The question everyone in AI is asking: How long before a GPU depreciates?

The rapid obsolescence of AI hardware, particularly GPUs, is creating a critical accounting challenge for tech giants planning $1 trillion in data center investments, with significant implications for reported profits and financing. While some companies, like CoreWeave, are depreciating assets over six years, citing continued value retention, skeptics including short-seller Michael Burry and Nvidia's CEO Jensen Huang argue for a much shorter 2-3 year useful life due to the accelerated pace of technological advancements. This divergence is already impacting market sentiment, with companies like Amazon adjusting depreciation schedules and others experiencing stock declines amid concerns over AI overspending.

Analysis

The rapid obsolescence of AI hardware, particularly GPUs, presents a significant accounting challenge for tech giants investing $1 trillion in data centers, impacting depreciation schedules and reported profitability. Amazon's reduction of server useful life from six to five years underscores the accelerated pace of technological development, creating a debate over asset longevity and earnings quality.

While CoreWeave maintains a six-year depreciation cycle for its Nvidia chips, citing strong demand, market sentiment is cautious; its stock plunged 16% post-earnings on guidance delays and is down 57% from its high. Oracle shares also plummeted 34% from their peak, reflecting broader concerns about AI overspending and potential earnings inflation, as suggested by short seller Michael Burry.

This skeptical view is reinforced by Nvidia CEO Jensen Huang's comments on rapid chip value deterioration with new annual releases, and Microsoft CEO Satya Nadella's caution against being "stuck with one generation." The divergence in depreciation estimates creates significant financing challenges and necessitates rigorous auditor scrutiny, influencing valuations and capital allocation decisions across the AI infrastructure sector.

Key Decisions for Investors

• Investors should scrutinize company depreciation policies for AI hardware, comparing stated useful lives against the accelerated pace of technological advancement and industry expert opinions.
• Monitor the financial health and guidance of companies heavily invested in AI infrastructure, particularly those experiencing stock declines like CoreWeave and Oracle, for signs of overspending or asset impairment risks.
• Evaluate the long-term sustainability of AI-driven revenue growth, considering the potential for rapid obsolescence to impact future profitability and the need for continuous, significant capital expenditure.