Core Insights - The era of U.S. hyperscalers generating significant free cash flow while maintaining capital efficiency is coming to an end [1] Group 1: Capital Expenditures - Wall Street consensus estimates project hyperscaler capital expenditures to reach $667 billion in 2026, reflecting a $127 billion increase since the start of the fourth-quarter earnings season and implying a 62% year-over-year growth [2] - Capex is expected to consume approximately 92% of hyperscaler cash flows from operations, a higher share than during the Dot Com Boom [2][3] - Goldman Sachs anticipates further upward revisions to 2026 capex estimates, potentially pushing spending toward $700–$725 billion under optimistic scenarios [5] Group 2: Shift in Capital Allocation - The allocation of cash flows to buybacks has decreased from roughly 43% at the start of 2023 to just 16% today, with collective gross buybacks declining 15% year-over-year in 2025 [4] - Management teams are now prioritizing scale in AI over shareholder distributions, indicating a full-scale reinvestment cycle rather than incremental spending [5][9] Group 3: Future Outlook - Capex growth is expected to peak and begin decelerating in the second half of 2026, suggesting a slower pace of fortifying AI infrastructure [6] - A deceleration in capex growth could provide visibility into a trough in free cash flow, allowing investors to value these companies based on earnings rather than reinvestment narratives [7] - The strategic bet is to spend aggressively today to protect monopoly-like positioning in the future, raising questions about whether future AI-driven revenue and margins will justify current investments [10]