As tech giants pour hundreds of billions into artificial intelligence (AI), an ostensibly mundane accounting practice—depreciation methods for critical equipment like AI chips—has unexpectedly become a focal point of market debate, raising questions about whether companies are leveraging it to flatter earnings.
Recent reports indicate that companies including Meta Platforms, Inc., Alphabet, Microsoft, and Amazon have extended the estimated useful lives of their server and network assets. This accounting adjustment directly reduces current depreciation expenses, thereby boosting reported profits—a move that has drawn scrutiny from some investors.
Renowned short-seller Michael Burry has labeled this practice as "one of the modern era's common frauds," arguing that extending asset lifespans artificially inflates profits and overstates valuations. With massive capital expenditures at stake, such accounting tweaks have heightened market sensitivity.
However, for investors tracking these tech behemoths, the issue may not be a simple matter of right or wrong. While extending depreciation periods does provide an immediate earnings lift, its impact on fundamentals may be marginal. The market’s focus remains on the ultimate returns from AI investments rather than accounting nuances. A deeper discussion has emerged around whether depreciation methodology could distort long-term assessments of AI’s value.
**The "Earnings Magic" of Accounting Adjustments** Depreciation allocates capital investment costs over time in financial statements. By lengthening an asset’s "useful life," companies spread costs over more years, lowering annual expenses and lifting current profits.
Tech giants have indeed embraced this approach. Meta Platforms, Inc., for instance, raised the useful life of most servers and network assets to 5.5 years in 2025, up from 4–5 years previously and just 3 years in 2020. This change reduced Meta’s depreciation expenses by $2.3 billion in the first nine months of 2025—though total depreciation still neared $13 billion against pre-tax profits exceeding $60 billion.
Similar trends appear elsewhere: Alphabet and Microsoft now apply 6-year depreciation for comparable assets, up sharply from 3 years in 2020. Amazon increased its lifespan from 4 years in 2020 to 6 years by 2024, though it trimmed some server and network equipment back to 5 years in 2025. Such multi-billion-dollar accounting shifts naturally invite skepticism.
**Straight-Line vs. Accelerated Depreciation: Which Reflects Reality?** The debate may center less on specific lifespans than on methodology. Most firms use straight-line depreciation, spreading costs evenly across an asset’s life.
Yet for rapidly evolving hardware like NVIDIA’s AI chips, this may misrepresent real value erosion. Data from Silicon Data shows a three-year-old H100 system resells at roughly 45% of its original price—suggesting steeper early-stage value declines followed by stabilization.
Here, accelerated depreciation—front-loading expenses to mirror faster value drops—might better reflect economic reality. Still, analysts note the difference between methods may not be drastic enough to overhaul financial assessments.
**What Should the Market Watch?** Ultimately, financial statements rely on estimates, and depreciation itself is an accounting construct with inherent imprecision. Even for high-demand tech gear, predicting exact useful lives remains challenging.
A more critical red flag for markets could be asset impairments—large write-downs required under accounting rules if values deteriorate significantly. But these typically follow severe stock declines, a scenario far removed from the current dominance of mega-cap tech.
Thus, while depreciation debates rage, they’re unlikely to sway investment decisions decisively. As analysis suggests, if investors someday conclude AI spending is wasteful, the culprit won’t be depreciation timelines. For these tech titans, the ultimate verdict will hinge on whether AI delivers substantial returns—not the accounting methods used along the way.
Comments