Enterprise AI pricing pressures mount as annualized hyperscaler + ecosystem capex approaches $1T, forcing revenue recovery through higher end-user costs

Josh Bersin's analysis argues that the aggregate 2026 run-rate for AI infrastructure investment, Big 4 hyperscalers (~$650B), plus Oracle/CoreWeave/xAI (~$500B annualized), plus semiconductor suppliers (Nvidia, TSMC, Micron, SK Hynix, ~$200–300B), is approaching $1T, requiring enterprise software and AI service prices to roughly double to generate a 15% compound return on a five-year depreciation schedule. Anthropic has already shifted Claude enterprise pricing from flat fees to consumption-based billing, and CIOs report bracing for materially higher and volatile AI spend; at least one firm is reconsidering offshore substitution. For financial institutions, this dynamic tests whether enterprise AI adoption produces durable ROI at higher unit costs, or whether rising prices slow deployment and compress the recognized-revenue growth story underpinning AI-exposed equity valuations.