Silicon Valley is spending over $200 billion on AI and losing money doing it

Alphabet, Microsoft, Meta, and Amazon are collectively spending over $200 billion a year on AI infrastructure and watching their margins compress in real time. Capital expenditure is money spent on physical assets: data centers, servers, and the specialized chips needed to train and run AI models. It shows up on the balance sheet as an asset, but it hits cash flow immediately. Alphabet, Microsoft, Meta, and Amazon have guided investors toward a combined annual capex figure that exceeds $200 billion, with the majority earmarked for AI hardware and cloud infrastructure. Meta has gone further than most, substantially increasing its 2026 capex outlook specifically because of purchases of AI accelerator chips. Those purchases flow through as operating expenses, which means they directly erode reported profits in the near term. Microsoft and Google Cloud have reported higher operational costs tied directly to AI investments, producing visible margin impacts in their earnings reports. Analysts at Goldman Sachs and Bernstein have flagged that the projected payback periods for these AI investments stretch into 2027 and 2028. Companies spending aggressively today should not expect to see tho...