Robin Wigglesworth explores the massive capital expenditure by AI hyperscalers and whether their investments will be profitable. Goldman Sachs estimates that US hyperscalers will spend $800 billion in 2026 and $1.1 trillion in 2027 on AI infrastructure. To break even, these companies need to generate approximately $300 billion in annual AI revenues, which is about four times the current run rate. For a 30% return on invested capital, revenues must reach $636 billion.

While cloud revenue backlogs exceed $1.5 trillion, the sustainability of these figures remains uncertain. The article also notes that AI software applications on top of this infrastructure will require additional spending, potentially pushing total consumer and enterprise AI spending to nearly $2 trillion.