UBS forecasts global AI capital expenditure will reach $998 billion in 2026 and $1.45 trillion in 2027, with memory costs dominating spending growth. Memory expenses are projected to surge from $71 billion to $923 billion by 2027, comprising 64% of total AI capex, while non-memory spending actually declines, signaling a shift toward semiconductor constraints rather than physical infrastructure expansion.
Twitter discussion on AI capital expenditure trends, including debate over frontier labs' CAPEX investments versus AI speed regulation concerns, and Goldman Sachs' projections for Micron's revenue growth driven by AI infrastructure demand for high-bandwidth memory.