Edward Conard

Top Ten New York Times Bestselling Author

  • “…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 fresh argument for the productive value of inequality.” - David Autor, Professor of Economics, Massachusetts Institute of Technology
  • “…a comprehensive explanation of the modern economy.” - Julian Robertson, Founder, Tiger Management
  • “Unintended Consequences offers deep and well-argued analyses on almost every issue.” - The New York Times
  • “…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
  • “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
  • “…serious thinking for serious thinkers. …a thought-provoking blueprint for growing middle- and working-class incomes.” - Mitt Romney, former Governor of Massachusetts
  • “…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
  • “…a fresh argument for the productive value of inequality.” - David Autor, Professor of Economics, Massachusetts Institute of Technology
  • “A full-throated defense of economic dynamism.” - The Wall Street Journal
  • “…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
Upside of Inequality Oxford Unintended Consequences

Macro Roundup


Friday, October 3, 2025

AI Summary. The author argues that large language models are compressing variance in knowledge domains such as writing, coding, finance, and reasoning, much as baseball lost its .400 hitters when top performers' relative distinctiveness faded. As rising general competence masks the tails of the distribution, societies become less able to identify, reward, and sustain true outliers.

Paul Kedrosky Applied Complexity
Date Posted:

Drawing a parallel with the disappearance of .400 hitters as average player performance improved, Kedrosky argues LLM adoption means rising baseline intellectual performance while “true outliers” are more difficult to find and sustain.

Do AI models make top performers harder to identify?

Core argument: The author argues that large language models are compressing variance across knowledge domains including writing, coding, finance, reasoning, and creating, pushing the tails of performance distributions closer to the mean and thinning them.

All distributions are compressing at once under the evolutionary pressure of LLMs. Like batting averages in baseball, variance in many/most knowledge domains is collapsing. The consequence: our ability to identify true outliers—our “.400 hitters” of thought, design, or creativity—erodes. The tails of the distribution exist, but they are closer to the mean, less populated, and masked by the rising competence of everyone else. This is not about tools “making us better.” It is about the compression of variance in creative domains where variance mattered: writing, coding, finance, reasoning, and creating. Just as baseball lost its .400 hitters—not because top hitters no longer existed, but because their relative distinctiveness disappeared—intellectual and creative .400 hitters will become much harder to detect. Evolution does not necessarily march toward higher peaks but toward tighter distributions around a plateau. Once variance collapses, the outliers vanish into the baseline. And societies that cannot recognize outliers cannot reward or sustain them. Or they seize upon stranger differentiators, divorced from actual performance, which is a topic for another day.

Takeaways by Macro Roundup® AI

  1. The author argues that large language models are compressing variance across knowledge domains including writing, coding, finance, reasoning, and creating, pushing the tails of performance distributions closer to the mean and thinning them.
  2. The author draws an analogy to baseball's disappearance of .400 hitters, which reflected the loss of top hitters' relative distinctiveness rather than their absence, to argue intellectual and creative outliers will become much harder to detect.

Related Articles:

  • Long-Term Shareholder Returns: Evidence From 64,000 Global Stocks — A study of about 64,000 global stocks estimates net global stock market wealth creation of $75.7tn, with returns highly concentrated: the top 159 firms (0.25%) account for half of net wealth creation and the top 1,526 (2.39%) account for all of it. Fewer than half of stocks beat one-month US Treasury bills, 42.6% of non-US and 44.8% of US stocks.
  • Which U.S. Stocks Generated the Highest Long-Term Returns? — Of the 29,078 U.S. common stocks in a database spanning nearly a century, 51.6% had negative cumulative returns, while 17 stocks returned more than 5m% — led by Altria at 265m%. Those top performers averaged annualized returns of just 13.47%, underscoring the value of time invested.
  • The Economics of Inequality in High-Wage Economies — Information technology created investment opportunities exceeding the supply of properly trained talent, while trade with low-wage economies, low-skilled immigration, trade deficits, and aging demographics eased constraints on low-skilled labor and risk-averse savings. The author argues this framework, rather than rising cronyism, explains high returns to success, slowing productivity growth, and low interest rates.

AI Summary. Citigroup estimates annual spending on artificial intelligence equipment has risen by 0.9% of GDP a year, a bigger surge than at the comparable point of the 1995-2004 internet boom, when productivity growth averaged 2.9% a year and economic growth reached 3.3%. Citigroup economists argue recent growth has been boosted by demand from that investment rather than by supply-side productivity gains.

Greg Ip Wall Street Journal
Date Posted:

Citing a Citi analysis, Greg Ip suggests that over the next several years, “AI could raise productivity growth by 0.5 to 1.5 percentage points a year—less than the ‘Internet Revolution of 1995–2000,’ except in the most bullish scenario.”

Is AI investment boosting growth through demand rather than productivity?

Core argument: Citigroup estimates annual spending on AI equipment has risen by 0.9% of GDP a year since 2023, a bigger surge than at the equivalent point of the 1995–2004 internet boom.

Economists at Citigroup recently compared the [1995-2000 “Internet Revolution” to today]. The prior boom began in 1995 with the initial public offering of Netscape and the full commercialization of network infrastructure developed by the National Science Foundation. In the first five years of the boom, annual internet-related investment rose by a massive 1.25% of GDP. Annual productivity growth averaged 2.9% from 1995 to 2004, double the rate of the prior two decades. At that rate, per capita incomes double in 24 instead of 47 years. The productivity boost lifted annual economic growth to 3.3%. Citi estimates annual spending on AI equipment has risen by 0.9% of GDP a year since 2023, a bigger surge than at the equivalent point of the 1995-2004 boom. Nathan Sheets of Citi said GDP growth has been boosted by the demand-side fuel of all that AI investment, not by its supply-side productivity benefits.

Takeaways by Macro Roundup® AI

  1. Citigroup estimates annual spending on AI equipment has risen by 0.9% of GDP a year since 2023, a bigger surge than at the equivalent point of the 1995–2004 internet boom.
  2. Annual internet-related investment rose by 1.25% of GDP in the first five years of the boom that began in 1995, while productivity growth averaged 2.9% from 1995 to 2004, double the rate of the prior two decades.
  3. Nathan Sheets of Citigroup said GDP growth has been boosted by the demand-side fuel of AI investment rather than by its supply-side productivity benefits.

Related Articles:

  • The Blob: Capital, China, Chips, Chicago and Chilliwack — Data center construction spending has eclipsed office construction, and the specialized power rates most data centers pay fall short of covering the cost of a new natural gas plant, leaving other customers to bear part of the bill. In the PJM grid region, data center demand accounted for 70% of the increase in electricity costs over the past year.
  • Honey, AI Capex is Eating the Economy — Capital spending on AI data centers is around 20% of railroads' peak spending as a share of GDP and is rising quickly, having already surpassed the dot-com era peak in telecom investment. Unlike long-lived rail infrastructure, data centers are short-lived assets requiring frequent hardware replacement, and the capital is being reallocated from other sectors.
  • The Productivity Slowdown in Advanced Economies: Common Shocks or Common Trends? — Total factor productivity is the main factor accounting for differences in labor productivity growth across countries and over time, and its growth has been lackluster across large advanced economies since the mid-2000s. The authors find this slowdown appears in both manufacturing and market services, with U.S. productivity growth slowing before the Great Recession.

AI Summary. Other English-speaking countries such as the UK, Canada, Australia, Ireland and New Zealand direct more pension spending toward preventing old-age poverty and far less toward middle- and upper-income retirees, with maximum US Social Security benefits running two to four times higher in dollar terms. One proposed reform would base benefits on years worked rather than contributions.

Andrew Biggs American Enterprise Institute
Date Posted:

Andrew Biggs notes that the maximum Social Security payout for a high-income couple is almost $100,000 annually, 2 to 4 times higher in dollar terms than the payouts for their peers in other Anglosphere countries.

Should public pensions target old-age poverty rather than higher-income retirees?

Core argument: The maximum U.S. Social Security benefit approaches $100,000 per year for a high-income couple, two to four times higher in dollar terms than benefits paid in the United Kingdom, Canada, Australia, Ireland and New Zealand.

Most Anglo countries, such as the United Kingdom, Canada, Australia, Ireland and New Zealand, focus more of their pension resources on preventing poverty in old age and less, much less, on paying benefits to middle and upper income seniors. The maximum Social Security benefit, which today approaches $100,000 per year for a high-income couple, is two to four times higher in dollar terms than is paid in these other countries. A middle-income couple retiring today receives over $58,000 in combined annual benefits. By 2050, they’ll get over $81,000. At some point, a safety net becomes something else entirely. [One example of a possible reform that would reduce costs] but maintain benefits for lower-income workers and provide a stronger safety net against poverty in old age, is that Americans would earn Social Security benefits based not on how much they paid into the program but upon how many years they worked, no matter how little they earned or paid into the system. But the plan would provide incentives to work longer, as continued work would result in higher benefits.

Takeaways by Macro Roundup® AI

  1. The maximum U.S. Social Security benefit approaches $100,000 per year for a high-income couple, two to four times higher in dollar terms than benefits paid in the United Kingdom, Canada, Australia, Ireland and New Zealand.
  2. Most Anglo countries concentrate more pension resources on preventing old-age poverty and much less on benefits for middle- and upper-income seniors, while a middle-income U.S. couple retiring today collects over $58,000 in combined annual benefits, rising to over $81,000 by 2050.
  3. One proposed reform would base Social Security benefits on years worked rather than amounts paid in, reducing costs while maintaining benefits for lower-income workers and providing higher benefits for those who continue working.

Related Articles:

  • Trends in the Distribution of Family Wealth, 1989 to 2022 — The Congressional Budget Office estimates that the top 10% of U.S. families held 60% of all wealth in 2022, up from 56% in 1989, with the top 1% holding 27%, up from 23%. The rest of the top half saw its share fall from 37% to 33%, while the bottom half held 6% in both years.
  • Social Security and Trends in Wealth Inequality — The authors estimate that including Social Security in household wealth leaves top wealth shares little changed over three decades, with the top 10% and top 1% shares rising 1.1 and 1.6 percentage points versus roughly 9 and 6 points for marketable wealth alone. Social Security wealth grew from $7tn to $39tn.
  • Europe’s Right-Wing Populists Swerve Left on the Economy — Right-wing populist parties across Europe, including the Sweden Democrats, Austria's Freedom Party, Germany's Alternative for Germany and France's National Rally, have shifted from small-government, low-tax, free-trade platforms toward calls for higher welfare benefits, subsidies and protectionism, often alongside pro-market positions.

AI Summary. An analysis of 250,000 adults in more than 50 countries finds time spent on social media peaked in 2022 and has since declined steadily, with developed-world adults aged 16 and older averaging 2 hours 20 minutes daily, almost 10% below 2022. North America is the exception, where usage continues to climb.

John Burn-Murdoch Financial Times
Date Posted:

According to an FT analysis, time spent on social media globally has declined 10% since its 2022 peak. An exception is North America, where in 2024 users spent 15% more time on social media than Europeans.

Is global time spent on social media declining?

Core argument: Adults aged 16 and older across the developed world spent an average of 2 hours 20 minutes per day on social media at the end of 2024, down almost 10% since the 2022 peak, according to GWI data on 250,000 adults in more than 50 countries compiled for the FT.

It has gone largely unnoticed that time spent on social media peaked in 2022 and has since gone into steady decline, according to an analysis of the online habits of 250,000 adults in more than 50 countries carried out for the FT by the digital audience insights company GWI. And this is not just the unwinding of a bump in screen time during pandemic lockdowns — usage has traced a smooth curve up and down over the past decade-plus. Across the developed world, adults aged 16 and older spent an average of two hours and 20 minutes per day on social platforms at the end of 2024, down by almost 10% since 2022. Notably, the decline is most pronounced among the erstwhile heaviest users — teens and 20-somethings. There is one notable exception to this promising international trend: North America, where consumption of social media’s diet of extreme rhetoric, engagement bait and slop continues to climb. By 2024 it had reached levels 15% higher than Europe.

Takeaways by Macro Roundup® AI

  1. Adults aged 16 and older across the developed world spent an average of 2 hours 20 minutes per day on social media at the end of 2024, down almost 10% since the 2022 peak, according to GWI data on 250,000 adults in more than 50 countries compiled for the FT.
  2. The decline in social media time is most pronounced among teens and 20-somethings, formerly the heaviest users, and usage has traced a smooth rise and fall over more than a decade rather than an unwinding of pandemic lockdown screen time.
  3. North America is the exception to the international decline, with social media consumption still climbing and reaching levels 15% above Europe by 2024.

Related Articles:

  • What Are TikTok’s New Owners Buying? — More than 116m Americans use TikTok each month, with 18-to-25-year-olds spending nearly an hour a day scrolling or posting videos. Teenagers spend more time on TikTok than on YouTube and Instagram combined.
  • Teens Used To Be Happier. What Changed? — The share of US high school seniors describing themselves as "very happy" peaked at 24.8% and remained around 24.4% before falling to 11.4% at its low, recovering only to 12.7%, well below the 17.9% recorded before the pandemic. The decline coincided with smartphones and social media becoming pervasive.
  • What’s Behind the Rapid Rise in LGBTQ Identity? — Young women use social media more than young men and are more likely to encounter accounts featuring LGBTQ topics, and research finds young women identifying as gay, lesbian, bisexual or queer are much more active on these platforms than straight women of the same age.

AI Summary. Inflation-adjusted US house prices are 2.7% below their 2022 peak. An affordability index built from median income, a 15% down payment, and 30-year fixed mortgage rates shows the monthly principal-and-interest payment on a $500,000 house rose 94% over four years as prices climbed 24% and mortgage rates went from 2.87% to 6.72%.

Bill McBride Calculated Risk
Date Posted:

Nationally, real house prices are now 2.7% below the 2022 peak and 9.8% above the 2006 peak. “Affordability” remains at levels last seen in the early 1980s, when 30-year mortgage rates were over 14%.

How have house prices and mortgage rates changed monthly payments?

Core argument: Inflation-adjusted U.S. house prices stand 2.7% below their 2022 peak, based on the Case-Shiller National Index, seasonally adjusted.

I’ve put together my own affordability index [using] median income from the Census Bureau (estimated 2024 and 2025), assumed a 15% down payment, and used a 2% estimate for property taxes, insurance and maintenance. This is probably low for high property tax states like New Jersey and Texas, and too high for lower property tax states. For house prices, I used the Case-Shiller National Index, Seasonally Adjusted (SA). Also, for the down payment - there wasn’t a significant difference between 15% and 20%. For mortgage rates, I used the Freddie Mac PMMS (30-year fixed rates). Here is what the index looks like (lower is more affordable). In July 2021, the payment on a $500,000 house, with a 20% down payment and 2.87% 30-year mortgage rates, would be around $1,658 for principal and interest. The monthly payment for the same house, with house prices up 24% over four years and mortgage rates at 6.72% in July 2025, would be $3,218 - an increase of 94%! Almost double the payment for the same house!

Takeaways by Macro Roundup® AI

  1. Inflation-adjusted U.S. house prices stand 2.7% below their 2022 peak, based on the Case-Shiller National Index, seasonally adjusted.
  2. The monthly principal-and-interest payment on a $500,000 house rose 94%, from about $1,658 in July 2021 at a 2.87% 30-year mortgage rate to $3,218 in July 2025 at 6.72%, with prices up 24% over four years.
  3. The affordability index uses Census Bureau median income with 2024 and 2025 estimated, a 15% down payment, a 2% estimate for property taxes, insurance and maintenance, and Freddie Mac PMMS 30-year fixed rates, with little difference between 15% and 20% down.

Related Articles:

  • Why Are Young Adults in the English-Speaking World So Unhappy? — The share of young adults regularly experiencing stress and anger has risen sharply in the US, Canada, UK, Ireland, Australia and New Zealand while remaining largely stable elsewhere in the West. Real house prices since 1995 are up 85% in the US and over 200% in the other five countries.
  • How America’s AI Boom Is Squeezing The Rest Of The Economy — About a sixth of America's 2% real GDP growth over the past year came from investment in computer and communications equipment and data centres, and one estimate puts the artificial intelligence boom's total contribution to growth at 40%. Average electricity bills rose 7%, attributed in part to data-centre demand, while housebuilding and non-AI business investment slumped and real consumption flatlined.
  • Rapidly Growing Treasury Supply Crowding Out Other Types of Credit Growth — Roughly half of all fixed income issuance coming to market over the past year has been US Treasuries, meaning investors must allocate a growing share of dollars to financing the government. The argument is that lower government debt would free up more funds for consumer borrowing and business investment.

AI Summary. New York City's Democratic mayoral front-runner has proposed ending the public school system's gifted and talented program for kindergartners while keeping current participants and a separate third-grade admissions track. The program enrolls about 2,500 of the city's 55,000 kindergartners, selected by teachers during pre-K, and is often viewed as a stepping stone to competitive middle and high schools.

Troy Closson New York Times
Date Posted:

Zohran Mamdani, the Democratic front-runner for Mayor in New York City, is campaigning on ending the city’s small gifted and talented programs for kindergarten students at public schools. The program enrolls 2,500 out of 55,000 kindergartners.

Should gifted education for kindergartners be eliminated in public schools?

Core argument: Zohran Mamdani, the Democratic front-runner for New York City mayor, said he would end the city's kindergarten gifted and talented program, which enrolls about 2,500 students out of 55,000 kindergartners.

Zohran Mamdani, the Democratic front-runner for mayor, said this week that he planned to end the city’s gifted and talented program for kindergarten students at public schools if he is elected. Students who are already in gifted classes would remain in the program, and his campaign said on Thursday that he would keep a separate track next year that admits students in third grade. New York’s program stands apart from many other major cities that admit children into gifted classes starting in later grades. It is small: New York’s school system enrolls about 2,500 students out of 55,000 kindergartners. Children are selected during pre-K by their teachers. The highly selective program is often viewed as a stepping stone to the city’s competitive middle and high schools. Many families clamor for access to gifted classes as a way of ensuring that high-achieving children can be challenged and receive prime educational opportunities in later grades.

Takeaways by Macro Roundup® AI

  1. Zohran Mamdani, the Democratic front-runner for New York City mayor, said he would end the city's kindergarten gifted and talented program, which enrolls about 2,500 students out of 55,000 kindergartners.
  2. Students already in New York's gifted classes would remain in the program, and Mamdani's campaign said he would keep a separate track next year admitting students in third grade.
  3. New York's program, in which children are selected during pre-K by their teachers, stands apart from many other major cities that admit children into gifted classes starting in later grades.

Related Articles:

  • Americans and Affirmative Action: How the Public Sees the Consideration of Race in College Admissions, Hiring — Survey research finds more Americans disapprove than approve of colleges considering race and ethnicity in admissions: half of U.S. adults said they disapprove of selective colleges weighing race and ethnicity to increase diversity, a third approved, and 16% were unsure. In another survey, 82% said colleges should not consider race or ethnicity, versus 17% who said they should.
  • Yale, Princeton and Duke Are Questioned Over Decline in Asian Students — Students for Fair Admissions, which won the Supreme Court case barring race-based admissions preferences, is threatening investigations and lawsuits against Princeton, Yale, and Duke over declines in Asian American enrollment that it says defied expectations. Asian American enrollment fell to 29% from 35% at Duke and to 24% from 30% at Yale.
  • At M.I.T., Black and Latino Enrollment Drops Sharply After Affirmative Action Ban — Massachusetts Institute of Technology's first undergraduate class admitted after the U.S. Supreme Court banned affirmative action saw Black enrollment fall to 5% from 15% and Hispanic and Latino enrollment fall to 11% from 16%. Asian American students rose to 47% from 40%, while white students were 37% compared with 38%.

Monday, October 6, 2025

AI Summary. Ten-year inflation-adjusted bond yields have been broadly stable at about 2%, but market-implied expectations for future real yields have risen, with the 10-year real yield expected ten years ahead now around 3%, about one percentage point higher than two years earlier. Expected real yields are at their highest levels since 2010.

Jesper Rangvid Rangvid's Blog
Date Posted:
Is Database:
Database

The market-implied 10-year real yield expected ten years from now is ~ 3%, its highest level since 2010, and ~2x what it was in early 2023. The real 20-year yield expected in 20 years is even higher, signalling high real yields for many years to come.

Do markets expect future real interest rates to rise?

Core argument: The 10-year real yield expected ten years from now stands at roughly 3%, about one percentage point higher than two years ago and roughly 1.5 percentage points higher than in early 2023.

Ten-year real yields have been broadly stable over the past two years. In late 2023, they stood at about 2%, and they remain close to that level today. By contrast, market-implied future real yields have risen noticeably. The 10-year real yield expected ten years from now is around 3%—one percentage point higher than two years ago and roughly 1.5 percentage points higher than in early 2023. The same increase is evident for the 10-year real yield expected in twenty years.The key point, thus, is not simply that market-implied forward yields exceed current yields—that is typical given an upward-sloping yield curve. Rather, the point is that expected yields have risen by more than current ones, signalling higher future real yields. These shifts represent a substantial rise in long-term real yields expectations. Indeed, the increase is so pronounced that expected real yields are now at their highest levels since 2010, as shown in Figure 2.

Takeaways by Macro Roundup® AI

  1. The 10-year real yield expected ten years from now stands at roughly 3%, about one percentage point higher than two years ago and roughly 1.5 percentage points higher than in early 2023.
  2. Ten-year real yields have been broadly stable near 2% since late 2023, while market-implied forward real yields have risen noticeably, signalling higher future real yields.
  3. Expected real yields are at their highest levels since 2010, with the same increase evident in the 10-year real yield expected twenty years ahead.

Related Articles:

  • Fiscal Risk from Rising Bond Yields — US interest rates fell for roughly 40 years before the pandemic, allowing maturing high-coupon debt to be refinanced at lower rates and keeping federal interest expense manageable even as debt rose. Because Congressional Budget Office projections assume the 10-year Treasury yield drifts down to 3.8% from about 4.1%, the author argues its debt-service projections face upside risk.
  • Our Thoughts on Large US Deficits and Their Impact on Bond Yields — Total borrowing in the US is about average despite high government deficits, as high public borrowing has contributed to the need for higher interest rates that have depressed private sector borrowing. Risk premiums on long-dated debt look about average, with no market signs that economic or debt limits are currently being hit.
  • Global Debt Report 2025 — About 60% of fixed-rate government debt in advanced economies maturing by 2027 — roughly $9tn — was issued before the recent interest-rate tightening, with weighted average yields below 2% versus projected 10-year rates near 3.6%, implying refinancing at nearly double the original cost. France, Spain, the UK, and the US face heightened vulnerability, with maturing debt exceeding 15% of GDP.

AI Summary. US tariff revenue has reached about $30bn per month, roughly $23bn of it new, and could raise about $2tn over a ten-year budget window if broad tariffs survive legal challenges. The authors argue this shift from income taxation to tariffs is hard to defend, estimating efficiency losses equal to about a third of tariff revenue.

Kimberly Clausing and Maurice Obstfeld National Bureau of Economic Research
Date Posted:
Is Database:
Database
Is Important:
Important

Tariffs generated $30B in August 2025; $27B of that was new revenue, and the current regime could raise ~$2T over the next decade. Clausing/Obstfeld’s analysis finds that the welfare losses associated with the tariffs offset over a third of the revenue.

Are tariffs an efficient substitute for income taxes?

Core argument: US tariff revenues reached about $30bn per month by August 2025, $23bn of it new relative to August 2024, and if broad tariffs survive legal challenges they could net roughly $2tn over the coming ten-year budget window.

The 2025 tax and budget legislation sharply reduced expected US income tax revenues. At the same time, a dramatic increase in the use of tariffs has caused tariff revenues to surge, reaching about $30 billion per month by August 2025, $23 billion of which is new revenue relative to August 2024. If such broad tariffs survive legal challenges, tariffs could net about $2 trillion in revenue over the coming ten-year budget window. While insufficient to pay for the recent tax legislation, they mark an important source of government revenue that may prove hard for future US administrations to forgo. [Yet] even at revenue-maximizing rates [~30 to 45%], tariff revenues would be only a fraction of current US federal income tax revenue. On conventional tax policy grounds, the US fiscal shift from income taxation to tariffs is difficult to defend. Tariffs are a particularly inefficient source of revenues that [reduce] gains from trade; the efficiency loss from tariffs at summer 2025 levels is about a third of their revenue yield, not including additional costs associated with rent-seeking. [Taking account of] intermediate goods could raise the deadweight costs of tariffs sharply [and raise projected stagflation].

Takeaways by Macro Roundup® AI

  1. US tariff revenues reached about $30bn per month by August 2025, $23bn of it new relative to August 2024, and if broad tariffs survive legal challenges they could net roughly $2tn over the coming ten-year budget window.
  2. Even at revenue-maximizing rates of roughly 30–45%, tariff revenues would amount to only a fraction of current US federal income tax revenue, and they are insufficient to pay for the 2025 tax and budget legislation that sharply reduced expected income tax revenues.
  3. The efficiency loss from tariffs at summer 2025 levels is about a third of their revenue yield, excluding costs associated with rent-seeking, and taking account of intermediate goods could raise those deadweight costs sharply.

Related Articles:

  • Tariffs Are Generating Meaningful New Revenue — New U.S. tariffs are estimated to generate $1.3tn in net new revenue through the end of the current presidential term and $2.8tn through 2034 before economic effects, about $600bn more than previously enacted tariffs. If a court ruling striking down some tariffs is upheld, the remainder could raise as little as $800bn through 2034.
  • The Fiscal Future — US debt is on an upward trajectory, and the author argues that extraordinary growth, default, money creation, and spending cuts are each implausible or unacceptable, leaving large tax increases. US governments collect about 28% of GDP in taxes versus 35% in the UK and 46% in France, and a value-added tax could close the roughly 4%-of-GDP fiscal gap.
  • Tariffs are Dumb Enough to (Almost) Work — A universal 15% tariff on imported goods could raise roughly $300–$400bn annually, a fraction of what a value-added tax would generate but real revenue given U.S. imports of about $4tn. The author argues tariffs are a poorly designed form of consumption tax that could build political groundwork for replacing them with a more efficient value-added tax.

AI Summary. France's prime minister resigned less than a month after appointment, becoming the shortest-serving premier since 1958 and the third to leave office since snap parliamentary elections left the assembly deeply divided. Markets sold off, with 10-year government borrowing costs rising 0.06 percentage points to 3.57% and the euro falling 0.4% against the dollar.

Ian Johnston, Adrienne Klasa, Ian Smith and Emily Herbert Financial Times
Date Posted:

French Prime Minister Sébastien Lecornu resigned less than a month after his appointment, damaging the ability of France to confront a 5.4% of GDP budget deficit. French 10-year yields rose to 3.57%.

Are markets reacting to political instability in France?

Core argument: French Prime Minister Sébastien Lecornu resigned less than a month after his appointment, the Élysée Palace said, making him the shortest-serving prime minister since the Fifth Republic was established in 1958.

France’s Prime Minister Sébastien Lecornu has resigned less than a month after his appointment, prompting a market sell-off amid concern about dysfunction in the Eurozone’s second-biggest economy. Lecornu, a long-serving ally of French President Emmanuel Macron, submitted his resignation on Monday morning, the Élysée Palace said, making him the shortest-serving prime minister since the Fifth Republic was established in 1958 and heaping pressure on the president. Lecornu was the third prime minister appointed by Macron since snap parliamentary elections in summer 2024, a vote that left the French assembly bitterly divided and has made governing almost impossible. All three have now left office. A fall in government bond prices pushed up the country’s 10-year borrowing costs by 0.06 percentage points to 3.57%. The euro fell 0.4% against the dollar.

Takeaways by Macro Roundup® AI

  1. French Prime Minister Sébastien Lecornu resigned less than a month after his appointment, the Élysée Palace said, making him the shortest-serving prime minister since the Fifth Republic was established in 1958.
  2. France's 10-year borrowing costs rose 0.06 percentage points to 3.57% as government bond prices fell, and the euro dropped 0.4% against the dollar amid concern about dysfunction in the Eurozone's second-biggest economy.
  3. Sébastien Lecornu was the third prime minister appointed by Emmanuel Macron since the summer 2024 snap parliamentary elections, a vote that left the French assembly bitterly divided; all three have now left office.

Related Articles:

  • ECB Yield Caps and TARGET2 Balances — Unlike International Monetary Fund programs, whose conditionality and stigma push borrowing countries to reform quickly, the euro system's internal cross-border payment imbalances carry neither and have persisted far longer. The author argues their persistence is evidence that the European Central Bank's de facto yield caps keep high-debt countries' borrowing costs artificially low.
  • Europe’s Right-Wing Populists Swerve Left on the Economy — Right-wing populist parties across Europe, including the Sweden Democrats, Austria's Freedom Party, Germany's Alternative for Germany and France's National Rally, have shifted from small-government, low-tax, free-trade platforms toward calls for higher welfare benefits, subsidies and protectionism, often alongside pro-market positions.
  • French Companies’ Borrowing Costs Fall Below Government’s As Debt Fears Intensify — Bond yields for French companies including L'Oréal, Airbus and Axa have fallen below those on French government debt of similar maturity, with 10 French firms trading at a negative spread to the sovereign — the most in data going back to 2006, according to Goldman Sachs. Sovereign yields typically set a baseline for corporate bonds.

AI Summary. Global data center power demand is projected to rise about 165% by 2030 versus 2023, with 60% of new US needs requiring roughly 72 GW of new generation, mainly natural gas (60%), plus solar and wind. US grid investment is estimated at $780bn through 2030, skewed toward distribution and transmission.

Eric Sheridan, James Schneider, Kash Rangan, Gabriela Borges, et al. Goldman Sachs
Date Posted:
Is Database:
Database

Goldman estimates overall American power demand is accelerating to a 2.5% CAGR (a growth rate not seen since the 1990s), ~ 40% of which is driven by datacenters. They estimate that 60% of the new generation capacity will be sourced from natural gas.

How much new power generation and grid investment will data centers require?

Core argument: Global datacenter power demand is expected to rise ~165% by 2030 versus 2023, with 60% of future US needs requiring new generation and an estimated 72 GW of incremental capacity.

Meeting [new datacenter electrical] demand will require a major build-out of generation capacity. We expect a ~165% rise in datacenter power demand globally by 2030 vs. 2023. In the US, we estimate that 60% of future needs will require new generation, leading to an estimated 72 GW of incremental capacity - sourced primarily from natural gas (60%), supported by solar (25-30%) and wind (10-15%). In the US, datacenters contribute ~1% CAGR to overall power demand, underscoring a persistent, secular lift as overall US power demand accelerates to a 2.5% CAGR, which has not been seen since the 1990s. Sustaining this growth requires significant investment in the power grid. Since July 2025, our US grid investment expectations have risen by $60bn to $780bn through 2030 (from previous $720bn), with a skew towards distribution infrastructure and faster growth in transmission capex to sustain datacenter supply growth & accommodate the broader increase in power demand.

Takeaways by Macro Roundup® AI

  1. Global datacenter power demand is expected to rise ~165% by 2030 versus 2023, with 60% of future US needs requiring new generation and an estimated 72 GW of incremental capacity.
  2. The estimated 72 GW of incremental US capacity would be sourced primarily from natural gas (60%), supported by solar (25–30%) and wind (10–15%).
  3. US grid investment expectations have risen by $60bn to $780bn through 2030, from $720bn previously, skewed toward distribution infrastructure with faster growth in transmission capex.

Related Articles:

  • The Blob: Capital, China, Chips, Chicago and Chilliwack — Data center construction spending has eclipsed office construction, and the specialized power rates most data centers pay fall short of covering the cost of a new natural gas plant, leaving other customers to bear part of the bill. In the PJM grid region, data center demand accounted for 70% of the increase in electricity costs over the past year.
  • Power Check: Watt’s Going On With The Grid? — Bank of America Global Research estimates US electricity demand will grow at a 2.5% compound annual rate through 2035, up from 0.5% annually over the prior decade, driven by building electrification, data centers, industrial demand, and electric vehicles. Meeting it requires greater utility investment in grid capacity.
  • Electricity 2025 — US electricity demand is projected to grow about 2% annually over 2025-2027, an upward revision from an earlier 1% forecast. A study commissioned by the US Department of Energy found data centre electricity use rose from about 60 TWh in 2014 to 176 TWh in 2023, and could reach 325-580 TWh by 2028.

AI Summary. OpenAI has agreed to buy tens of billions of dollars of chips from AMD, totaling 6 gigawatts of power consumption, under a deal that could give OpenAI a roughly 10% stake in the $270bn chipmaker over time. OpenAI executives estimate 1GW of capacity costs about $50bn to bring online.

George Hammond and Michael Acton Financial Times
Date Posted:

OpenAI will buy chips from AMD with power demand of 6GW, roughly equal to Singapore’s mean demand. Over the next five years, OpenAI has committed to 23GW of new capacity, which will cost over $1T to bring online.

Can a chipmaker grant an AI buyer an ownership stake?

Core argument: OpenAI agreed to buy tens of billions of dollars' worth of AMD chips and could take a roughly 10% stake in the $270bn chipmaker over time, with no total dollar figure disclosed.

OpenAI has agreed to buy tens of billions of dollars’ worth of chips from AMD as part of a deal that could also see the ChatGPT maker take a roughly 10% stake in the $270bn chipmaker over time. The San Francisco-based artificial intelligence start-up said it had agreed to purchase processors with a total power consumption of 6 gigawatts, roughly equivalent to Singapore’s average demand. The companies did not put a total dollar figure on the transaction, but OpenAI executives estimate that 1GW of capacity costs about $50bn to bring online, with two-thirds of that spent on chips and the infrastructure to support them. The deal comes just a fortnight after AMD’s rival Nvidia announced it planned to invest $100bn in OpenAI, with the two companies pledging to deploy 10GW of new data centre capacity.

Takeaways by Macro Roundup® AI

  1. OpenAI agreed to buy tens of billions of dollars' worth of AMD chips and could take a roughly 10% stake in the $270bn chipmaker over time, with no total dollar figure disclosed.
  2. The AMD processors OpenAI agreed to purchase carry a total power consumption of 6GW, roughly equivalent to Singapore's average demand.
  3. OpenAI executives estimate that 1GW of capacity costs about $50bn to bring online, with two-thirds spent on chips and supporting infrastructure.

Related Articles:

  • The Blob: Capital, China, Chips, Chicago and Chilliwack — Data center construction spending has eclipsed office construction, and the specialized power rates most data centers pay fall short of covering the cost of a new natural gas plant, leaving other customers to bear part of the bill. In the PJM grid region, data center demand accounted for 70% of the increase in electricity costs over the past year.
  • Data Centers: More Money, Fewer Buyers — Data center developers can earn 12–15% returns on new construction while stabilized hyperscale campuses trade at 4–5% capitalization rates, drawing capital into development. Because long, flat leases to Amazon, Microsoft and Google make finished assets behave like annuities, few buyers exist, so developers turn to securitization instead of selling.
  • The Kanye/Data Center Crossover — Data center capital spending differs from fiber and railroad investment because most of the cost sits in processors that are replaced every 3-4 years rather than in long-lived shells, power, and cooling connections. That short replacement cycle requires returns high enough to exceed a risk-adjusted cost of capital of 12-14% before the hardware must be replaced.

AI Summary. South Korea's fertility rate fell below 1.7 in 1985 and its share of population aged 20-40 has fallen 20% since 2000, yet it ranked first in Bloomberg's Innovation Index and twelfth in a global soft power ranking. The author argues low fertility does not inevitably cause technological and cultural stagnation.

Adair Turner Project Syndicate
Date Posted:

Adair Turner, citing South Korea’s continued ability to innovate despite plummeting fertility, argues that population decline doesn’t necessarily imply a slowdown in growth; only a small fraction of the population drives technological advances.

Does low fertility inevitably cause technological and cultural stagnation?

Core argument: South Korea's fertility rate fell below 1.7 in 1985 and the share of its population aged 20–40 has declined by 20% since 2000, yet the country ranked first in Bloomberg's 2021 Innovation Index.

South Korea is a case in point. Its fertility rate fell below 1.7 in 1985, and the share of its population aged 20-40 has declined by 20% since 2000, yet the country still ranked first in Bloomberg’s 2021 Innovation Index. Earlier this year, it ranked twelfth in the Brand Finance Global Soft Power Index 2025, a “meteoric ascent” attributed to its “dominance in arts and entertainment” and the worldwide appeal of its cultural exports, whether K-pop, K-dramas like Squid Game, or K-beauty products. At just 0.8, South Korea’s fertility rate may eventually sap its innovative vitality. But the idea that low fertility rates inevitably result in technological and cultural stagnation is supported by neither logic nor empirical evidence.

Takeaways by Macro Roundup® AI

  1. South Korea's fertility rate fell below 1.7 in 1985 and the share of its population aged 20–40 has declined by 20% since 2000, yet the country ranked first in Bloomberg's 2021 Innovation Index.
  2. South Korea ranked twelfth in the Brand Finance Global Soft Power Index 2025, an ascent attributed to its dominance in arts and entertainment and the worldwide appeal of cultural exports including K-pop, K-dramas such as Squid Game, and K-beauty products.
  3. Neither logic nor empirical evidence supports the claim that low fertility rates inevitably produce technological and cultural stagnation, though South Korea's fertility rate of 0.8 may eventually sap its innovative vitality.

Related Articles:

  • The End of Children — South Korea's fertility rate of 0.7 is the lowest of any nation and may be the lowest in recorded history; if sustained, each generation would be about a third the size of the one before, with every 100 people of childbearing age producing roughly 12 grandchildren. Annual births fell from 1m in 1970 to 230,000 in 2023.
  • Depopulation Globally and in the Asia-Pacific: The Shape of Things to Come — Population aging and fertility below replacement levels are connected to a revolution in family structure in which families grow smaller, fewer people marry, and voluntary childlessness becomes common across countries. As a result, families become less able to bear the demands that may be placed on them.
  • World Depopulation: Prospects and Implications — The US is projected to account for a growing share of the rich world's labor force, youth, and highly educated talent, with continuing population and labor force growth and only moderate aging through 2050. Its fertility, though below replacement, exceeds that of East Asian and nearly all European countries, and sustained legal immigration is essential to this demographic advantage.