One year, $800 billion
Depreciation is how companies spread the cost of an asset over its years of use. The longer the depreciation period, the lower the annual expense and the higher the reported profit. Given the hundreds of billions of dollars that tech giants are spending, a small change in this assumption can make an enormous difference.
According to Breakingviews, shortening the depreciation period by 12 months would raise expenses and cut profits, and, applied to current valuations, could wipe out nearly $800 billion of the four groups' market value. The author argues that skepticism is warranted given the faster-than-ever pace of technology upgrades.
Five years or six: a contested number
According to company filings compiled by various analysts, the tech giants' server depreciation policies currently look like this:
- Alphabet depreciates servers and network equipment over about six years, up from three years in 2020.
- Microsoft discloses a range of two to six years for servers and network equipment, having extended to six years from fiscal 2023.
- Meta extended the useful life of most of its servers to 5.5 years from early 2025, cutting about $2.9 billion of depreciation expense that year.
- Amazon went against the trend, shortening a subset of servers from six to five years in 2025, citing the rapid pace of development in AI and machine learning.
The fact that Meta and Amazon moved in opposite directions at the same time, for the same type of asset, is precisely what leads analysts to question the objectivity of these assumptions.
When chips age faster than the books
Nvidia now launches a new chip architecture almost every year, each generation significantly more powerful and energy-efficient. That means processors bought a few years ago quickly lose their competitive edge.
According to Fortune, Chris Brightman, a partner at Research Affiliates, warns that AI equipment can become nearly worthless in as little as three years.
“The economic life of AI hardware is a lot shorter than its accounting life.” — Chris Brightman
According to analyses compiled from company filings, investor Michael Burry, famous for predicting the 2008 housing crash, has estimated that depreciation at the big cloud providers could be understated by about $176 billion between 2026 and 2028.
The defenders' case
Not everyone agrees with the gloomy scenario. Defenders argue that older chips can still be used for less demanding inference tasks or leased out, so their real working life may be longer than product launch cycles. Moreover, if assumptions prove wrong, companies will have to adjust depreciation or book losses on early retirements, as Amazon has done.
Some accounting commentators note that depreciation is only an estimate, and investors already know that the real returns from AI lie in the future. In their view, if the market one day concludes that much AI investment was wasted, the cause will not be the depreciation schedules, but business performance itself.
What it means for investors
The pressure is growing as capital spending soars. According to figures compiled from financial statements, the four groups spent about $434 billion on property and equipment in the four quarters through March 2026, nearly three times their reported depreciation. Alphabet alone expects 2026 capital spending of about $195–205 billion, with a sharp rise in 2027. Even Microsoft CEO Satya Nadella has acknowledged that the company deliberately spaces out AI chip purchases to avoid the risk of being stuck with outdated hardware.
“Stuck with four or five years of depreciation on one generation.” — Satya Nadella
For investors, the lesson is to look beyond earnings per share and pay attention to free cash flow and the revenue generated per dollar of capital spending. As the gap between accounting life and economic life narrows, adjustments could arrive suddenly and expensively.

















