Edward Conard

Top Ten New York Times Bestselling Author

  • “Unintended Consequences provides a provocative interpretation of the causes of the global financial crisis and the policies needed to return to rapid growth. Whether you agree or not, this analysis is well worth reading.” - Nouriel Roubini, New York University; Chairman, Roubini Global Economics
  • “…a comprehensive explanation of the modern economy.” - Julian Robertson, Founder, Tiger Management
  • “…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
  • “Unintended Consequences offers deep and well-argued analyses on almost every issue.” - The New York Times
  • “…serious thinking for serious thinkers. …a thought-provoking blueprint for growing middle- and working-class incomes.” - Mitt Romney, former Governor of Massachusetts
  • “…serious thinking for serious thinkers. …a thought-provoking blueprint for growing middle- and working-class incomes.” - Mitt Romney, former Governor of Massachusetts
  • “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
  • “…a fresh argument for the productive value of inequality.” - David Autor, Professor of Economics, Massachusetts Institute of Technology
  • “…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 must-read for serious students of economic policy.” - Glenn Hubbard, Dean, Columbia Business School, and former Chairman of the Council of Economic Advisers
  • “…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
Upside of Inequality Oxford Unintended Consequences

Macro Roundup


Monday, June 8, 2026

AI Summary. The authors estimate that AI coding tools raise developers' task-level productivity by about 40% with autocomplete, 140% including synchronous agents, and 180% with asynchronous agents, with larger effects for less active developers. Gains attenuate up the production chain: synchronous agents raise lines of code 741% and pull requests 65% but releases only 20%.

Mert Demirer, Leon Musolff and Liyuan Yang National Bureau of Economic Research
Date Posted:

Event studies indicate that adoption of AI coding tools raised “commits” (saved code updates) ~180%, but releases by only ~30%. Large upstream productivity gains thus far show only modest translation into shipped and used software.

How much do AI coding tools raise code writing versus shipping?

Core argument: AI coding tools raise developer task-level productivity measured by commits by roughly 40% with autocomplete, about 140% cumulatively with synchronous agents, and 180% with asynchronous agents through agent-authored commits.

Our results are summarized in Figure 1. We find that AI coding tools substantially increase developer task-level productivity, with effects that grow across successive generations of tools. Focusing on the number of commits [saved code changes], autocomplete increases task-level productivity by approximately 40%; the cumulative effect including sync agents rises to approximately 140%; and additional use of async agents raises it to 180% through agent authored commits. These effects are larger for less active developers. The results point to a substantial productivity shift in software development with the rise of agentic AI tools. [Yet] our estimates across the layers of the production hierarchy reveal a clear pattern of attenuation. After adopting AI tools, developers write substantially more code and work on more features, but this translates into much smaller increases in shipped software. Sync agents lead to a 741% increase in lines of code and a 65% increase in pull requests, yet releases rise by only 20%. Autocomplete shows the same pattern: a 228% effect on lines of code attenuates to 36% on commits and just 10% on releases. AI tools appear to relax upstream constraints in producing and publishing software, but developers must still test, polish, and iterate toward market fit, and these higher-level tasks may remain constrained [once] AI lowers the cost of writing code.

Takeaways by Macro Roundup® AI

  1. AI coding tools raise developer task-level productivity measured by commits by roughly 40% with autocomplete, about 140% cumulatively with synchronous agents, and 180% with asynchronous agents through agent-authored commits.
  2. Productivity gains attenuate up the production hierarchy: synchronous agents increase lines of code by 741% and pull requests by 65%, yet releases rise only 20%.
  3. Autocomplete shows the same attenuation pattern, with a 228% effect on lines of code falling to 36% on commits and 10% on releases, and estimated effects are larger for less active developers.

Related Articles:

  • Commoditization, Orchestration, and the New AI Stack — The author argues that as incremental gains from larger AI training runs shrink, the return on additional computing power falls, weakening the case for the training spending that dominates data center demand. Survivors of model commoditization will be those owning the orchestration layer or the applications where outcomes are priced.
  • America Is Experiencing A Productivity Miracle — America's post-2019 productivity acceleration came largely from professional services and management—about 10% of output and heavy users rather than producers of new technology—and from oil and gas. The information industry grew productivity around 6% annually, no faster than its 2000-19 average, with its output share steady at 5.3–5.5%.
  • Attention (And Money) Is All You Need: Why Universities Are Struggling to Keep AI Talent — The annual compensation gap between top industry and academic artificial intelligence researchers widened from roughly $294,000 to $1.5m, coinciding with accelerating academic out-migration toward large incumbent technology firms. Researchers who move to industry cut paper production by about 63% while increasing patent filings nearly sixfold, consistent with a shift from open science toward proprietary innovation.

AI Summary. Measuring developer productivity by software lines of code captures output rather than value, and AI inflates the metric by generating boilerplate, wrappers, tests, and speculative code at near-zero cost. More code also creates liabilities through added maintenance, complexity, attack surface, and debugging load.

Paul Kedrosky Applied Complexity
Date Posted:

Anthropic data shows Mythos drove an 8-fold increase in mean lines of code per developer. Kedrosky cautions against measuring output over value: “The best engineers create value by writing less code, not more.”

Do lines of code measure developer productivity?

Core argument: Kedrosky argues an 8x increase in software lines of code generally corresponds to a far smaller increase in actual engineering throughput, since AI agents generate boilerplate, wrappers, tests, and what-if code at near-zero cost.

[Kedrosky warns a new Anthropic paper based on internal data on increased developer productivity, as measured by software lines of code (SLOC) per developer measures output, not value]. Ten thousand lines of generated code are often less valuable than deleting a thousand lines of technical debt. SLOC as a measure rewards quantity, not usefulness. AI systematically inflates the metric. Agents generate boilerplate, wrappers, tests, and what-if code at near-zero cost. An 8x increase in SLOC generally corresponds to a far smaller increase in actual engineering throughput. Code volume creates liabilities. More code means more maintenance, complexity, attack surface, and debugging load. The best engineers create value by writing less code, not more. Most evidence is about implementation, not automating research. The strongest results involve coding, experiment execution, and debugging. Recursive self-improvement depends on solving an unsolved problem.

Takeaways by Macro Roundup® AI

  1. Kedrosky argues an 8x increase in software lines of code generally corresponds to a far smaller increase in actual engineering throughput, since AI agents generate boilerplate, wrappers, tests, and what-if code at near-zero cost.
  2. Kedrosky warns that an Anthropic paper using internal data measures developer productivity by lines of code per developer, a metric that captures output rather than value and rewards quantity over usefulness.
  3. Kedrosky notes that more code brings more maintenance, complexity, attack surface, and debugging load, and that the strongest evidence covers coding, experiment execution, and debugging rather than automated research, leaving recursive self-improvement dependent on an unsolved problem.

Related Articles:

  • Claude Code is the Inflection Point — Claude Code authors 4% of public GitHub commits, and the authors project it will reach 20%+ of daily commits by the end of 2026 and drive revenue growth that lets Anthropic outgrow OpenAI. They argue falling costs and rising quality of AI tokens have inverted the coder-tool relationship.
  • We’re Using So Much AI That Computing Firepower Is Running Out — Surging AI demand is straining computing capacity: Anthropic's Claude programming interface recorded 98.95% uptime over a 90-day span, below the 99.99% software companies typically promise, as its annualized revenue run rate climbed from $9bn to $14bn to $30bn. The company has capped how much usage customers can consume during weekday peak morning hours.
  • Commoditization, Orchestration, and the New AI Stack — The author argues that as incremental gains from larger AI training runs shrink, the return on additional computing power falls, weakening the case for the training spending that dominates data center demand. Survivors of model commoditization will be those owning the orchestration layer or the applications where outcomes are priced.

AI Summary. Revenue per employee is rising for the largest seven US technology companies and falling for small-cap companies, while profit margins are rising for those seven and flat for the rest of the S&P 500. There are not yet signs that AI is boosting revenues or profit margins outside that group.

Torsten Sløk Apollo
Date Posted:
Is Database:
Database

Net the Mag 7, revenue per employee for the S&P 500 has been flat and for smaller firms, falling. Sløk points to the higher cost of capital, and notes “there are not yet signs that AI is boosting revenues or profit margins outside the Mag 7.”

Are AI gains spreading beyond the largest technology companies?

Core argument: Revenue per employee is rising for the Magnificent 7 while falling for small-cap companies, with no sign yet that AI is boosting revenues or profit margins outside the Magnificent 7.

Revenue per employee is rising for the Mag 7 and falling for small-cap companies. Profit margins are rising for the Mag 7 but flat for the S&P 493. The bottom line is that Fed hikes and higher all-in yields continue to bite for companies with weaker credit fundamentals, and there are not yet signs that AI is boosting revenues or profit margins outside the Mag 7.

Takeaways by Macro Roundup® AI

  1. Revenue per employee is rising for the Magnificent 7 while falling for small-cap companies, with no sign yet that AI is boosting revenues or profit margins outside the Magnificent 7.
  2. Profit margins are rising for the Magnificent 7 but flat for the S&P 493.
  3. Fed rate hikes and higher all-in yields continue to bite for companies with weaker credit fundamentals.

Related Articles:

  • The Record Divide Between Corporate Profits and Worker Pay — Labor's share of gross domestic income fell to 51%, the lowest since records began in 1947, while profits' share rose to 12.1%, the highest since 1950. Inflation-adjusted hourly wages are up 3% since the end of 2019, compared with a 50% rise in profits.
  • Competitive Advantage Period: The Neglected Value Driver — Data indicate persistence in returns on invested capital has increased since 2000, with large companies the main beneficiaries; the information technology sector's aggregate return was 19% versus a 10.8% median from 1970 to 2024. Of about 13,800 U.S. public firms listing from 1976 to 2019, 18% survived.
  • Have We Entered an Era of High Productivity Growth? — Labor productivity has grown solidly while the San Francisco Federal Reserve's total factor productivity index, which adjusts for the equipment and technology workers use, has grown modestly, suggesting gains so far reflect better tools rather than broader advances. A regime-switching model puts the probability of a high-productivity regime at 57% using labor productivity and 21% using total factor productivity.

AI Summary. U.S. job gains have averaged 188,000 a month over the past three months while the unemployment rate has held steady at 4.3%, challenging the view that an immigration crackdown lowered the pace of hiring needed to keep unemployment stable.

Matt Grossman The Wall Street Journal
Date Posted:

The American economy added ~500,000 jobs btw March and May, the strongest stretch of job creation in two years. The unemployment rate was steady at 4.3%, suggesting the breakeven rate may not have fallen.

Is the job growth needed to hold unemployment steady higher than thought?

Core argument: U.S. job gains averaged 188,000 a month over the past three months while the unemployment rate held steady at 4.3%, suggesting the labor market's neutral gear may not be as low as economists had thought.

The resurgence poses a puzzle for economists. Many came to believe that the immigration crackdown had pushed the economy into a new, more sedate equilibrium, where even a trickle of new jobs each month would be enough to employ an American workforce whose growth had slowed substantially. The country, some economists proposed, simply didn’t need to create hundreds of thousands of new jobs a month anymore to keep the economy stable, a trend called a falling break-even rate of job creation. If that downshift were happening, slower job creation wouldn’t lead to more unemployment—but weaker growth and a less dynamic economy would likely follow. More recent data is calling that conclusion into question. Job gains have averaged 188,000 each month over the past three months. At the same time, however, the unemployment rate hasn’t really fallen. Instead, it has been steady at 4.3%—signaling that the labor market’s neutral gear might not be as low as economists had thought.

Takeaways by Macro Roundup® AI

  1. U.S. job gains averaged 188,000 a month over the past three months while the unemployment rate held steady at 4.3%, suggesting the labor market's neutral gear may not be as low as economists had thought.
  2. Many economists had concluded that the immigration crackdown lowered the break-even rate of job creation, meaning slower hiring would suffice to employ a workforce whose growth had substantially slowed.
  3. Under the falling break-even thesis, slower job creation would not raise unemployment, though weaker growth and a less dynamic economy would likely follow.

Related Articles:

  • Labor Force Growth, Breakeven Employment, and Potential GDP Growth — Slowing net immigration could cut US labor force growth so much that the breakeven pace of job creation falls to nearly zero — under 10,000 jobs a month — far below any point in the past 65 years. Monthly payroll declines as large as 100,000 would then not be unusual even with output at potential.
  • The U.S. Job Market is (Still) Inflationary — Prices for locally produced U.S. services are rising about 1–1.5 percentage points faster than in the years just before the pandemic, leaving underlying inflation as far from the Federal Reserve's goals as it was three years ago. Prices for manufacturing components and data-center-related loaded circuit boards have also risen.
  • A Changing Job Market Leans Against Men — The share of men aged 16 and over who are employed has fallen from a 1990s average of 70.9% to 64.1%, partly reflecting an aging population, while women's employment share has held up better at 54.5%. Healthcare and social assistance, where women outnumber men three to one, drove recent private-sector job gains.

AI Summary. The US Department of Agriculture confirmed two more detections of New World screwworm in Texas, bringing the total to four and raising concern the parasitic fly is spreading in the largest cattle-producing state. The detections add pressure on a US beef industry already facing tight supply, with the domestic herd at a 75-year low.

Ilena Peng Bloomberg
Date Posted:

The US Department of Agriculture confirmed two new cases of New World screwworm in Texas, bringing the total number of cases to four. The US’s domestic cattle herd is at a 75-year low.

Is a parasitic pest spreading through the largest cattle-producing state?

Core argument: The US Department of Agriculture confirmed two additional New World screwworm cases in Texas, bringing total detections to four in the biggest cattle-producing state and raising concerns that the parasite is spreading.

The US Department of Agriculture confirmed an additional two cases of New World screwworm in Texas, bringing the total number of detections to four and raising concerns that the parasite is spreading in the biggest cattle-producing state. The parasitic fly was detected in a calf in La Salle County and a dog in Andrews County; early reports indicate the dog was recently in Mexico. [The screwworm brings] more worries for a US beef industry already struggling with supply tightness as the domestic herd has dwindled to a 75-year low.

Takeaways by Macro Roundup® AI

  1. The US Department of Agriculture confirmed two additional New World screwworm cases in Texas, bringing total detections to four in the biggest cattle-producing state and raising concerns that the parasite is spreading.
  2. The screwworm was detected in a calf in La Salle County and a dog in Andrews County, with early reports indicating the dog had recently been in Mexico.
  3. The screwworm detections add to pressures on a US beef industry already facing supply tightness, with the domestic herd down to a 75-year low.

Related Articles:

  • Flesh-Eating Pest Confirmed in U.S. Cattle — The New World screwworm, a fly whose larvae feed on the flesh of livestock and other warm-blooded animals, has been detected in a Texas calf — the first U.S. cattle case since its eradication in the 1960s — prompting a quarantine zone and release of millions of sterile flies.

AI Summary. Failures in India's medical entrance exam and other testing problems have sparked a viral satirical online campaign demanding the education minister's resignation, with its founder calling the episode evidence of government incompetence. Under-30s make up about half of India's 1.4bn people, and a university report found nearly 40% of graduates aged 15-25 and 20% of those aged 25-29 are jobless.

Andres Schipani, Jyotsna Singh and Michael Stott Financial Times
Date Posted:

There is widespread outrage in India after results of the nation-wide medical entrance exam were voided because answers were leaked prior to the test. 2.2mm applicants were competing for 130,000 slots.

Are exam failures and youth joblessness fueling discontent in India?

Core argument: Nearly 40% of Indian graduates aged 15–25 and 20% of those aged 25–29 are jobless, according to a March report by Azim Premji University in Bengaluru.

It has also helped spark a rare challenge to Modi in the form of the Cockroach Janta Party, a viral new satirical online campaign whose creator has called for a mass demonstration to demand the education minister’s resignation. Party founder Abhijeet Dipke, who is based in Boston but plans to return to New Delhi on Saturday, has said the medical entrance exam fiasco along with other leaks, marking mistakes and system failures demonstrate the “sheer incompetence and failure of the government.” Under-30s make up about half of India’s 1.4bn people — the world’s largest youth population — and according to a March report by Azim Premji University in Bengaluru, nearly 40% of graduates aged 15-25 and 20% of those aged 25-29 are jobless.

Takeaways by Macro Roundup® AI

  1. Nearly 40% of Indian graduates aged 15–25 and 20% of those aged 25–29 are jobless, according to a March report by Azim Premji University in Bengaluru.
  2. Under-30s make up about half of India's 1.4bn people, the world's largest youth population.
  3. The Cockroach Janta Party, a satirical online campaign founded by Boston-based Abhijeet Dipke, has called for a mass demonstration demanding the education minister's resignation, with Dipke citing the medical entrance exam fiasco, leaks, marking mistakes and system failures as government incompetence.

Related Articles:

  • India’s Population Will Soon Be Falling—Probably Quite Fast — India's fertility has fallen further and faster than expected, with the total fertility rate at 1.3 in Tamil Nadu and West Bengal and 1.5 across urban India, raising concern that the country is following China, whose population has been shrinking. Poorer northern states appear to be converging with wealthier ones.
  • India’s Statistics Chief Counts GDP Data Reform As Success — India's statistics ministry says its reformed GDP methodology, which treats inflation differently and draws on additional data sources, showed growth was higher than previously reported in two of the past three fiscal years, countering claims that official figures flattered growth. Economists question the manufacturing and import inflation adjustments.
  • India’s 20 Years of GDP Misestimation: New Evidence — The authors argue that India's 2015 GDP methodology used inappropriate deflators and indicators, breaking the historical link between reported GDP and exports, credit, taxes, electricity use, and industrial production. They estimate the economy grew 4–4½% on average from 2011 to 2023 rather than the reported 6%.

Tuesday, June 9, 2026

AI Summary. An analysis of U.S. counties estimates that where wages grew faster than prices, voters supported the incumbent party more, while counties with larger real wage losses than average shifted more toward the opposition. A one-standard-deviation larger real wage loss is associated with shifts of roughly 0.10 and 0.19 percentage points in vote share and margin.

Juan Felipe Riaño and Francesco Trebbi National Bureau of Economic Research
Date Posted:

Given state fixed effects, demographics, and local inflation, a county whose real wage loss was 1SD > that of the mean county shifted its Presidential vote in 2024 towards the GOP by ~ 0.1pp in vote share and ~0.19pp in margin. Inflation per se was not “punished.”

Do voters punish incumbents for real wage losses rather than inflation?

Core argument: Using total local inflation to deflate wages, the estimated coefficient on real wage growth is −0.052 for county vote-share change and −0.048 for margin change, and is negative in every column of Table 3.

Table 3 presents estimates that decompose the economic environment into the wage–price gap G(k) and nominal price growth ∆P(k). Across vote-share and margin outcomes, the coefficient on real wage growth is negative in every column. With total [local] inflation used as the deflator of wages, the estimates are −0.052 for vote-share change and −0.048 for margin change. This sign is consistent with a basic retrospective accountability intuition applied to real income: counties where wages outpaced prices more strongly (higher real wage growth) tended to support the incumbent party more and shift less toward the opposition party. A county whose standardized real wage loss is one standard deviation larger than that of the mean county is predicted to shift toward the Republican ticket by 0.052 standard deviations in vote share change and 0.048 standard deviations in margin change (total-inflation column), or approximately 0.10 and 0.19 percentage points, respectively. Thus, among counties that lost real purchasing power on average, those that lost more relative to the mean shifted more strongly toward the Republican party, consistent with retrospective “punish-the-incumbent” voting.

Takeaways by Macro Roundup® AI

  1. Using total local inflation to deflate wages, the estimated coefficient on real wage growth is −0.052 for county vote-share change and −0.048 for margin change, and is negative in every column of Table 3.
  2. A county whose standardized real wage loss exceeds the mean county's by one standard deviation is predicted to shift toward the Republican ticket by 0.052 standard deviations in vote-share change and 0.048 in margin change, roughly 0.10 and 0.19 percentage points.
  3. Counties where wages outpaced prices more strongly tended to support the incumbent party more and shift less toward the opposition party, a pattern consistent with retrospective "punish-the-incumbent" voting on real income.

Related Articles:

  • Real Wages Start To Shrink In Developed Countries — US annual inflation of 3.8% outpaced average hourly earnings growth of 3.6%, leaving prices rising faster than pay for the first time in two years, while Eurozone workers face a fresh energy-driven setback. JPMorgan attributes the real wage decline to Middle East conflict and expects growth to resume if energy prices retreat.
  • US Consumer Sentiment Slides to Record Low on Price Concerns — The University of Michigan's consumer sentiment index fell 5 points to 44.8, a record low and weaker than all economist projections in a Bloomberg survey. Consumers expect prices to rise 3.9% annually over the next five to 10 years, up from 3.5%, and 4.8% over the coming year.
  • The Record Divide Between Corporate Profits and Worker Pay — Labor's share of gross domestic income fell to 51%, the lowest since records began in 1947, while profits' share rose to 12.1%, the highest since 1950. Inflation-adjusted hourly wages are up 3% since the end of 2019, compared with a 50% rise in profits.

AI Summary. The post-World War II productivity boom, which doubled real wages and living standards in a generation, came largely from technologies that already existed decades earlier rather than radical new inventions. It was followed by sluggish productivity growth from 1973 to 1995, then an information-technology boom that lasted roughly a decade before measured productivity growth slowed again.

Paul Krugman Krugman Wonks Out
Date Posted:
Is Database:
Database

Krugman highlights the “Solow Paradox.” The highest TFP growth in the postwar US took place btw 1948–73, “an era of continuity” based on pre-existing technologies, whereas later periods with much greater technological ferment saw far less growth.

Do productivity booms come from existing technologies rather than new inventions?

Core argument: The quarter-century postwar boom in total factor productivity doubled real wages and living standards in a generation, yet the U.S. economy in 1963 was still achieving rapid gains from a technological suite that existed in 1913, what Peter Drucker in 1969 called an "age of continuity.".

There was an extraordinary, quarter-century-long boom in total factor productivity after World War II. Real wages and living standards doubled in a generation. It’s reasonable to think that this era of unprecedented economic progress must have been driven by radical new technologies that transformed the economy. Remarkably, however, it wasn’t. As the great management theorist Peter Drucker observed in 1969, the postwar economic miracle took place in an “age of continuity.” The U.S. economy in, say, 1963 was still achieving rapid productivity gains from a technological suite that already existed in 1913. This was followed by two-plus decades of very sluggish progress, the era of the “Solow paradox” — a combination of highly visible new technologies but disappointing economic payoff. Then came the IT boom that began circa 1995 but petered out around a decade later. Despite widespread perceptions that the productivity drought of 1973-1995 was over, the tech boom only lasted around a decade, after which productivity growth, at least as measured, once again slowed to a crawl.

Takeaways by Macro Roundup® AI

  1. The quarter-century postwar boom in total factor productivity doubled real wages and living standards in a generation, yet the U.S. economy in 1963 was still achieving rapid gains from a technological suite that existed in 1913, what Peter Drucker in 1969 called an "age of continuity.".
  2. The 1973–1995 "Solow paradox" era combined highly visible new technologies with disappointing economic payoff, producing two-plus decades of sluggish productivity progress.
  3. The IT boom beginning around 1995 lasted roughly a decade before measured productivity growth slowed to a crawl again, despite widespread perceptions that the earlier productivity drought had ended.

Related Articles:

  • Have We Entered an Era of High Productivity Growth? — Labor productivity has grown solidly while the San Francisco Federal Reserve's total factor productivity index, which adjusts for the equipment and technology workers use, has grown modestly, suggesting gains so far reflect better tools rather than broader advances. A regime-switching model puts the probability of a high-productivity regime at 57% using labor productivity and 21% using total factor productivity.
  • America Is Experiencing A Productivity Miracle — America's post-2019 productivity acceleration came largely from professional services and management—about 10% of output and heavy users rather than producers of new technology—and from oil and gas. The information industry grew productivity around 6% annually, no faster than its 2000-19 average, with its output share steady at 5.3–5.5%.
  • Attention (And Money) Is All You Need: Why Universities Are Struggling to Keep AI Talent — The annual compensation gap between top industry and academic artificial intelligence researchers widened from roughly $294,000 to $1.5m, coinciding with accelerating academic out-migration toward large incumbent technology firms. Researchers who move to industry cut paper production by about 63% while increasing patent filings nearly sixfold, consistent with a shift from open science toward proprietary innovation.

AI Summary. Alphabet, Microsoft, Amazon, Meta and Oracle are projecting $700bn in capital spending in 2026, with Bain & Company estimating the tech industry must generate $2tn in annual AI revenue to justify current data centre spending trends. Electrical equipment suppliers describe AI data centres as a core growth driver.

Clara Murray Financial Times
Date Posted:

According to an FT analysis, more than 200 public firms linked to the AI buildout outperformed the MSCI, which had risen ~21% y/y as of June 9. The firms are largely industrial companies, ranging from Caterpillar, to glass maker Corning, to GE Vernova.

Are electrical equipment suppliers benefiting from AI data center spending?

Core argument: Alphabet, Microsoft, Amazon, Meta and Oracle are projecting $700bn in combined capital spending in 2026 alone, raising questions over whether the investment wave can last if demand from the small number of companies driving the build-out slows.

Alphabet, Microsoft, Amazon, Meta and Oracle [are] projecting $700bn in capital spending in 2026 alone. Questions are being raised over whether the wave of investment can last — particularly if demand from the relatively small number of companies driving the build-out slows. Bain & Company has estimated that the tech industry must generate $2tn in AI revenue per year to justify current data centre spending trends. Schneider Electric told the FT AI data centres were a “core growth driver” for the company, while Siemens’ head of data centres Ciaran Flanagan said AI demand was a “long-term, structural growth trend rather than a short-term cycle,” adding that its data centre business grew 40% in 2025.

Takeaways by Macro Roundup® AI

  1. Alphabet, Microsoft, Amazon, Meta and Oracle are projecting $700bn in combined capital spending in 2026 alone, raising questions over whether the investment wave can last if demand from the small number of companies driving the build-out slows.
  2. Bain & Company estimates the tech industry must generate $2tn in AI revenue per year to justify current data centre spending trends.
  3. Siemens' head of data centres Ciaran Flanagan called AI demand a "long-term, structural growth trend rather than a short-term cycle," noting Siemens' data centre business grew 40% in 2025, while Schneider Electric called AI data centres a "core growth driver.".

Related Articles:

  • How Much Is Donald Trump Costing America’s Economy? — Excluding artificial-intelligence-related categories, US non-residential fixed investment has contracted at roughly a 3% annual rate over the past year, versus average growth above 5% in the prior decade, running about $130bn below trend and cutting GDP growth by roughly 0.4 percentage points. Credit remains plentiful, with tight corporate bond spreads.
  • The Macro Implications of the AI Capex Boom — Artificial intelligence capital spending is estimated to add roughly 1.4 percentage points to US growth in one year and about 1.5 points the next, comparable to business investment's contribution during the tech bubble. A material part of that estimate reflects profits at Nvidia and a few other chipmakers that are unlikely to be recycled back into the economy.
  • Capital Is Making a Comeback — Tangible investment in property, factories and equipment fell from 12.5% to 8.5% of private gross domestic product, while intangible investment in intellectual property, software and process knowledge rose from about 11.5% to 16.75%. The author argues demand for tangible capital investment will now have to rise.

AI Summary. The author argues that inventory drawdowns temporarily masked the size of an oil supply shortfall, and that as stockpiles deplete the effective shortfall rises toward its full extent. On this view prices could climb toward the top of a projected 60–80% range but should not set new highs, since inventory depletion leaves the underlying supply shock unchanged.

Robin Brooks Robin Brooks Substack
Date Posted:

Since inventories cushioned the oil shock more than Brooks first assumed, the initial supply shortfall was smaller. As stocks deplete, prices may move from the bottom to the top of his 60–80% range, but with the underlying shock unchanged, not further.

Does inventory depletion push oil prices to new highs?

Core argument: Oil prices sit at the bottom of the author's 60–80% range, which could rise toward the top end as inventories deplete, though the author sees no reason for new highs.

All that’s really going on is that the initial supply shortfall was less than the ten million barrels I assumed at the time (due to inventories being run down), while it’s now - as stockpiles deplete - rising towards that. Nothing about the upper bound for oil prices has changed and of course it shouldn’t. The underlying supply shock is the same, it’s just that a temporary source of relief is ending. We’re currently at the bottom of my 60 to 80% range. I can see how we rise towards the top end of this range as inventories fall away, but I see no reason for us to make new highs. Inventory depletion changes nothing about the underlying shock and so there shouldn’t be new highs.

Takeaways by Macro Roundup® AI

  1. Oil prices sit at the bottom of the author's 60–80% range, which could rise toward the top end as inventories deplete, though the author sees no reason for new highs.
  2. The initial supply shortfall was smaller than the assumed ten million barrels because inventories were run down, and it is now rising toward that figure as stockpiles deplete.
  3. Inventory depletion leaves the underlying oil supply shock unchanged and ends only a temporary source of relief, so the upper bound for oil prices is unchanged.

Related Articles:

  • Global Market Views: Shortages — A ceasefire with Iran and reduced risk of extreme downside outcomes let markets compress risk premia, with US equities, high-carry and commodity currencies, and emerging markets recovering even as oil prices and yields stayed elevated. Shortages in commodities and the artificial intelligence supply chain are driving capital flows, pointing to higher volatility alongside further equity gains.
  • Shifting Conclusions From The Crises Of The Past 30 Years — Past financial crises taught countries to accumulate US Treasuries, while war-related financial sanctions taught them to accumulate gold; the author argues a Strait of Hormuz crisis will push countries to stockpile industrial commodities, energy and food while reinvesting in their own power grids. Funding this would drain liquidity from global financial markets for years.
  • You Can’t Print Molecules — Rebuilding after a disruption to a key shipping strait would require governments to finance defense, strategic reserves, domestic energy production, and infrastructure at the same time that inflation expectations are raising funding costs. The author argues capital will rotate out of asset-light sectors that thrived on open sea lanes and cheap energy into asset-heavy physical ones.

AI Summary. The author argues the US is maintaining an illusion of abundance by drawing down oil inventories rather than adding supply: the Strategic Petroleum Reserve has fallen from above 415m to roughly 357m barrels and Cushing storage from 33m to about 24.5m, with gasoline stocks drawing for 15 straight weeks.

Jeff Currie Carlyle
Date Posted:

Currie argues that the increase in US energy exports is suppressing “the very price signal that would trigger the investment response the market needs” to increase supply. Without an opening of Hormuz in the near term, he forecast a non-linear repricing.

Is US oil abundance being sustained by drawing down inventories?

Core argument: The Strategic Petroleum Reserve has fallen from above 415 million barrels in March to roughly 357 million, on a trajectory reaching operational minimums by early autumn.

The United States is cashing out its insurance policy in order to maintain the illusion of abundance. The SPR has fallen from above 415 million barrels in March to approximately 357 million today, on a trajectory reaching operational minimums by early autumn. Cushing[2] has declined from 33 to roughly 24.5 million barrels, weeks from nearing the floor below which the futures settlement mechanism begins to break down. Commercial crude inventories have erased their entire 2026 build in five weeks. Gasoline stocks have drawn for fifteen consecutive weeks. Diesel stocks are near critical levels before the summer driving season. This is not supply responding to price. It is inventory responding to price. And inventory, unlike production, has a floor.

Takeaways by Macro Roundup® AI

  1. The Strategic Petroleum Reserve has fallen from above 415 million barrels in March to roughly 357 million, on a trajectory reaching operational minimums by early autumn.
  2. Cushing inventories have declined from 33 million to roughly 24.5 million barrels, weeks from nearing the floor below which the futures settlement mechanism begins to break down.
  3. Commercial crude inventories have erased their entire 2026 build in five weeks, gasoline stocks have drawn for fifteen consecutive weeks, and diesel stocks sit near critical levels before the summer driving season.

Related Articles:

  • The New Joule Order — Ample physical capacity keeps inflation low and stable and allows lower interest rates, but when demand catches up to capacity constraints, better returns in the physical economy pull capital back to fund expansion, making a commodity super-cycle a capital-expenditure cycle. In real terms, total energy investment is below its level 15 years ago.
  • Power Politics: Energy Self-Sufficiency in a Modern Mercantilist World — China leads in adding power to the grid, with more than 30,000 miles of ultra-high-voltage transmission lines versus zero in the US and lead times for large transformers of 48 weeks against a US average of 143 weeks. This positions China to accommodate new demand sources such as artificial intelligence.
  • Fighting Words: The Energy Transition in 2026 — Estimates based on primary energy overstate how much electricity is needed to replace oil in passenger cars, since electric motors convert 80%-90% of electricity into motion versus about 18% for gasoline cars. Measuring useful final energy net of waste heat avoids this "primary energy fallacy": the US consumed 45 exajoules of useful final energy versus 95 exajoules of primary energy.

AI Summary. No Gulf producer's oil output has fallen to zero, as engineers have rotated well shutdowns, choked flows to a trickle, and used pipelines bypassing the Strait of Hormuz to avoid clogging and pressure loss on restart. The author argues that reopening the strait will surprise observers with its speed.

Javier Blas Bloomberg
Date Posted:

Blas argues tanker traffic will likely return to pre-war levels within 30 days of Hormuz opening. He forecasts 50% of the region’s total production capacity will return within a matter of days and 75% within a few weeks.

How quickly can Gulf oil output rebound once shipping resumes?

Core argument: No oil producer in the region has cut output to zero, because domestic demand must be met and Saudi Arabia and the UAE use pipelines that bypass the Strait of Hormuz.

Nowhere in the region has output fallen to zero, because of the need to meet domestic oil demand and, in the cases of Saudi Arabia and the UAE, the use of pipelines bypassing Hormuz. Thus, petroleum engineers have kept some output running continuously, purposely selecting the wells most likely to have problems when reopening if they’d been out of service. In other cases, they’ve rotated the shutdowns, keeping some wells closed for a couple of weeks, then reopening them while shutting down others in an effort to never have an individual well out of business for more than a few weeks. They have choked down flows to reduce output to a trickle, but still keeping a few barrels coming out. By doing so, they try to avoid problems down the road, like clogging or loss of pressure. Oil engineers are, above all, expert problem solvers. So while the restart process won’t be easy once traversing the strait is possible again, don’t mistake difficult for impossible. The reopening of Hormuz, when politics allows it, will surprise with its speed.

Takeaways by Macro Roundup® AI

  1. No oil producer in the region has cut output to zero, because domestic demand must be met and Saudi Arabia and the UAE use pipelines that bypass the Strait of Hormuz.
  2. Petroleum engineers have kept selected wells producing continuously, rotated shutdowns so no well stays closed more than a few weeks, and choked flows to a trickle to avoid later clogging and loss of pressure.
  3. The restart of oil flows through the Strait of Hormuz will be difficult but not impossible, and the reopening, once politics allows it, will surprise with its speed.

Related Articles:

  • How Long Can the US Be the Oil Supplier of Last Resort? — US net exports of crude and refined products have reached a record average of 5.9m barrels a day, up from 3.3m a year earlier, versus net imports above 5m a decade ago. Strategic Petroleum Reserve releases hit a record weekly pace above 1.2m barrels a day, which would leave the reserve at 242m barrels.
  • China’s Invisible Hand Is Rebalancing the Oil Market — Chinese state-owned oil companies have been reselling cargoes to European and Asian buyers, a sign of surplus that has capped benchmark crude prices and collapsed the premiums paid for physical barrels from about $30 above benchmarks to as little as $1. Satellite data show Chinese commercial stockpiles still rising.
  • How Fast Can Gulf Production Recover After Reopening? — Gulf oil production is likely to mostly recover within a few months of a full and safe reopening of the Strait, given limited reported physical damage to oil fields and the likelihood that Saudi Arabia and the United Arab Emirates deploy spare capacity. Full recovery could take several quarters.