“…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
“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
“Unintended Consequences is full of substance, it is one of the must-read books of the year, and once I finish it I will be giving it a second read through right away.” - Tyler Cowen, Professor, George Mason University
“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
“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 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 must-read for serious students of economic policy.” - Glenn Hubbard, Dean, Columbia Business School, and former Chairman of the Council of Economic Advisers
“A full-throated defense of economic dynamism.” - The Wall Street Journal
“…a comprehensive explanation of the modern economy.” - Julian Robertson, Founder, Tiger Management
“…reminds us that inequality sends a signal of what society lacks most, in America’s case, entrepreneurship and risk taking.” - Lawrence Lindsey, CEO, The Lindsey Group, former Director of the National Economic Council
“…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
AI Summary.Large US cloud and artificial-intelligence companies have raised $48bn in European-currency bonds, more than triple the prior year's total, spread across €27bn, £13bn and CHF7.5bn. The issuance adds highly rated US borrowers to European credit indices and deepens the thin market for longer-dated maturities, with few signs so far of crowding out other issuers.
Huw van SteenisApollo
Date Posted:
Hyperscalers have raised $48B in European-currency bonds this year, more than 3x all of 2025. They now account for 3% of euro IG issuance, 10% of sterling and 22% of Swiss franc, versus 8% in the US.
Are European bond markets funding the US AI buildout?
Core argument: Hyperscalers have raised $48 billion in bonds denominated in European currencies this year — €27 billion, £13 billion and CHF7.5 billion — more than triple the full-year 2025 total.
Hyperscalers have raised $48 billion in bonds in European currencies this year, already more than triple the entire amount raised last year. This breaks down as €27 billion, £13 billion, and CHF 7.5 billion. There are few signs of crowding out: non-hyperscaler IG issuance remains strong and little changed YoY, and corporate spreads remain tight. Rather, hyperscalers are adding highly rated US borrowers (AA- or higher) to European credit indices that have traditionally had more A-to-BBB exposure. Hyperscalers account for 7% of euro issuance with maturities of 10 years or more, versus 3% across all maturities, and 3% of the 10-year-plus euro IG index versus 1% of the overall index. This adds depth at longer-dated maturities—thin outside sovereign issuance—at a time when some governments, including the UK, have been shortening duration.
Takeaways by Macro Roundup® AI
Hyperscalers have raised $48 billion in bonds denominated in European currencies this year — €27 billion, £13 billion and CHF7.5 billion — more than triple the full-year 2025 total.
Hyperscaler issuance shows few signs so far of crowding out European borrowers, with non-hyperscaler investment-grade issuance little changed year-over-year and corporate spreads still tight.
Hyperscalers account for 7% of euro issuance with maturities of 10 years or more versus 3% across all maturities, adding borrowers rated AA- or higher to indices traditionally weighted toward A to BBB.
Related Articles:
Financing the AI Buildout— The authors estimate roughly 57 gigawatts of operating US data center capacity against a planned development pipeline of about 509 gigawatts, which is not a completion forecast since projects may be delayed, scaled back, or abandoned. Their central scenario has 182.8 gigawatts of added capacity becoming operational through 2032, 117.2 gigawatts completed later, and 226.9 gigawatts never completed.
AI Set to be Largest CapEx Cycle Ever … and Soon Majority Externally Financed— Hyperscaler and data-center investment is tracking toward $1.5tn annually by 2030, which in real terms would exceed the 2005-06 housing boom, the 1990s fiber peak, and the 1882 railroad peak combined. External debt and equity are projected to fund close to 90% of the buildout, making it dependent on credit markets.
AI Is Driving Up Treasury Yields: ‘It Just Touches Everything’— Investment-grade companies have sold nearly $1.5tn of bonds, up 36% from a year earlier, and some investors have sold US Treasuries to buy higher-yielding corporate debt. Bank of America economists estimate the surge in corporate-debt sales, along with more mortgage-backed securities issuance, pushed 10-year rates up by about 0.3 percentage point.
AI Summary.Utilization-adjusted total factor productivity from the Federal Reserve Bank of San Francisco sits slightly below zero with no acceleration since the artificial intelligence investment cycle began, even as output per hour runs near 2.5%, above its post-2005 average. That combination is the signature of capital deepening rather than a technology shock.
Torsten SløkApollo
Date Posted:
The FRBSF’s utilization adjusted TFP series shows no sign of acceleration since the start of the AI capex cycle. Slok reiterates that the rise in output per hour “alongside flat TFP is the signature of capital deepening, not of a technology shock.”
Does productivity data show evidence of an AI technology shock?
Core argument: Utilization-adjusted total factor productivity from the San Francisco Fed sits slightly below zero with no acceleration since the AI capex cycle began, while output per hour runs near 2.5%.
The chart shows utilization-adjusted TFP from the San Francisco Fed, and it is currently sitting slightly below zero with no sign of acceleration since the AI capex cycle began. Output per hour, by contrast, is running near 2.5%, comfortably above its post-2005 average, and that strength is exactly what gets cited as evidence that AI is already working. But strong output per hour alongside flat TFP is the signature of capital deepening, not of a technology shock. TFP has swung between roughly -3% and +4% over the past 40 years with no discernible trend, so the current sub-zero reading is not in itself unusual. Electricity and IT both took a decade or more to show up in aggregate numbers.
Takeaways by Macro Roundup® AI
Utilization-adjusted total factor productivity from the San Francisco Fed sits slightly below zero with no acceleration since the AI capex cycle began, while output per hour runs near 2.5%.
Output per hour above its post-2005 average alongside flat total factor productivity is the signature of capital deepening rather than a technology shock, despite that strength being cited as evidence that AI is already working.
Total factor productivity has swung between roughly -3% and +4% over the past 40 years with no discernible trend, and both electricity and IT took a decade or more to show up in aggregate numbers.
Related Articles:
AI and Productivity— The recent pickup in US labor productivity appears driven by higher utilization of existing capital—running factories, server racks, and hotel rooms harder—rather than by firm-level efficiency gains from AI. Adoption remains too thin and the cross-sectional evidence too weak to conclude that AI-driven productivity gains are already widespread.
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.
Writing Code vs. Shipping Code: Productivity Effects Across Generations of AI Coding Tools— 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%.
AI Summary.Canada's core inflation is around 2% and its budget deficit about 2.4%, versus roughly 6% in the US, yet Canadian bond yields move with US yields. The author argues global AI-related debt issuance, including multibillion-dollar Canadian-dollar bonds from Google and Amazon, helps transmit higher US borrowing costs into Canada.
Paul PodolskyA Letter From Paul
Date Posted:
Canada’s 10-year yield rose from ~3% to ~4% despite a modest 2.4% budget deficit, slow domestic growth and 2% core inflation. Podolsky sees the US bond selloff as raising yields globally. “It can seem odd that bond markets trade in sync, but they do.”
Are US borrowing costs being exported to Canada through global AI debt?
Core argument: Canada’s core CPI runs near 2% and its budget deficit near 2.4%, versus roughly 6% in the US, which the author attributes to Canada not being a meaningful part of the AI boom.
Core CPI in Canada is around 2%. This is because Canada is not a meaningful part of the AI boom and is also not running a big budget deficit. Canada’s budget deficit is around 2.4%, as opposed to about 6% in the US. Canada’s bond-market reward for the US’s truculent, guns-and-butter policy is higher borrowing costs and slower growth. In part this is because the AI debt issuance is global. In May, Google raised C$8.5B and in June Amazon raised C$14B, or about one-third of net new Canadian corporate issuance this year. Over time, however, markets often revert to pricing that accurately reflects local conditions. [The chart shows] yields in the US, Canada, and Canadian unemployment. As you can see, as US yields sell off, Canada sells off, even if Canadian employment conditions are tepid. Obviously, the US (orange line) has sold off more.
Takeaways by Macro Roundup® AI
Canada’s core CPI runs near 2% and its budget deficit near 2.4%, versus roughly 6% in the US, which the author attributes to Canada not being a meaningful part of the AI boom.
Canadian bond yields sell off alongside US yields even when Canadian employment conditions are tepid, leaving Canada with higher borrowing costs and slower growth, though US yields have sold off more.
Google raised C$8.5B in May and Amazon C$14B in June, together about one-third of net new Canadian corporate issuance this year, reflecting the global nature of AI-related debt issuance.
Related Articles:
Foreign Capital Flows Into US Stocks Hit Record As Appetite For Debt Fades— Foreign investors made a record $942bn in net purchases of US stocks and investment fund shares over 12 months, the highest rolling 12-month total in Treasury data going back to 1985. Foreign demand for US debt securities moved in the opposite direction, falling to a net $188bn in the second quarter from $314bn in the first.
AI Is Driving Up Treasury Yields: ‘It Just Touches Everything’— Investment-grade companies have sold nearly $1.5tn of bonds, up 36% from a year earlier, and some investors have sold US Treasuries to buy higher-yielding corporate debt. Bank of America economists estimate the surge in corporate-debt sales, along with more mortgage-backed securities issuance, pushed 10-year rates up by about 0.3 percentage point.
Is the AI Buildout Pushing Up Yields?— US saving-investment data show that non-financial corporations remain net savers, unlike the early-2000s information technology boom when they flipped to net borrowers and their capital spending accounted for the entire current account deficit. Government dissaving through the budget deficit, not the artificial intelligence buildout, is absorbing national resources.
AI Summary.The authors estimate AI capital spending of $560bn in 2026 and, in a high-investment case, $3.5tn of net AI capital in service by 2030, requiring roughly $1.3tn in annual infrastructure revenue once depreciation, operating costs and a 15% return are covered. That implies 7.4x growth from current annualized revenues, with downstream revenues of $1.9–2.6tn.
William Gildea and Azeem AzharExponential View
Date Posted:
Is Database:
Database
Beginning with capex projections, Gildea/Azhar derive for their high-growth case a $1.3T annual AI-infrastructure revenue hurdle for 2030. This covers $771B of depreciation and operating costs, plus a 15% return on $3.5T of net capital in service.
How much annual revenue must AI infrastructure generate to justify investment?
Core argument: In a high investment case with $3.5tn of net AI capital in service, the authors estimate AI infrastructure must generate $1.3tn of annual revenue by 2030, requiring 7.4x growth (59% CAGR) from today’s annualized revenues.
We estimate AI CapEx to be $560B in 2026, 64% of the total declared by the hyperscalers. Unfinished facilities can’t generate revenue, so the investment bill and the productive asset base will grow at different speeds. [In the high investment case], by 2030 we see $3.5T of net AI capital in service. The largest component [of the revenue hurdle] is the $555B in compute depreciation. We model lifespans at the current 5-6 years. Shell depreciation and OpEx combined contribute about a further $216B to the required revenue hurdle. [Adding a 15% return on net capital brings required annual infrastructure revenue to $1.3T]. Reaching the $1.3T hurdle would represent 7.4x growth (59% CAGR) from today’s annualized revenues. For its customers to earn 30-50% gross margins, downstream revenues need to be $1.9-2.6T. [Editor’s note: Van Nieuwerburgh’s far larger figure of $3.7T covers mature revenues from projects completed through 2032, using commissioning capital and a 50% infrastructure operating cash-flow margin. This estimate instead charges depreciation and operating costs plus 15% on net capital in service in 2030. Different horizons and cost conventions prevent a direct comparison. Van Nieuwerburgh also has a much higher capex estimate.]
Takeaways by Macro Roundup® AI
In a high investment case with $3.5tn of net AI capital in service, the authors estimate AI infrastructure must generate $1.3tn of annual revenue by 2030, requiring 7.4x growth (59% CAGR) from today’s annualized revenues.
Compute depreciation of $555bn, modeled on 5-6 year lifespans, is the largest component of the 2030 revenue hurdle, with shell depreciation and operating expenses adding about $216bn and a 15% return on net capital completing the $1.3tn total.
Related Articles:
Financing the AI Buildout— The authors estimate roughly 57 gigawatts of operating US data center capacity against a planned development pipeline of about 509 gigawatts, which is not a completion forecast since projects may be delayed, scaled back, or abandoned. Their central scenario has 182.8 gigawatts of added capacity becoming operational through 2032, 117.2 gigawatts completed later, and 226.9 gigawatts never completed.
What Investment Data Implies about the AI Transition— The five largest U.S. technology firms are scaling AI capital spending about 50% a year, from $155bn to a forecast $755bn. Embedding this in a two-sector model with rare productivity booms, the authors estimate AI adds roughly 5, 20, or 58pp to cumulative growth over a seven-year transition, depending on scenario.
Rear Window— Major cloud providers posted annual revenue growth of 37% at Amazon Web Services, 43% at Microsoft Azure and 82% at Google Cloud, with rising revenue backlogs of $514bn and $496bn respectively at Google Cloud and Amazon Web Services. Earnings growth appears highly contingent on continued hyperscaler capital spending and research, which rose from $460bn to $870bn.
AI Summary.Wall Street analysts covering technology project the sector's operating cash flow will more than double to roughly $2.4tn by 2028, a gain of over $1.2tn, while analysts covering the rest of the S&P 500 — technology's customers — project those firms will add far less cash flow. These sector-by-sector forecasts are internally inconsistent.
Torsten SløkApollo
Date Posted:
Is Database:
Database
Tech analysts expect sector operating cash flow to more than double to ~$2.4T by 2028. Analysts covering other S&P 500 sectors (tech’s customers) see far less cash growth in their sectors. Slok “raises the question of who exactly will be writing all those checks to buy AI services.”
Are tech cash flow forecasts inconsistent with customers' projected growth?
Core argument: Analysts covering the tech sector expect its operating cash flow to more than double to roughly $2.4tn by 2028, an increase of over $1.2tn.
Wall Street equity analysts work in sector silos, and when you add up their forecasts, the numbers are internally inconsistent. The analysts covering tech expect the sector's operating cash flow to more than double to roughly $2.4 trillion by 2028, an increase of over $1.2 trillion. Meanwhile, the analysts covering the other sectors in the S&P 500, which are tech's customers, expect those companies to add much less operating cash flow. In other words, the tech silo is betting on a future in which demand for AI and tech services explodes, while the silos covering the companies that would pay for those services see a much more modest outlook. Both cannot be right at the same time. The bottom line is that either tech's customers will generate a lot more cash than their analysts expect, or tech's cash flow forecasts are too optimistic, which raises the question of who exactly will be writing all those checks to buy AI services.
Takeaways by Macro Roundup® AI
Analysts covering the tech sector expect its operating cash flow to more than double to roughly $2.4tn by 2028, an increase of over $1.2tn.
Analysts covering the other S&P 500 sectors, which are tech's customers, expect those companies to add much less operating cash flow, leaving the combined forecasts internally inconsistent.
Either tech's customers will generate much more cash than their analysts expect, or tech's cash flow forecasts are too optimistic, which raises the question of who will pay for AI services.
Related Articles:
A Catalyst For The AI Bubble Break— Historically, controlling for valuations, a 1% rise in the share of the market represented by initial public offerings has been associated with roughly a 4% decline in prices. With large artificial-intelligence company listings expected and share issuance exceeding buybacks, equity supply could shift from shrinking about 1% a year to growing near 5%, implying returns roughly 20% below average.
Financing the AI Buildout— The authors estimate roughly 57 gigawatts of operating US data center capacity against a planned development pipeline of about 509 gigawatts, which is not a completion forecast since projects may be delayed, scaled back, or abandoned. Their central scenario has 182.8 gigawatts of added capacity becoming operational through 2032, 117.2 gigawatts completed later, and 226.9 gigawatts never completed.
Rear Window— Major cloud providers posted annual revenue growth of 37% at Amazon Web Services, 43% at Microsoft Azure and 82% at Google Cloud, with rising revenue backlogs of $514bn and $496bn respectively at Google Cloud and Amazon Web Services. Earnings growth appears highly contingent on continued hyperscaler capital spending and research, which rose from $460bn to $870bn.
AI Summary.The author argues that a slowdown in artificial-intelligence capital spending would remove a stimulus that has been masking weakness, colliding with overcapacity, falling prices, refinancing needs and weaker labor income. On this view the main risk is deflationary, transmitting through investment, employment, consumption, credit, housing and asset prices at once.
Paul KedroskyApplied Complexity
Date Posted:
Kedrosky argues that the “inevitable transition” from the current inflationary AI-buildout regime to a new regime in which AI-driven productivity gains exert deflationary pressure is “where the real macro risk lies.”
Would an AI spending slowdown create deflationary rather than inflationary risk?
Core argument: The author argues the dangerous phase arrives when the AI capex boom slows, removing stimulus just as overcapacity, falling prices, refinancing needs and weaker labor income hit together, with a refinancing wall waiting in 2028.
If AI investment has been masking underlying weakness, a slowdown removes the stimulus just as overcapacity, falling prices, refinancing needs and weaker labor income hit together. A refinancing wall sits waiting in 2028. Failure is over-determined. This moment sits at the intersection of the four forces that have produced the most consequential boom-bust cycles in Western economic history: loose credit, government policy, technology, and real estate. There are, as a result, myriad catalysts, each with a low probability of happening, but which collectively guarantee collapse. That is why the principal risk is deflationary rather than inflationary. An AI bust could transmit through capex, employment, consumption, credit, housing, and asset prices simultaneously.
Takeaways by Macro Roundup® AI
The author argues the dangerous phase arrives when the AI capex boom slows, removing stimulus just as overcapacity, falling prices, refinancing needs and weaker labor income hit together, with a refinancing wall waiting in 2028.
The principal risk from an AI bust is deflationary rather than inflationary, according to the author, because it could transmit simultaneously through capex, employment, consumption, credit, housing, and asset prices.
The author places this moment at the intersection of loose credit, government policy, technology, and real estate — the four forces behind the most consequential boom-bust cycles in Western economic history — and sees a plausible decade of Japan-style balance recession with fiscal stimulus constrained by massive deficits.
Related Articles:
Financing the AI Buildout— The authors estimate roughly 57 gigawatts of operating US data center capacity against a planned development pipeline of about 509 gigawatts, which is not a completion forecast since projects may be delayed, scaled back, or abandoned. Their central scenario has 182.8 gigawatts of added capacity becoming operational through 2032, 117.2 gigawatts completed later, and 226.9 gigawatts never completed.
A Catalyst For The AI Bubble Break— Historically, controlling for valuations, a 1% rise in the share of the market represented by initial public offerings has been associated with roughly a 4% decline in prices. With large artificial-intelligence company listings expected and share issuance exceeding buybacks, equity supply could shift from shrinking about 1% a year to growing near 5%, implying returns roughly 20% below average.
AI Set to be Largest CapEx Cycle Ever … and Soon Majority Externally Financed— Hyperscaler and data-center investment is tracking toward $1.5tn annually by 2030, which in real terms would exceed the 2005-06 housing boom, the 1990s fiber peak, and the 1882 railroad peak combined. External debt and equity are projected to fund close to 90% of the buildout, making it dependent on credit markets.
AI Summary.An analysis estimates that for US data center owners to break even on AI investments, the sector would need about $3.725tn in annual compute revenue by 2032 — roughly 9.2% of projected 2032 GDP. The estimate assumes 183 GW of US capacity completed, a $9.61tn capital base, a 10% discount rate, and a 50% operating cash flow margin.
Hanno LustigThe Two Cents
Date Posted:
Is Database:
Database
The US AI-buildout is projected to average ~3.6% of GDP over 2025–32. Lustig notes data-center operators will need ~$3.725T in annual compute revenue by 2032, 9.2% of projected GDP, to earn a 10% unlevered return on the buildout.
How much annual revenue must AI data centers generate to break even?
Core argument: Stijn’s baseline estimate holds that data center owners would need $3.725 trillion of annual revenue by 2032 — roughly 9.2% of projected 2032 GDP — to break even on AI investment at a 10% discount rate.
What do you have to believe about future revenue in order for all data center owners to to break even on their AI investment? Using a 10% discount rate, Stijn’s baseline number is 3.725 trillion of annual revenue by 2032, which would amount to roughly 9.2% of projected 2032 GDP. This number is the top line earned from selling compute in aggregate by whoever operates the 182.7 GW of US capacity completed by 2032.2. The assumptions behind that number: 183 GW of U.S. data-center capacity completed between 2025 and 2032. $8.54 trillion of construction cost, grossed up to a $9.61 trillion capital base to account for the return on capital tied up during construction. A 10% discount rate. A 6-year life for IT equipment (68% of cost) and a 20-year life for everything else (32%). A 50% operating cash flow margin.
Takeaways by Macro Roundup® AI
Stijn’s baseline estimate holds that data center owners would need $3.725 trillion of annual revenue by 2032 — roughly 9.2% of projected 2032 GDP — to break even on AI investment at a 10% discount rate.
The break-even calculation assumes 183 GW of U.S. data-center capacity completed between 2025 and 2032 at $8.54 trillion of construction cost, grossed up to a $9.61 trillion capital base for return on capital tied up during construction.
The estimate assumes a 6-year life for IT equipment, which accounts for 68% of cost, a 20-year life for the remaining 32%, and a 50% operating cash flow margin.
Related Articles:
Financing the AI Buildout— The authors estimate roughly 57 gigawatts of operating US data center capacity against a planned development pipeline of about 509 gigawatts, which is not a completion forecast since projects may be delayed, scaled back, or abandoned. Their central scenario has 182.8 gigawatts of added capacity becoming operational through 2032, 117.2 gigawatts completed later, and 226.9 gigawatts never completed.
A Catalyst For The AI Bubble Break— Historically, controlling for valuations, a 1% rise in the share of the market represented by initial public offerings has been associated with roughly a 4% decline in prices. With large artificial-intelligence company listings expected and share issuance exceeding buybacks, equity supply could shift from shrinking about 1% a year to growing near 5%, implying returns roughly 20% below average.
Rear Window— Major cloud providers posted annual revenue growth of 37% at Amazon Web Services, 43% at Microsoft Azure and 82% at Google Cloud, with rising revenue backlogs of $514bn and $496bn respectively at Google Cloud and Amazon Web Services. Earnings growth appears highly contingent on continued hyperscaler capital spending and research, which rose from $460bn to $870bn.
AI Summary.The authors estimate roughly 57 gigawatts of operating US data center capacity against a planned development pipeline of about 509 gigawatts, which is not a completion forecast since projects may be delayed, scaled back, or abandoned. Their central scenario has 182.8 gigawatts of added capacity becoming operational through 2032, 117.2 gigawatts completed later, and 226.9 gigawatts never completed.
Stijn Van NieuwerburghBrookings Papers on Economic Activity
Date Posted:
Is Database:
Database
Is Important:
Important
Stijn Van Nieuwerburgh projects ~$10.3T will be spent on the AI buildout over 2025–2032, averaging 3.63% of GDP. He estimates 2/3 of that spending reflects IT equipment and one-third real estate and associated power infrastructure.
How much of the planned data center pipeline gets built?
Core argument: Using Cleanview project-level data from July 2026, the authors estimate roughly 57 GW of operating U.S. data center capacity against a planned development pipeline of roughly 509 GW, which they stress is not a completion forecast.
The required investment is enormous. A representative 200 MW AI training campus costs roughly $8.2 billion, about two-thirds of which reflects IT equipment and one-third the real estate and associated power infrastructure. We estimate that 183 GW is completed by 2032 with a further 118 GW completed after 2032. This pipeline will trigger about $10.3 trillion dollars of investment over 2025–2032. At an average of 3.63% of GDP per year, the projected buildout would be larger relative to the economy than the major U.S. canal, railroad, electrification, highway, and telecommunications investment booms. It would double the electricity consumption of the entire U.S. residential sector.
Takeaways by Macro Roundup® AI
Using Cleanview project-level data from July 2026, the authors estimate roughly 57 GW of operating U.S. data center capacity against a planned development pipeline of roughly 509 GW, which they stress is not a completion forecast.
The authors’ central scenario allocates the announced pipeline so that 182.8 GW of additional U.S. capacity becomes operational during 2025–32, including 19.0 GW tied to projects already operational in 2025 or 2026.
The authors’ scenario has 117.2 GW of planned data center capacity completed after 2032 and 226.9 GW never completed, applying lower realization rates to more distant and less fully documented projects.
Related Articles:
Rear Window— Major cloud providers posted annual revenue growth of 37% at Amazon Web Services, 43% at Microsoft Azure and 82% at Google Cloud, with rising revenue backlogs of $514bn and $496bn respectively at Google Cloud and Amazon Web Services. Earnings growth appears highly contingent on continued hyperscaler capital spending and research, which rose from $460bn to $870bn.
AI Set to be Largest CapEx Cycle Ever … and Soon Majority Externally Financed— Hyperscaler and data-center investment is tracking toward $1.5tn annually by 2030, which in real terms would exceed the 2005-06 housing boom, the 1990s fiber peak, and the 1882 railroad peak combined. External debt and equity are projected to fund close to 90% of the buildout, making it dependent on credit markets.
The Summer I Turned Pretty— A warning sign in a boom cycle is when companies closest to final demand weaken while capital spending beneficiaries keep thriving, as when telecom operators' shares rolled over while communications equipment stocks kept rising. Semiconductor stock prices and free cash flow are soaring while large cloud providers' shares stagnate and their free cash flow plummets.
