A forecast that U.S. AI infrastructure investment could exceed $10 trillion over the next eight years drew a reaction from Tesla CEO Elon Musk (일론 머스크), who referenced a “K2 economy.” Musk did not directly assess whether the scale of investment is excessive.

On Sept. 25 (local time), Cryptopolitan reported that Musk wrote on X (Twitter) the previous day, citing a U.S. AI infrastructure investment chart shared by investor Andrew Chen (앤드루 첸), “An Earth economy is less than a trillionth the size of a K2 economy.” Here, K2 refers to a Type II civilization on the Kardashev Scale, meaning a civilization that can use nearly all the energy emitted by its parent star. The “one trillionth” comparison is Musk’s wording and is not a research finding.

The chart is based on the Brookings Papers on Economic Activity paper “Financing AI Buildout,” written by Columbia University professor Stijn Van Nieuwerburgh (스테인 판 니우베르그). The study estimates that U.S. investment in data center buildings and power systems, networks and AI semiconductors will total $10.3 trillion in 2025 to 2032. On an annual average basis, that equals 3.63 percent of U.S. gross domestic product (GDP).

That is higher than past major U.S. infrastructure investment booms. Railroad investment averaged 2.24 percent of GDP in 1870 to 1890, the interstate highway program 1.13 percent in 1956 to 1973, and telecoms and fiber optics about 1.1 percent in 1996 to 2003. The researcher said that if the outlook materialises, AI investment relative to the size of the economy would surpass those investment booms.

A risk factor cited was the financing structure. In addition to direct investment by large companies, financing conducted off balance sheet is increasing, including joint ventures, private credit, securitisation, special purpose vehicles (SPVs) and leasing. These structures are affected by AI demand and technological change, power and semiconductor supply, and the creditworthiness of a small number of data center tenants.

Still, Van Nieuwerburgh said it was too early to conclude that AI infrastructure investment has already created systemic risk comparable to past credit bubbles. He instead pointed out that it could be difficult to identify interconnected risk exposures before an economic downturn, and that there is a need to improve measurement and transparency of funding flows.

An Earth economy is less than a trillionth the size of a K2 economy https://t.co/CiVPMCuGf3