The AI Data Center Boom Is Warping the US Economy (www.wired.com)

Tech giants are pouring unprecedented capital into AI data centers — Microsoft, Alphabet, Meta and Amazon expect roughly $370 billion in 2025 capex, with Microsoft alone spending nearly $35 billion in one quarter. That surge has powered stock-market gains (AI-related names drove ~75% of S&P 500 returns since ChatGPT) and, according to Harvard economist Jason Furman, nearly all U.S. GDP growth in H1 2025. But the frenzy carries financial risks: firms are using accounting and financing maneuvers (SPVs, massive bond sales) and assuming long GPU depreciation lives even though Nvidia refreshes hardware roughly every two years — a faster upgrade cycle would erode profits and could expose a potential AI investment bubble. The buildout also has material technical and societal spillovers. Modern training centers can house tens of thousands of GPUs performing trillions of operations, creating intense cooling and electricity demands that are stretching U.S. grid capacity, pushing utilities to seek nearly $30 billion in rate hikes and prompting warnings from OpenAI about power constraints jeopardizing U.S. leadership. Labor effects are mixed: big tech is cutting corporate roles even as it automates entry-level work (Amazon projects avoiding 160k hires by 2027), while massive capital flows into data centers crowd out investment in other sectors like manufacturing, contributing to softer overall job growth.