The AI-driven data centre spending surge is creating both acute shortages and looming overcapacity risks. Hyperscalers like Alphabet, Amazon, Meta, Microsoft, and Oracle are set to spend $800bn, exceeding their combined cash flow. GPU rental costs have risen 50% in six months, with Nvidia B200 chips costing over $6/hour.
SpaceX claims its AI infrastructure investments pay back in under a year. However, Brookings Institution estimates AI services must account for 9.2% of US GDP by 2032 to justify returns. Gartner forecasts $8.5tn in needed revenue versus $3.6tn projected.
AI cloud startups like Core Weave and Nebius face volatility. While productivity gains could support demand, the AI miracle remains largely unproven. Political backlash, power shortages, or rising interest rates may slow construction and mitigate overbuilding risks.
Source: Financial Times Companies · Summarized by HeadlinesBriefing