AI Capex Surge Masks Depreciation Risks as Hyperscalers Shorten Asset Lives Amid Rapid Tech Shifts

Big Tech's AI infrastructure buildout continues at a blistering pace, with Microsoft, Alphabet, Amazon, and Meta collectively guiding toward $600-725 billion in capital expenditures for 2026 alone, up sharply from prior years.[1][2] This spending spree funds data centers, servers, and GPUs essential for training and inference, yet accounting choices around asset useful lives introduce underappreciated pressures on future profitability.

Amazon explicitly shortened the estimated useful life of a subset of its servers and networking equipment from six to five years effective January 1, 2025, citing the 'increased pace of technology development, particularly in the area of artificial intelligence and machine learning.' The change is projected to reduce 2025 operating income by approximately $700 million, with additional accelerated depreciation from early retirements.[3][4][5] While Meta extended certain server lives to 5.5 years and others like Microsoft and Alphabet maintain six-year schedules for computer equipment, analysts and investors including Michael Burry highlight that Nvidia's annual chip cycles and high-intensity AI workloads may render economic lives closer to two-to-three years, potentially understating depreciation by tens of billions industry-wide through 2028.[6][7]

Alphabet has already tapped debt markets for over $85 billion and announced plans to raise up to $85 billion in equity to sustain its $175-190 billion 2026 capex target, underscoring cash flow strains even as operating cash flows approach coverage.[8][9] Historical extensions of server lives from three-to-four years pre-AI to five-to-six years provided earnings tailwinds, but the reversal at Amazon signals recognition that AI-specific hardware obsolesces faster than general-purpose servers. Combined with power, chip supply, and infrastructure constraints, these factors suggest replacement cycles could accelerate, layering ongoing capex atop initial build costs and challenging the narrative of sustainable AI returns.

Mainstream coverage emphasizes demand growth and revenue potential from AI services, yet the depreciation time bomb—evident in rising annual depreciation (nearly doubled for the quartet to $116 billion recently) and selective life adjustments—remains a critical variable for long-term margins.