The Hidden Depreciation Trap Behind Big Tech’s AI Spending Spree

The scale of investment is staggering, rivaling half of the entire UK government’s annual spending. Yet, these tech giants face hard ceilings, from the physical scarcity of power and water to the growing difficulty of funding such massive builds through debt and equity alone. Alphabet’s recent $85 billion debt raise and plans for an additional $80 billion in equity illustrate the desperation to keep the momentum alive.

Beyond the initial construction, a quieter but more dangerous cost is emerging: rapid depreciation. AI servers, which represent two-thirds of data center build costs, have a lifespan of just three to six years. As innovation accelerates, this hardware becomes obsolete faster than traditional equipment. Annual depreciation across the four firms has already nearly doubled to $116 billion.

Amazon has already signaled the shift by shortening the expected useful life of its data center assets from six years to five, citing the breakneck pace of machine learning development. When Meta, Microsoft, and Alphabet inevitably follow suit, their balance sheets will face further pressure. The industry is currently betting that the infrastructure will eventually pay for itself, but the math is becoming increasingly difficult to justify as replacement cycles shorten and maintenance costs climb.