AI boom is unsustainable unless tech spending goes 'parabolic' (fortune.com)

Deutsche Bank warned that the current AI boom may be unsustainable because tech capital spending—particularly on data centers, GPUs and the massive power infrastructure that hyperscalers need—cannot “remain parabolic.” The bank says AI-related capex is large enough to be propping up U.S. growth this year; Goldman Sachs estimates AI capex hit roughly $368 billion through August. Bain & Co. adds that meeting projected AI compute demand by 2030 requires about $2 trillion in annual revenues to fund that capacity, leaving an estimated $800 billion shortfall even after anticipated AI-driven savings. Deutsche Bank singled out NVIDIA and other hardware suppliers as bearing a disproportionate share of the investment cycle. The practical risk: GDP gains today are coming from building AI factories, not necessarily from broad-based productivity or end-user AI revenues, so a slowdown in infrastructure spending could reveal the boom’s fragility. Markets already show extreme concentration—about half of the S&P 500’s gains this year are tied to the Magnificent 7—and analysts warn of overexposure if capex cools. Counterpoints exist: Goldman expects meaningful productivity lifts (roughly +0.4% near term and ~1.5% cumulatively long term) as AI adoption diffuses. The debate centers on whether AI will transition from an infrastructure-driven investment cycle to widespread, revenue-generating productivity gains before spending tapers.