How Valuable Are AI Chips? The Investor Focus: Depreciation

Global tech giants plan to invest $1 trillion over the next five years to build AI data centers. One key financial consideration stands out: depreciation.

Depreciation spreads the cost of fixed assets over their useful life. In the AI era, this concept is critical: the lifespan and value retention of hundreds of thousands of NVIDIA GPUs directly affect profit and financial planning. Companies like Google, Oracle, and Microsoft estimate servers may last up to six years, but depreciation cycles can range from two to six years. For investors and lenders, the longer an asset retains value, the smaller the depreciation impact on profits.

The Challenge of AI GPUs
AI GPUs are relatively new, lacking long-term usage benchmarks. Since NVIDIA introduced its first data center AI processor in 2018, the AI boom exploded after ChatGPT’s release in late 2022. NVIDIA’s data center revenue jumped from $15 billion to $115 billion as of this fiscal year.

Some customers believe AI chips hold long-term value; older models remain useful. CoreWeave, which rents GPUs, has set a six-year depreciation cycle since 2023. CEO Michael Intrator notes A100 chips are fully booked, and 2022 H100 chips were rebooked at 95% of original price, indicating sustained infrastructure value.

Yet the market is cautious. CoreWeave shares plunged 16% after earnings, reflecting concern over AI overspending. Oracle shares have fallen 34% since September highs. Short-seller Michael Burry claims Meta, Oracle, Microsoft, Google, and Amazon overstate AI chip lifespan and understate depreciation, with real server life closer to 2–3 years.

Hoppers Chips Depreciation Reminder
AI chips depreciate through wear or replacement by new generations. NVIDIA CEO Jensen Huang joked, “When Blackwell chips ship in volume, Hoppers chips won’t be worth anything.” With annual chip updates and AMD following closely, rapid tech iterations make depreciation unpredictable.

Microsoft CEO Satya Nadella highlights spreading AI chip purchases to avoid over-investment in one generation, noting new chips compete with their predecessors. Depreciation estimates must consider technology obsolescence, maintenance, historical life, and engineering analysis, all audited carefully for financial accuracy.

In the AI investment surge, depreciation cycles are a core measure of chip value retention.