the AI capex cycle economically unsustainable? GPUs fail every 2-3 years, new chips cost 40% more each cycle, but rental

Unpopular opinion: The AI infrastructure bubble will pop when people realize hyperscalers are trapped in an unsustainable capex cycle. GPUs fail every 2-3 years, new generations obsolete every 18 months at rising costs, but revenue per chip is collapsing. How long before Meta/Microsoft/Google admit they’ve been over-investing?

CryptoAlerta — análise de criptomoedas e mercado em tempo real