Skip to main content

Posts

Following AI exposure across borders, and finding where the statistics go dark

Earlier this month I wrote about Stijn Van Nieuwerburgh’s NBER paper, “Financing the AI Buildout” . His numbers are hard to forget. A one-gigawatt AI campus costs about $41 billion, and a U.S. buildout of 188 gigawatts by 2032 would need nearly $9 trillion of investment, about 3.2 percent of GDP a year. As the hyperscalers’ capital spending outgrows their cash flow, the financing is moving into leases, joint ventures, project debt, private credit, securitization and special-purpose vehicles. In “The AI Boom’s Invisible Balance Sheet” , I mapped part of that financing network from public filings. My conclusion was that the main problem is visibility: the people best placed to judge the risk cannot yet see the whole picture. That post asked who owes whom inside the United States. This one asks the next question: who outside the United States holds the AI boom, and how much of that can official statistics actually see? I am writing the answer up as a working paper. What follows is a pr...
Post a Comment
Read more
Recent posts

The AI Boom's Invisible Balance Sheet

What public filings reveal about who funds AI data centers and who bears the risk, and what oversight should look like Stijn Van Nieuwerburgh's “Financing the AI Buildout” paper describes an investment cycle of nearly $9 trillion over 2025–32 that is increasingly funded through leases, joint ventures, project bonds, private credit and special-purpose vehicles. These structures expand funding capacity, but they also raise leverage at the level of the asset and keep much of it off the balance sheets of the firms that use the capacity. The template is Meta's Hyperion campus in Louisiana. A joint venture, 80 percent owned by Blue Owl funds and 20 percent by Meta, owns the campus and leases it back to Meta; Beignet Investor, the Blue Owl vehicle that holds the 80 percent stake, sold about $27 billion of bonds to pay for it. The paper's central worry is that the resulting exposures are layered, correlated and hard to observe. This post asks how hard. I tried to map the fina...
Post a Comment
Read more

Exposed Twice: AI, Labor, and the Public Purse

Most of the debate about AI and labor markets is a debate about workers. Almost none of it is about the entity standing quietly behind every payroll. Moving from employment-share exposure to labor-tax-revenue exposure changes which occupations matter — and which policies do. Results draw on ongoing joint work with Dhushyanth Raju applying the Richmond (2026) AI transition framework to Türkiye's 2024 Household Labour Force Survey Most of the debate about AI and labor markets is a debate about workers: who gets displaced, who gets augmented, who gets retrained. Almost none of it is about the entity standing quietly behind every payroll: the treasury. Governments in most middle- and high-income countries raise roughly half of their revenue from labor — personal income tax plus social security contributions. If AI-driven capital deepening shifts factor income from labor toward capital, that base erodes at precisely the moment demand for social spending r...
Post a Comment
Read more

Six Questions for a Billion-Dollar Promise

A Skeptic’s Guide to Economic Impact Claims A Skeptic’s Guide to Economic Impact Claims Six questions anyone can ask about the economic promises attached to a big development — and what I learned building the model that produced them. When a large project comes to your county, it arrives with a number. Billions in economic impact. Thousands of jobs. Tens of millions in new revenue. That number is almost always paid for by the people who want the project approved, calculated using software whose workings are not public, and presented to a council as settled fact. Residents get to react to it. They almost never get to check it. Last year I built an economic model of Montgomery County, Maryland, from public data. It estimates what a given amount of spending actually does to the local economy. Building it taught me something more useful than any of its own numbers: the specific places where impact claims go wrong, and how to spot them without needing a mo...
Post a Comment
Read more

The Labor Market Impacts of AI in Türkiye: Beyond Occupational Exposure Analysis

Beyond Exposure AI and Labor Markets · Türkiye Occupational exposure to AI identifies where technology may affect work, but not whether employment will contract, reorganize, or expand. Applying the Richmond (2026) AI Jobs Transition Framework to Türkiye's 2024 Household Labour Force Survey, the dominant high-exposure archetype is not automation. It is reorganization. Mpumelelo Nxumalo and Dhushyanth Raju · Summary of the working paper Beyond Exposure: AI and Labor Market Reorganization in Türkiye Most discussion of AI and jobs runs on a single number: how much of an occupation's task content a model can perform. That number identifies where technology reaches. It does not say what happens next. An exposed occupation may shed workers, or it may keep them while the work itself is rebuilt around the technology. Those are different outcomes, they call for different policy, and exposure alone cannot tell them apart. Applying the Richmond (202...
Post a Comment
Read more