In his article "How AI Conquered the US Economy—and What Happens Next" Derek Thompson uses a historical yardstick to place the AI investment boom within the panorama of U.S. economic development, giving us a more comprehensive understanding of these figures. In Q2 and Q3 of 2025, tech giants' capital expenditure on AI data centers has reached $400 billion, nearly $100 billion per quarter, accounting for approximately 1.36% of U.S. GDP during the same period. This figure not only exceeds the U.S. government's annual budget for education, employment, and social services, but can also be compared to the entire European defense spending.
Thompson emphasizes that AI-related capital expenditure approaches 2% of U.S. GDP. While the investment intensity has not yet reached the peak of the 19th-century railroad boom (in terms of GDP proportion, it's currently about 20% of the railroad investment peak), it has far surpassed the historical levels of the internet bubble and telecommunications investment.
More remarkably, in the first half of 2025, AI capital expenditure's marginal contribution to GDP growth rate surpassed consumer spending, the traditional engine, for the first time, becoming the biggest driver of economic expansion this year.
These funds primarily come from tech giants' free cash flow, capital market financing, and high investor expectations for the AI sector. Microsoft, Amazon, Google, Meta, and other companies have bet almost all their profits and financing on data center and AI computing power expansion, creating a concentrated flow of capital to very few sectors. Meanwhile, traditional manufacturing and consumer-oriented startups face cold financing conditions, with the economy's "diversified engines" being replaced by a "computing power monopoly."
Can these investments justify their returns? Not yet. AI hardware suppliers like NVIDIA have achieved doubled revenue and profit growth in 2025, driving related companies' market values to soar. The rises in the S&P 500 and NASDAQ are almost entirely driven by the "AI Seven Giants," while traditional consumer and manufacturing sectors show weakness. However, this income structure is highly dependent on the continued expansion of AI computing power demand. Once application-layer innovation or demand expectations cool down, related sectors may face dramatic adjustments in revenue and market value.
The employment landscape similarly shows structural bias. In May and June 2025, U.S. non-farm employment increases were only 19K and 14K respectively, both experiencing significant downward revisions, while July's new employment was only 73K, of which healthcare contributed 55K, accounting for 75%. Except for healthcare and social services, other industries showed almost zero or negative growth, with employment growth increasingly dependent on a few sectors. Meanwhile, the talent war in the AI field has reached an "NBA level." Meta has offered compensation packages worth hundreds of millions or even billions to top researchers, far exceeding the compensation levels of historical national technology projects such as the Manhattan Project and Apollo moon landing. These sky-high salaries not only make AI researchers worth more than most NBA stars but also highlight Silicon Valley's crazy bet on "superintelligence."
Finally, a bonus point: since the rise of LLMs, academic research has undergone tremendous changes. In 2024, the occurrence of the word "delves" exceeded the historical average by 2700%... It's estimated that 1/7 of abstracts have been processed by AI (including this article, of course)...
Regardless of the outcome of this investment boom, this is definitely no longer just a game that only affects the capital markets...
== Sources ==
- US AI Capital Expenditure and GDP Contribution Data for Q2-Q3 2025
- Fortune: https://fortune.com/2025/08/06/data-center-artificial-intelligence-bubble-consumer-spending-economy/
- Phillips & Company: https://phillipsandco.com/blog/the-trump-economys-tightrope-booming-tech-limping-labor/
- Sherwood News: https://sherwood.news/markets/the-ai-spending-boom-is-eating-the-us-economy/
- Historical comparison and bubble risk discussion
- Latest U.S. non-farm employment data
