Harvard economist Jason Furman calculated that if you strip out investment in information processing equipment and software, GDP growth in the first half of the year would have been a pathetic 0.1 per cent on an annualised basis. In other words, nearly all of the country’s growth came from tech firms building giant data centres stuffed with GPUs.
Furman posted his analysis on 27 September, pointing out that while tech investment accounted for just 4 per cent of GDP, it made up 92 per cent of the growth. He noted that the economy might have seen a bit more movement if the AI boom had not inflated energy prices and borrowing costs, but that would only make up “about half of what we got from the AI boom.”
The surge has been driven by the usual suspects such as Microsoft, Google, Amazon, Meta and Nvidia, all pouring tens of billions into facilities built to power large language models and other AI workloads.
Morgan Stanley Wealth Management, chief investment officer Lisa Shallet said hyperscaler capital expenditure has “risen fourfold and is nearing $400 billion annually.” She warned the pace and concentration of that spending are now distorting the wider economy.
Meanwhile, everyone outside tech appears to be treading water. Manufacturing, real estate and retail all barely moved the needle.
Apollo Global Management, chief economist Torsten Sløk said economists have been “wrong since January” in predicting a slowdown that never arrived, joking that the profession needs to “look ourselves in the mirror.”
The disconnect between anaemic job growth and booming GDP has become what Morgan Stanley, chief economist Michael Gapen calls “the mystery” of the 2025 economy. He believes it reflects companies cutting labour costs instead of raising prices, while tech investment masks the weakness elsewhere.
The line that best captures the whole mess came from Rusty Foster of Today in Tabs: “Our economy might just be three AI data centres in a trench coat.”
Even Amazon founder Jeff Bezos has weighed in, describing the current frenzy as an “industrial bubble,” not a financial one. He insists the mountain of money pouring into compute infrastructure will prove worthwhile eventually. Whether it props up the wider economy for long enough to prove him right is another matter entirely.
