“Unintended Consequences represents the most cogent and persuasive analysis of the Financial Crisis to date.” - Andrei Shleifer, 1999 John Bates Clark Medal Winner
“Unintended Consequences should be read by anyone who takes for granted the superiority of progressive taxation and has not thought carefully about the trade-offs involved.” - The New Republic
“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 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
“…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
“…serious thinking for serious thinkers. …a thought-provoking blueprint for growing middle- and working-class incomes.” - Mitt Romney, former Governor of Massachusetts
“There are an amazing number of good ideas and interesting points made in Unintended Consequences. The thinking underlying it, and the obvious depth of understanding of the author, are very impressive.” - Steven Levitt, coauthor of Freakonomics; 2004 John Bates Clark Medal
“…a fresh argument for the productive value of inequality.” - David Autor, Professor of Economics, Massachusetts Institute of Technology
“…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 very valuable contribution.” - Larry Summers, former Secretary of the Treasury and director of the National Economic Council, president emeritus, Harvard University
AI Summary.In both a successful and failed AI scenario, long-term interest rates fall: AI success drives deflationary growth, while AI failure triggers an equity crash and a flight to bonds.
Torsten SløkApollo
Date Posted:
Sløk argues that the next six months will see either a large increase in revenue associated with AI or the start of a major market correction. He argues either scenario will bring long rates down.
Do long-term rates fall regardless of whether AI succeeds or fails?
Core argument: Regardless of whether AI succeeds or fails, long-term Treasury rates are poised to fall: AI-driven productivity would suppress inflation, while an AI bust would trigger an equity selloff—Nasdaq potentially down ~50%—and a flight-to-safety bid for Treasuries.
If AI succeeds and tech companies generate trillions in revenue, AI will be massively deflationary and push rates lower. If AI does not work out, the bubble bursts and the Nasdaq is down 50% as investors rotate out of equities into Treasuries and long rates fall dramatically. Over the next six months, the market will make up its mind about which AI scenario is playing out. The narrative in rates today is all about inflation and fiscal problems. But the narrative going into 2027 is going to be all about either the success or failure of AI. And in both scenarios, long rates are going to be lower.
Takeaways by Macro Roundup® AI
Regardless of whether AI succeeds or fails, long-term Treasury rates are poised to fall: AI-driven productivity would suppress inflation, while an AI bust would trigger an equity selloff—Nasdaq potentially down ~50%—and a flight-to-safety bid for Treasuries.
Related Articles:
The Summer I Turned Pretty— Capital spending booms historically peak when end-demand companies stagnate while equipment suppliers still thrive; today, semiconductor profits are rising even as the large cloud companies funding AI infrastructure see earnings and cash flow decline, raising doubt over who will sustain AI investment.
The Buyers of AI Are Still Waiting for the Payoff— AI capital spending is generating profit margin gains for technology sellers but not for the companies buying and deploying AI across other sectors of the economy.
The Market Is Asking Questions— AI infrastructure debt spreads are widening as markets question whether returns on massive, front-loaded capital spending will outpace financing costs before assets depreciate. If compute demand plateaus from efficiency gains or slow adoption, the industry faces a glut of expensive, rapidly depreciating capacity.
AI Summary.AI model pricing is converging toward commodity levels, where an 80% annual price decline requires 400% unit growth just to maintain flat revenue. Ceding lower-tier markets to defend premium pricing has historically failed against low-cost competitors, making trillion-dollar valuations difficult to sustain alongside heavy capital spending.
Paul KedroskyApplied Complexity
Date Posted:
US labs have “conceded commoditization” for models behind the frontier, while still trying to “preserve scarcity rents” for their frontier models. Kedrosky warns “this market-ceding tactic has failed in most markets historically, especially when competing with China.”
Does AI pricing collapse force a choice between growth and profitability?
Core argument: DeepSeek forced every Western lab to justify higher costs, then demonstrated retained pricing power by tripling effective output price while remaining within the low-cost band—exposing the fragility of Western labs’ commodity-tier positioning.
Good-enough models are converging in capability and price. That makes model switching easier and durable margins harder to defend. US labs are cutting lower- and mid-tier offerings while keeping their best models expensive. They are conceding commoditization below the frontier while trying to preserve scarcity rents at the top. This market-ceding tactic has failed in most markets historically, especially when competing with China. This is irreconcilable with trillion-dollar valuations and high & growing debt loads. OpenAI and Anthropic must finance enormous capital spending while competition pushes the models likely to generate the most volume toward commodity pricing. An 80% year-over-year price decline requires 400% unit growth just to stand still, and much more to deliver overall growth.
Takeaways by Macro Roundup® AI
DeepSeek forced every Western lab to justify higher costs, then demonstrated retained pricing power by tripling effective output price while remaining within the low-cost band—exposing the fragility of Western labs’ commodity-tier positioning.
U.S. labs ceding mid- and lower-tier segments to defend frontier scarcity rents repeats a market-retreat pattern that has historically failed against low-cost competitors, compounding capital-spending strain at OpenAI and Anthropic where an 80% annual price decline requires 400% unit growth merely to sustain revenue.
Related Articles:
OpenAI and Anthropic In Price War as Chinese AI Rivals Gain Ground— AI model pricing is falling as competition intensifies, with leading models cutting token costs by up to 80%. Higher-priced models can deliver lower total costs by completing tasks in fewer tokens or attempts.
Who’s Afraid of Chinese Models?— Frontier AI labs can sustain lower inference prices as inference revenue scales faster than training costs, making high per-token margins less necessary to fund model development.
The AI Trade Is Losing One of Its Key Signals— AI token prices have fallen over 90% while total spending has roughly doubled, expanding the market overall. However, a 46% gap between AI investment and actual sales — wider than the 32% divergence seen during the 2001 telecom collapse — raises the risk that current infrastructure spending is outpacing real
AI Summary.A leading AI company valued at ~$1tn faces investor scrutiny over Chinese competition, regulatory tensions, and data-center opposition, but argues its premium positioning insulates it from lower-cost rivals whose capabilities trail frontier models by several months.
Berber Jin, Corrie Driebusch and Kate ClarkWall Street Journal
Date Posted:
Anthropic is meeting with investors ahead of a planned IPO in September or early October. The firm has downplayed the potential impact of Chinese competition, arguing “most users want the most intelligent AI systems available at any given time.”
Does premium positioning protect AI leaders from cheaper competitors?
The $965 billion artificial-intelligence juggernaut is staring down a host of fresh challenges that are leading investors to take a more critical look at the company’s business. Company executives have played down the impact of Chinese competition in meetings, telling investors that Anthropic is hyperfocused on offering cutting-edge AI models, the people said. Chief Executive Dario Amodei and other top U.S. AI leaders have publicly said that most users want the most intelligent AI systems available at any given time, suggesting that Chinese systems are less of a threat since their capabilities generally trail those of top AI models by at least a few months.
Related Articles:
They’re Shooting at Everyone!: The Yossarian Effect in Capital Markets— Concentrated leveraged portfolios force broad asset liquidations when a single large position is disturbed, as funds sell liquid winners to fund rebalancing, amplifying losses across correlated holdings through ETF-embedded leverage.
The Summer I Turned Pretty— Capital spending booms historically peak when end-demand companies stagnate while equipment suppliers still thrive; today, semiconductor profits are rising even as the large cloud companies funding AI infrastructure see earnings and cash flow decline, raising doubt over who will sustain AI investment.
OpenAI Leans Toward Waiting Until Next Year for I.P.O.— A major AI company is delaying its public stock offering until next year, as falling tech valuations and weak retail investor appetite for AI stocks make current market conditions unfavorable for a successful listing.
AI Summary.Heavy AI infrastructure spending is compressing free cash flow across major technology companies even as earnings per share rise, creating a widening gap between reported profits and actual cash generation.
John AuthersBloomberg
Date Posted:
As the Mag 7’s free cash flow has turned negative, that of the S&P 500 net the Mag 7 has ticked up since the start of the year.
Is earnings growth masking a cash flow problem for tech companies?
Core argument: S&P 500 earnings per share have outpaced free cash flow over a sustained period, a divergence that began with pandemic-era cost cuts and accelerated sharply after ChatGPT’s launch.
Questions are mounting over how quickly the capex ploughing into AI will translate into profitability, but tech’s earnings growth is still impressive — it’s just that the cost of that investment is increasingly hard to ignore. Their free cash flow is deteriorating at an unprecedented pace as spending on data centres gobbles it up, even as the rest of the corporate sector generates more cash. The numbers for the overall S&P 500, as shown by Bloomberg’s Graph Fundamentals function, confirm a protracted run in which earnings per share have risen much faster than free cash flow. This started after companies had made big pandemic cost cuts and then went into overdrive after the arrival of ChatGPT.
Takeaways by Macro Roundup® AI
S&P 500 earnings per share have outpaced free cash flow over a sustained period, a divergence that began with pandemic-era cost cuts and accelerated sharply after ChatGPT’s launch.
Big-tech free cash flow is deteriorating at an unprecedented pace as data-center capital expenditure consumes cash, even as the broader corporate sector expands cash generation.
Related Articles:
The Buyers of AI Are Still Waiting for the Payoff— AI capital spending is generating profit margin gains for technology sellers but not for the companies buying and deploying AI across other sectors of the economy.
The Summer I Turned Pretty— Capital spending booms historically peak when end-demand companies stagnate while equipment suppliers still thrive; today, semiconductor profits are rising even as the large cloud companies funding AI infrastructure see earnings and cash flow decline, raising doubt over who will sustain AI investment.
A Slower AI Payoff Would Be Everyone’s Problem— Slower-than-expected AI revenue growth would compress hyperscaler cash flows while committed capital spending and depreciation hit on schedule, spreading losses from a handful of dominant tech stocks across broader equity markets and raising recession risk.
AI Summary.AI infrastructure providers earn the highest margins in the AI value chain, but those margins are funded by capital raised at the unprofitable application layer rather than by end-customer revenue, making upstream profitability contingent on continued investor financing of downstream losses.
Torsten SløkApollo
Date Posted:
AI profits have so far come only from the buildout itself, while the end-product generates losses. Slok notes that in the absence of revenue from models and applications, raising capital “can bridge the gap for a while, but not indefinitely.”
Does AI infrastructure profitability depend on funding downstream losses?
In business, profit margins are frequently higher for the owner of the end-customer relationship. But that is not the case for AI. In AI, profit margins are higher the further you get from the end user. This is important because it means the AI boom's profits are currently being funded by investors rather than earned from customers. The upstream margins are real, but they are paid for out of capital raised by the layer losing money, not out of cash generated by end demand. That makes the 41% contingent on the -59% continuing to be financeable. The bottom line is that the most profitable part of the AI value chain depends on the least profitable part continuing to grow revenue or raise capital.
Related Articles:
The State of the AI Economy— AI infrastructure revenues across major cloud providers narrowly cover depreciation costs when compute assets are written down over 6 years. Demand for AI compute still exceeds supply, and every 10% price cut drives 12–18% more token usage, meaning total spending rises as prices fall.
AI: The ROI Runway Could Be Long Outside the Tech Sector— AI valuations embed assumptions of rapid profit growth across the broader economy, but outside the tech sector, returns on AI investment are slow to materialize, creating a gap between current market prices and actual cash flows that could force a sharp repricing.
A Slower AI Payoff Would Be Everyone’s Problem— Slower-than-expected AI revenue growth would compress hyperscaler cash flows while committed capital spending and depreciation hit on schedule, spreading losses from a handful of dominant tech stocks across broader equity markets and raising recession risk.
AI Summary.Data-center investment is expanding at nearly twice the speed of the housing boom at its peak, rising from 1.4% to 3.1% of GDP in two years (~0.85 percentage points per year) versus housing's fastest pace of 0.5 percentage points per year.
Torsten SløkApollo
Date Posted:
Sløk forecasts hyperscaler capex to peak at 3.2% of GDP, less than half housing’s 6.6% peak, but it is projected to rise ~70% more quickly – 0.85pp of GDP per year at its fastest, versus 0.5pp for housing.
Is artificial intelligence infrastructure growing unsustainably fast?
Core argument: Data-center capex is accelerating at ~0.85 percentage points of GDP per year—nearly twice the housing boom’s peak pace of 0.5 points per year and roughly six times the telecom fiber cycle’s rate.
In level, the ongoing data-center buildout sits between the fiber and housing cycles: more than twice the fiber peak, less than half the housing peak. In cumulative change, what matters is not the level of the share but how much it moves. On this measure, the data-center buildout is the bigger capex cycle. In speed, the contrast is sharper still, and it holds even when each cycle is measured over its own fastest stretch. Data-center capex adds 1.7 percentage points in just two years, from 1.4% of GDP in 2025 to 3.1% in 2027, or roughly 0.85 percentage points a year. Housing's quickest phase, from 5.1% in 2002 to 6.6% in 2005, ran at 0.5 percentage points a year, and telecom's at around 0.15. The AI cycle is building at close to twice the pace of the housing boom at its fastest.
Takeaways by Macro Roundup® AI
Data-center capex is accelerating at ~0.85 percentage points of GDP per year—nearly twice the housing boom’s peak pace of 0.5 points per year and roughly six times the telecom fiber cycle’s rate.
Data-center investment is projected to reach 3.1% of GDP by 2027, a 1.7-point rise from 1.4% in 2025, representing the largest cumulative capex swing of the three cycles on a change basis.
In level terms, the data-center buildout sits between the fiber and housing peaks—more than twice the fiber cycle’s peak yet still below half the housing cycle’s 6.6% of GDP.
Related Articles:
The Market Is Asking Questions— AI infrastructure debt spreads are widening as markets question whether returns on massive, front-loaded capital spending will outpace financing costs before assets depreciate. If compute demand plateaus from efficiency gains or slow adoption, the industry faces a glut of expensive, rapidly depreciating capacity.
Semiquincententacles— Frontier AI labs are shifting to usage-based pricing that reflects total costs, raising token prices and pushing some enterprises toward cheaper open models for routine tasks, while frontier models remain essential for high-performance applications like cybersecurity and scientific discovery.
AI: The ROI Runway Could Be Long Outside the Tech Sector— AI valuations embed assumptions of rapid profit growth across the broader economy, but outside the tech sector, returns on AI investment are slow to materialize, creating a gap between current market prices and actual cash flows that could force a sharp repricing.
AI Summary.U.S. stocks are historically expensive, with the gap between earnings yields and inflation-adjusted bond yields at 1.4% — well below the 4.7% long-run average. When this gap is this narrow, 10-year stock returns have historically been poor.
Martin WolfFinancial Times
Date Posted:
US equities account for 55% of the value of global stock markets. Shiller’s excess Cape yield implies US stocks will return 1.4% over the next 10 years, well under their long-term mean of 4.7%.
Are historically narrow valuation gaps predicting weak stock returns ahead?
[Shiller's] "excess Cape yield” measures the difference between the inverted Cape ratio — or cyclically adjusted aggregate earnings per share — and inflation-adjusted yields on Treasury bonds. When stocks are expensive, the excess yield is low. When stocks are cheap, the excess yield is high. Crucially, when the excess yield is low, the subsequent 10-year excess returns on stocks have normally been poor. The excess yield is a mere 1.4% in July 2026, far below its long-run average of 4.7%. With stocks this expensive, the chances of healthy future returns must, again, be relatively low. Cape is far from a perfect predictor of an imminent crash: if it were, it would not be; well-informed investors would then not allow markets to reach extreme positions in the first place, and so crashes would be far less likely.
Related Articles:
US Stock Market To Stop Shrinking For First Time In 23 Years— US equity supply is turning positive for the first time in over two decades, as a surge in IPOs and large share sales by major technology companies outweighs the buybacks and privatizations that have shrunk the stock market since 2003.
Opportunities and Expectations: The Present Value of Growth Opportunities in Valuation— A stock index's price can be split into steady-state earnings value and the value of future growth opportunities; historically, the growth opportunity component has averaged 35% of total price, and periods when this share is below average have preceded stronger 10-year returns.
One Hundred Years in the U.S. Stock Markets— Btw January 1926 and December 2025, 60% of US firms had negative total returns relative to T-bills. 46 firms accounted for half of the $91T in net wealth…
AI Summary.Data-center construction spending rose 23% year-over-year, while manufacturing construction fell 22%, and materials costs for nonresidential construction are up more than 55% since early 2020. Surging demand for electrical equipment from data centers and solar projects has created order backlogs exceeding two years, driving up costs across all nonresidential construction.
Bob TitaWall Street Journal
Date Posted:
Spending on data center construction in May rose 23% y/y, while construction spending on manufacturing buildings fell 22%. The data center buildout could be crowding out non-tech-related fixed investment.
Is AI infrastructure crowding out other construction investment?
Core argument: Manufacturing construction spending dropped 22% year-over-year to $174bn, results in data centers capturing equipment supply chains and extending electrical equipment.
Spending on data center construction in May rose 23% from a year earlier, according to the U.S. Census Bureau. Data centers accounted for 8% of the total spending on private, nonresidential construction. Construction spending on manufacturing buildings dropped 22% year-over-year in May to a seasonally adjusted annual rate of $174 billion. Manufacturing makes up the largest part of private, nonresidential construction, accounting for nearly a quarter of the spending in May, according to the Census Bureau. Materials costs for nonresidential construction are up more than 55% since early 2020 and even higher for fabricated steel, copper wire, and some other individual materials, according to the government’s producer price index.
Takeaways by Macro Roundup® AI
Manufacturing construction spending dropped 22% year-over-year to $174bn, results in data centers capturing equipment supply chains and extending electrical equipment.
Nonresidential construction materials costs surged 55%+ since early 2020 with transformer prices up 70% in five years, leads to extended.
Related Articles:
Midyear Outlook— AI investment is crowding out rival capital projects by absorbing scarce physical inputs—grid capacity, construction labor, metals, and engineering talent—while semiconductor prices surge where supply constraints meet surging demand. With hyperscaler free cash flow exhausted, over 80% of future AI capital spending requires external financing, including fresh equity issuance.
The Other US Capex Question— Weak non-AI business investment in the U.S. is driven primarily by near-zero labor force growth from tightened immigration policy, not by AI spending crowding out capital, since corporate savings are sufficient to fund both simultaneously.
