Nvidia's Accounting Question: Cloud Companies Reclassifying GPU Spending as Revenue-Generating Assets
A growing industry debate examines whether cloud providers' classification of AI GPU purchases as long-lived capital assets — rather than short-lived equipment — accurately reflects the rapid depreciation of AI hardware.
Key Takeaways
Cloud companies are reclassifying GPU spending from operating expenses to long-term capital assets, extending their 'useful life' to boost revenue metrics. The accounting practice has raised questions on Wall Street about whether it accurately reflects the rapid depreciation of AI hardware.
A financial accounting question is generating significant debate among Wall Street analysts and corporate governance experts: how should the hundreds of billions of dollars that cloud computing companies spend on Nvidia GPUs and other AI accelerators be classified on their balance sheets? The answer has material implications for reported earnings, asset valuations, and investor assessments of AI infrastructure profitability.
The Classification Debate
At the center of the debate is the 'useful life' assigned to AI hardware. Cloud providers including Microsoft, Google, and Meta have been extending the depreciation periods for their data center equipment — in some cases from three years to five or six years. This means the cost of GPU purchases is spread over a longer period, resulting in lower annual depreciation charges and higher reported earnings.
Critics argue that extending depreciation periods for AI hardware may not reflect economic reality. AI chip performance improves rapidly with each generation, and today's cutting-edge GPUs may become relatively obsolete long before their extended depreciation period expires. If companies are forced to write down or replace equipment before it is fully depreciated, the accounting adjustments could create sudden earnings hits.
The Scale of the Issue
With total industry AI capital expenditure projected at approximately $670 billion for 2026, even small changes in how these investments are accounted for have enormous implications. A one-year change in average depreciation period for the combined AI capex of the top four cloud providers alone could shift tens of billions of dollars in reported profits.
Why Some Analysts Say It's 'Overblown'
Not all analysts view the accounting question with alarm. Several have argued that the concerns are 'overblown,' noting that cloud providers' AI hardware generates substantial revenue through cloud services — meaning the assets are genuinely productive over extended periods, even if newer hardware offers better performance. Under this view, the extended depreciation periods reflect a legitimate business reality: companies don't replace functional hardware simply because newer models exist.
The debate is unlikely to be resolved quickly. As AI infrastructure spending continues to accelerate and hardware generations turn over, the tension between accounting conventions and technological reality will remain a critical issue for investors evaluating the profitability of AI infrastructure investments.