggests that the economy could survive less data center buildout just fine. Investment would fall, but so would imports and interest rates, and GDP would be little changed. A broader fall in the stock market would be a bigger deal, The bottom line: so far the AI economy is less a story about data centers than about the stock market—and that is also where the risk is.
Takeaways by Macro Roundup® AI
If the AI data center buildout ended and the stock market fell about 30%, the FRB/US model predicts growth would slow to roughly 1½ percent, short of even a shallow recession.
The economy could withstand less data center buildout, making the AI economy so far less a story about data centers than about the stock market, which is also where the risk lies.
Public policy changes or an inability to monetize AI investments could stop the AI boom, and financial-system complexities such as unwinding leverage, absent from the FRB/US model, pose a tail risk resembling 2001 more than 2007–09.
Related Articles:
Change Is Accelerating Because It’s All Interconnected— The authors estimate that artificial intelligence equity valuations discount roughly 3.5% annual productivity growth and trend nominal output growth well above the recent past, while consensus forecasts and broader markets imply growth closer to recent experience. On that basis, either AI stocks or broader macro markets are meaningfully mispriced.
The S&P 500 And Cembalest’s “AI Universe” After The Midterms— An "AI universe" index of 42 stocks groups companies into three categories: 28 direct AI firms including NVIDIA, Microsoft, Alphabet, Amazon, Meta and Broadcom; 8 AI-linked utilities such as Vistra, NextEra and Constellation; and 6 capital equipment makers including Eaton, Quanta and GE Vernova.
US Labor Productivity Increases at Fastest Pace in a Year— US labor productivity, measured as nonfarm business output per hour, rose at a 3.6% annualized rate, up from an upwardly revised 2.1% pace, while unit labor costs fell 0.7%. The acceleration suggests a capital spending boom is translating into greater efficiency, according to Bureau of Labor Statistics data.
AI Summary.US labor productivity, measured as nonfarm business output per hour, rose at a 3.6% annualized rate, up from an upwardly revised 2.1% pace, while unit labor costs fell 0.7%. The acceleration suggests a capital spending boom is translating into greater efficiency, according to Bureau of Labor Statistics data.
Julia FanzeresBloomberg
Date Posted:
Is Database:
Database
Is Important:
Important
Nonfarm business productivity grew at a 3.6% annualized rate in Q3, the fastest pace in a year. Real hourly compensation rose 1.1%, while the labor share fell to 53%, the lowest in data going back to 1947.
Is a capital spending boom lifting worker productivity?
Core argument: US nonfarm business output per hour rose at a 3.6% annualized rate in the third quarter, the fastest pace in a year, following an upwardly revised 2.1% pace in the prior quarter, according to the Bureau of Labor Statistics.
US labor productivity accelerated in the third quarter at the fastest pace in a year, suggesting a capital spending boom is translating into greater efficiency at a time when input costs have soared. Productivity, or nonfarm business output per hour, increased at a 3.6% annualized rate after an upwardly revised 2.1% pace in the prior quarter, data from the Bureau of Labor Statistics showed Thursday. Unit labor costs — what businesses pay employees to produce one unit of output — fell 0.7%, the first decline since the second quarter of 2025.
Takeaways by Macro Roundup® AI
US nonfarm business output per hour rose at a 3.6% annualized rate in the third quarter, the fastest pace in a year, following an upwardly revised 2.1% pace in the prior quarter, according to the Bureau of Labor Statistics.
Unit labor costs, what businesses pay employees to produce one unit of output, fell 0.7% in the third quarter, the first decline since the second quarter of 2025.
The third-quarter productivity acceleration suggests a capital spending boom is translating into greater efficiency at a time when input costs have soared.
Related Articles:
Change Is Accelerating Because It’s All Interconnected— The authors estimate that artificial intelligence equity valuations discount roughly 3.5% annual productivity growth and trend nominal output growth well above the recent past, while consensus forecasts and broader markets imply growth closer to recent experience. On that basis, either AI stocks or broader macro markets are meaningfully mispriced.
Higher Utilisation Explains The Recent Surge In Productivity Growth— Measured total factor productivity growth is estimated to come almost entirely from higher input utilisation rather than underlying efficiency gains, based on a method that infers unobserved capital and labour utilisation from movements in hours per worker. On this evidence, recent productivity gains reflect "working harder" rather than "working smarter.
US Focus: The Effect Of Soaring Profits— US nominal pre-tax corporate profits in the national accounts grew 23% over the past year, a pace not seen outside post-recession spikes since the mid-2000s. Wider margins were the key driver, as domestic profit growth far exceeded the 8% rise in corporate value added, pushing margins near all-time highs while the labor share hit new lows.
AI Summary.Nearly every major local ballot measure opposing data centers won, putting an estimated 37.5 GW of planned capacity at risk of a pause and another 15 GW subject to conditions such as grid cost-sharing, water limits, and ended tax incentives. The 37 GW expected in 2027, 75% already under construction, is largely unaffected.
Michael CembalestJ.P. Morgan
Date Posted:
Is Database:
Database
Is Important:
Important
Before the election, 96 GW of data center capacity was expected to come online in 2028. Based on the positions of incoming governors, Cembalest estimates 37.5 GW may be subject to a “pause” and another 15 GW to “conditions,” together more than half of the total.
Are local ballot measures putting planned data center capacity at risk?
Core argument: Nearly every major local ballot measure opposing data centers won or is winning, with North Dakota the only exception, and based on those results 37.5 GW of planned capacity may face a “pause” and another 15 GW conditions.
The biggest market issue may be AI over the medium term; nearly every major local ballot measure opposing data centers won or is winning, with only exception in North Dakota. What we’re now watching: governor imposed constraints on new builds. This is not an issue for the 37 GW expected to be delivered in 2027, 75% of which is under construction. It’s more an issue for the 96 GW expected in 2028. Based on election results, 37.5 GW may be subject to a “pause” with another 15 GW subject to “conditions”. While pauses are the more impactful outcome, conditions can be onerous: requiring data centers to pay grid costs to protect ratepayers, ending tax incentives, imposing water limits, setting requirements for onsite backup power, mandating labor requirements and banning non-disclosure agreements during permitting (most of which seem sensible to me). If the data center buildout starts to slow at some point, upgrading to better NVIDIA chips might be the only way to generate more tokens.
Takeaways by Macro Roundup® AI
Nearly every major local ballot measure opposing data centers won or is winning, with North Dakota the only exception, and based on those results 37.5 GW of planned capacity may face a “pause” and another 15 GW conditions.
The 37 GW of data center capacity expected in 2027, 75% of which is under construction, is less exposed to local opposition than the 96 GW expected in 2028.
Conditions imposed on data centers can include paying grid costs to protect ratepayers, ending tax incentives, water limits, onsite backup power requirements, labor mandates, and bans on non-disclosure agreements during permitting.
Related Articles:
Financing the AI Buildout— The authors estimate roughly 57 gigawatts of operating US data center capacity against a planned development pipeline of about 509 gigawatts, which is not a completion forecast since projects may be delayed, scaled back, or abandoned. Their central scenario has 182.8 gigawatts of added capacity becoming operational through 2032, 117.2 gigawatts completed later, and 226.9 gigawatts never completed.
Global Views: Downside Growth Risks— The growth boost from the artificial intelligence boom is projected to peak around late 2026 or early 2027, with AI investment as a share of GDP slowing and the estimated 0.4pp wealth-effect lift to US consumer spending fading. Across US nonmanufacturing industries, AI adoption shows a statistically significant negative correlation with employment growth and positive correlation with unemployment.
Change Is Accelerating Because It’s All Interconnected— The authors estimate that artificial intelligence equity valuations discount roughly 3.5% annual productivity growth and trend nominal output growth well above the recent past, while consensus forecasts and broader markets imply growth closer to recent experience. On that basis, either AI stocks or broader macro markets are meaningfully mispriced.
AI Summary.The authors estimate that artificial intelligence equity valuations discount roughly 3.5% annual productivity growth and trend nominal output growth well above the recent past, while consensus forecasts and broader markets imply growth closer to recent experience. On that basis, either AI stocks or broader macro markets are meaningfully mispriced.
Greg JensenBridgewater Associates
Date Posted:
Is Database:
Database
Bridgewater estimates that AI equities are pricing in ~3.5% annual productivity growth, while real growth and productivity implied by breakeven earnings in the broader stock market are slightly below current levels. They conclude: “Either AI stocks or broader macro markets are very meaningfully mispriced.”
Do AI stock valuations and broader markets imply incompatible growth expectations?
Core argument: AI equities are priced for roughly 3.5% annual productivity growth, more than is estimated as needed to make US debts sustainable, and across a reasonable range of assumptions imply trend nominal GDP growth well above the recent past.
Different markets are pricing very different futures; most are far from pricing an unrecognizable world. The AI boom has contributed tens of trillions of dollars of market cap in AI companies. As in any gold rush, trillions of dollars of “picks and shovels” only make sense if a great deal of gold (profits) is found. The pricing of a broader universe of AI companies tells us something about the scale of transformation that is discounted. If we calculate the incremental revenue required to sustain the valuation of AI companies and estimate how much of that can be earned by cannibalizing the revenue of existing companies, we can back out a productivity boost required to generate the rest of the incremental revenues. Today, we estimate that AI equities are discounting something like 3.5% annual productivity growth. That’s more than we estimate is needed for the US to make its debts sustainable. Across a reasonable range of assumptions, we assess AI stocks to be pricing trend nominal GDP growth well above the recent past. By contrast, consensus estimates and broader markets imply a world that looks more like the recent past. Making sense of the pricing in AI stocks requires a boom in productivity and growth. Consensus estimates for growth and productivity (from economist and Fed estimates) are slightly lower than current measures, as are rates of real growth and productivity implied by breakeven earnings in the broader stock market.
Takeaways by Macro Roundup® AI
AI equities are priced for roughly 3.5% annual productivity growth, more than is estimated as needed to make US debts sustainable, and across a reasonable range of assumptions imply trend nominal GDP growth well above the recent past.
Consensus growth and productivity estimates from economists and the Fed, along with real growth and productivity implied by breakeven earnings in the broader stock market, sit slightly below current measures, describing a world resembling the recent past.
Either AI stocks or broader macro markets are very meaningfully mispriced, since the AI boom has added tens of trillions of dollars of market cap that only makes sense if a great deal of profit is found.
Related Articles:
No Signs of AI Payoff Outside the Tech Sector— Revenue per worker and profit margins are rising for the seven largest technology companies but flat for the rest of the S&P 500, with no signs of artificial intelligence boosting productivity growth. The author argues that as long as spending on artificial intelligence stays confined to a small part of the economy, its macroeconomic effects will remain very limited.
Califoreboding— A basket of 42 AI stocks accounts for 60%-80% of all S&P 500 earnings growth, price returns and capital spending, while 60% of S&P 500 stocks sit 20% or more below their highs and only 30% outperform the index. AI-related construction spending has risen 38%, the only positive trend as manufacturing, commercial real estate and residential construction decline.
AI and Productivity— The recent pickup in US labor productivity appears driven by higher utilization of existing capital—running factories, server racks, and hotel rooms harder—rather than by firm-level efficiency gains from AI. Adoption remains too thin and the cross-sectional evidence too weak to conclude that AI-driven productivity gains are already widespread.
AI Summary.The cyclically adjusted price-earnings ratio for the S&P 500 stands at 40.7, implying an earnings yield of 2.46% and, combined with an estimated 1.8% annual real per-share earnings growth, a real expected return near 3% a year. Against a 2.75% inflation-protected 10-year Treasury yield, that leaves an equity premium of roughly 25 basis points, assuming the elevated multiple holds.
Hanno LustigThe Two Cents
Date Posted:
Is Database:
Database
At 40.7, CAPE is nearing its December 1999 peak of 44.2. For the S&P 500 to beat TIPS over the next decade, today’s near-record P/E multiple must persist for ten years, or real earnings must grow more than 5.8% a year—a pace exceeded in only two decades since 1880: the 1920s and 2010s.
How much do stock valuations offer over inflation-protected Treasuries?
Core argument: The S&P 500’s cyclically adjusted price-earnings ratio stands at 40.7, close to an all-time high, implying an earnings yield of just 2.46% for buyers of the index.
The CAPE is currently trading at 40.7, which is close to an all-time high. You’re paying close to 41 times annual earnings when you buy the S&P 500. The implied earnings yield is only 2.46%. According to AQR, a reasonable estimate for real earnings growth (on a per share basis) is 1.8% per annum. If you plug in today’s numbers, you end up with a real expected return of around 3% per year for U.S. equity. That’s a surprisingly low number. When you compare it to the 10-year TIPS yield of 2.75%, you’re left with a razor-thin equity premium. The stock market will only yield an extra 25 basis points per annum over the next, say, 10 years to compensate you for taking on equity risk. You might think this is a very conservative estimate, but it’s actually not, because we’re implicitly assuming that the CAPE stays at 40.7 over the next 10 years. That’s quite a lofty multiple. Even casual eyeball econometrics suggests that that’s not a great assumption, because the price-earnings ratio tends to revert back to its mean.
Takeaways by Macro Roundup® AI
The S&P 500’s cyclically adjusted price-earnings ratio stands at 40.7, close to an all-time high, implying an earnings yield of just 2.46% for buyers of the index.
Combining the 2.46% earnings yield with AQR’s estimate of 1.8% annual real per-share earnings growth yields a real expected return of roughly 3% per year for U.S. equity.
Against a 10-year TIPS yield of 2.75%, U.S. equities offer an equity premium of about 25 basis points per year, an estimate that assumes the CAPE stays at 40.7 for a decade despite the price-earnings ratio’s tendency to revert to its mean.
Related Articles:
Financing the AI Buildout— The authors estimate roughly 57 gigawatts of operating US data center capacity against a planned development pipeline of about 509 gigawatts, which is not a completion forecast since projects may be delayed, scaled back, or abandoned. Their central scenario has 182.8 gigawatts of added capacity becoming operational through 2032, 117.2 gigawatts completed later, and 226.9 gigawatts never completed.
A Catalyst For The AI Bubble Break— Historically, controlling for valuations, a 1% rise in the share of the market represented by initial public offerings has been associated with roughly a 4% decline in prices. With large artificial-intelligence company listings expected and share issuance exceeding buybacks, equity supply could shift from shrinking about 1% a year to growing near 5%, implying returns roughly 20% below average.
One Hundred Years in the U.S. Stock Markets— Over a century, U.S. publicly traded stocks generated roughly $91tn in shareholder wealth enhancement, with returns strongly positively skewed and dominated by a relatively small number of high-performing firms. Because most possible outcomes fall below the mean in a skewed distribution, the authors argue financial planning should focus on median outcomes or the full distribution.
AI Summary.A National Bureau of Economic Research paper projects U.S. medical spending, 18% of GDP in 2024, reaching 19%–26% of GDP by 2040, with federal medical spending consuming 45%–61% of federal revenue excluding Social Security contributions. The author argues only the low-spending path is consistent with current tax levels and other commitments.
David CutlerNational Bureau of Economic Research
Date Posted:
Is Database:
Database
From 1970 to 2010, growth in US medical spending outpaced GDP growth by ~2.4pp annually. From 2010 to 2024, that growth differential fell to just 0.3pp. If it persists, medical spending reaches only 19% of GDP by 2040; if excess spending growth reverts to 2.4pp, it reaches 26%.
How much federal revenue will medical spending consume by 2040?
Core argument: David Cutler projects U.S. medical spending reaching 26%, 22.3%, or 19% of GDP by 2040 under high, benchmark, and low scenarios, versus 18% in 2024 and 6.9% in 1970.
Figure 1 shows that in 1970, medical care was 6.9% of GDP. In 2024, it was 18%. From 1970 through around 2010; mean medical spending growth exceeded GDP growth by 2.4 pp annually. Since 2010, the growth of medical spending has exceeded the growth of GDP by only 0.3 pp annually. There is no comparable time period with anything like the sustained slowdown in spending growth since 2010. Patient cost-sharing has increased enormously [indirectly reducing overall expenditure via incentives] as shown in Figure 5. An additional role [was] played by the diffusion of cost-reducing technologies. Because they are often less invasive, second-round therapies typically have lower unit prices. [mechanism in Fgure 6]. Finally, there [are] fewer expensive new innovations than in the past. [However], recent data show an uptick in spending growth in 2024, which continued in 2025. Some of this is a result of new technologies— prime examples are GLP-1 medications and new drugs for cancer that are taken on an ongoing basis. To consider a scenario with [a reversion to] more rapid spending increases, I assume that the growth of medical spending above GDP growth after 2024 returns to the average rate realized between 1970 and 2010. This scenario is shown as the High-spending scenario in figure 1, in which medical care in 2040 rises to 26% of GDP.
Takeaways by Macro Roundup® AI
David Cutler projects U.S. medical spending reaching 26%, 22.3%, or 19% of GDP by 2040 under high, benchmark, and low scenarios, versus 18% in 2024 and 6.9% in 1970.
Federal medical spending consumed 45% of “discretionary revenue”—total federal revenue less Social Security contributions—in 2024, up from 13% in 1987, and Cutler projects 61% by 2040 under the high-spending scenario.
Cutler states that only the low-spending scenario is consistent with current tax levels and non-medical spending commitments, attributing the post-2010 slowdown to high-deductible plans, prior authorization, cost-reducing technology, and reduced administrative friction.
Related Articles:
Has the United States Bent the Health Care Cost Curve?— The authors estimate that a slowdown in US health care spending reflects technology-associated changes in health and site of care (21% of the slowdown, double that in Medicare), more elastic long-run supply (6%), market changes reducing and making demand more elastic (11–27%), improved population health (7%), and slower price growth net of upcoding (24%).
Can AI Avert the Impending Federal Budget Crisis?— U.S. GDP per capita, about $87,000, is projected to reach $102,000 by 2035 with 1.8% annual productivity growth, or $110,000 with 2.7% growth. Because the healthcare sector has been slow to adopt AI, health spending is projected to rise from 18.3% of GDP to 21% under slower growth and 21.7% under faster growth.