The Information Machine
Concluded·following since 9 Aug 2026·Day 3·9 sources·updated 17 Aug 2026

Columbia's US AI infrastructure spending research

The gist

Columbia research places US AI infrastructure at 2.8% of GDP, exceeding every prior American buildout

The 2.8% GDP share, if sustained, would make AI the largest single-sector capital commitment in US economic history, and Columbia's research implies the economy has grown dependent enough on the buildout that a slowdown carries recession risk. Debt markets are being pushed toward volumes that may cross historical issuance records, with investor absorption capacity already in question.

The full picture

Columbia Business School's Stijn Van Nieuwerburgh has calculated that AI infrastructure investment is projected to average roughly 2.8% of GDP between 2025 and 2032, a share that exceeds every prior US buildout, including the railroad boom (~2%), the interstate highway system (~1%), and the 1990s telecom buildout (~1%). Van Nieuwerburgh's research found that AI-related investment accounted for essentially all observed US GDP growth in Q4 2025, and a Columbia Business School research brief warns that the buildout is reshaping capital markets and creating new economic vulnerabilities, with the US economy at risk of recession if data center investment falters.

Scaling to roughly 200 gigawatts of data center capacity implies a total buildout cost near $8.2 trillion, with roughly half financed externally. Amazon, Alphabet, Meta, Oracle, Nvidia, and SpaceX issued roughly $182 billion in investment-grade bonds in 2026 alone, a 1,300% increase year-over-year, and AI-related debt has reached roughly $1.2 trillion, making it the single largest sector in the investment-grade bond market, larger than US banks. Reuters reported that AI-fueled debt volumes may push total investment-grade issuance above $2 trillion for the first time ever in 2026; hyperscaler deals had already surpassed their full-year 2025 total partway through the year, BNP Paribas forecasts hyperscaler issuance at $250 billion for the year, and investors are beginning to question whether the market can continue to absorb the supply. Bloomberg reported that hyperscalers, historically funded primarily from operating cash flow, are diversifying into bonds, leases, joint ventures, and project financing; JPMorgan expects technology, media, and telecom bond sales to reach $540 billion and estimates more than $2 trillion may be needed for AI chips over the next five years.

On the electricity side, BloombergNEF has revised its US data center power capacity forecast three times in fifteen months: from 78 gigawatts in April 2025 to 106 gigawatts in December 2025, then to 194 gigawatts by 2035, an 83% increase over the December figure. Data centers currently account for about 5.9% of US electricity consumption and are projected to reach roughly 20% by 2035. PJM Interconnection, spanning Virginia to Illinois, is projected to send 34% of its electricity to data centers by 2035, and ERCOT, covering most of Texas, is projected to devote 22% of its generating capacity to data centers. EPRI has more than doubled its 2024 demand estimate, and S&P's forecast rose by more than a third between October and April. BloombergNEF's base case requires the US grid to add new data center capacity at its all-time record pace of 7.1 gigawatts per year every year for the next decade, and a chip-based demand scenario projects need at 207 gigawatts by 2033, leaving a roughly 19-gigawatt shortfall even after new gas generation.

How it developed
17 August 2026

A Columbia Business School research brief on August 16 warned the US could enter recession if data center investment falters, extending Van Nieuwerburgh's calculation that AI infrastructure will average 2.8% of GDP from 2025 to 2032, more than any prior US buildout.

Reuters reported AI issuance may push investment-grade totals above $2 trillion in 2026 for the first time, with investors questioning absorption capacity. Regional grid data from ERCOT and EPRI corroborated BloombergNEF's 194-gigawatt US data center capacity forecast for 2035.

10 August 2026

Milk Road reports Columbia Business School research placing the AI buildout at 2.8% of GDP with $182 billion in bonds issued by major tech companies in 2026 alone and AI debt reaching $1.2 trillion

9 August 2026

Bloomberg reports hyperscalers diversifying from cash-flow funding into bonds, leases, joint ventures, and project financing at unprecedented scale

Sources
4 more sources
The daily email

Want this in your inbox?

I send a short email each morning with the stories that moved. If you would rather just read here, that works too.

Subscribe free