WireSift
Findings · AI reality check

How long does an AI server live? The hyperscalers can’t agree, and billions in profit ride on it.

How long a company assumes its servers last decides how fast it depreciates them, and therefore how much profit it reports. Since 2022, Microsoft, Amazon, Alphabet, Meta and Oracle all stretched those assumed lives toward six years, each booking a one-time earnings tailwind of roughly 4 to 5% of net income. Then Amazon did something the others have not: it shortened a subset of servers back to five years, citing the pace of AI hardware turnover. Same hardware, opposite calls. Here is the full record, from the filings.

5
hyperscalers, one accounting question
$8-9B+
one-time net-income tailwind from life extensions
~4 to 4.6%
of net income, each in its change year
1 of 5
has reversed course (Amazon), citing AI turnover
Assumed server useful life (years) · 5 hyperscalers · as of 2026-05-24

Four extended toward six years and held. Amazon alone stepped back.

A longer assumed life means less depreciation and higher reported profit. Amazon stretched its servers out with everyone else, reaching six years, then in 2025 cut a subset back to five, citing the pace of AI hardware turnover.

4.5420214.5520225.5520235.562024652025Others extend toward 6Amazon steps back ↓
The other four (Microsoft, Alphabet, Meta, Oracle), averageAmazon

Source: company 10-K filings (property-and-equipment useful-life disclosures). The four are shown as their rounded average; exact per-company values are in the ledger below. WireSift Research.

Share / save
By the numbers · trailing 12 months

Capex is running nearly 4x depreciation. The assumption sets how fast that closes.

Measured over the same recent twelve months, so the five are comparable. They are capitalizing AI hardware far faster than they expense it. The longer the assumed life, the slower depreciation catches up, and the higher reported profit stays in the meantime.

AmazonAMZN
Capex$151B
Deprec$46B
NI$91B
Pre-tax$115B
MicrosoftMSFT
Capex$97B
Deprec~$30B*
NI$125B
Pre-tax$155B
AlphabetGOOGL
Capex$110B
Deprec$23B
NI$160B
Pre-tax$194B
MetaMETA
Capex$76B
Deprec$20B
NI$71B
Pre-tax$89B
OracleORCL
Capex$48B
Deprec$6B
NI$16B
Pre-tax$18B

Across the five, trailing-12-month capex of ~$482B is nearly 4x their ~$125B of PP&E depreciation. Stretch the assumed life and that gap, and reported profit, both widen.

Trailing twelve months (trailing 12 months ending ~Mar 2026 (Oracle ~Feb 2026)), summed from each company's 10-Q and 10-K filings. PP&E depreciation; Amazon includes finance-lease assets. *Microsoft discloses PP&E depreciation only rounded, so its figure is estimated. Alphabet and Meta period earnings include large non-operating one-offs (an investment gain and a one-time tax charge, respectively).

The ledger

Every change, every disclosed dollar.

Each company’s server useful-life history and the earnings impact it disclosed, with a link to the filing. All figures are from the companies’ own 10-Ks.

AmazonAMZN · FY Dec 31
6 yrs (subset 5)
4→5 (Jan 2022) · 5→6 (Jan 2024) · 6→5 subset (Jan 2025)
Jan-2022: +$2.8B benefit to net loss ($0.28 EPS). 5→6 (FY2024): +$2.5B net income ($0.23 EPS). 6→5 subset: −$0.7B 2025 op income + ~$920M Q4’24 accelerated depreciation. [FY2024 10-K]
Extended, then reversed
MicrosoftMSFT · FY Jun 30
6 yrs
4→6 (FY2023), held
FY2023: +$3.7B operating income / +$3.0B net income / +$0.40 EPS. [FY2023 10-K]
Extended, held
AlphabetGOOGL · FY Dec 31
6 yrs
4→6 / 5→6 (FY2023), held
FY2023: −$3.9B depreciation / +$3.0B net income / +$0.24 EPS. [FY2023 10-K]
Extended, held
MetaMETA · FY Dec 31
5.5 yrs
→5.5 (Jan 2025), held
FY2025: −$2.92B depreciation / +$2.59B net income / +$1.00 EPS. [FY2025 10-K]
Extended, held
OracleORCL · FY May 31
6 yrs
4→5 (FY2023) · 5→6 (FY2025), held
FY2025 (5→6): −$733M opex / +$573M net income / +$0.20–0.21 EPS. (FY2023 4→5: −$434M opex.) [FY2025 10-K]
Extended, held
What’s at stake

Five companies, five accounting teams, one number: ~4 to 5% of profit.

In the year each company extended its server lives, the change lifted net income by a strikingly consistent ~4 to 4.6%: Microsoft +$3.0B (4.1%), Alphabet +$3.0B (4.1%), Amazon +$2.5B (4.2%), Meta +$2.59B (4.3%), Oracle +$0.57B (4.6%). Real money, and a real single-year boost to reported earnings per share.

But notice what that is and is not. It is a one-time, disclosed tailwind in a single year, around 4 to 5% of profit. It is not a claim that earnings are fabricated, and it is an order of magnitude smaller than the cumulative figures circulating in the bubble debate.

The claim vs. the filings

Michael Burry has accused the hyperscalers of inflating earnings by roughly $176B over 2026 to 2028, putting Meta at ~21% and Oracle at ~27% overstated. That is his own forward projection, and it does not match what the filings disclose: none of these companies break out a separate GPU asset class, and the disclosed, single-year impacts run around 4 to 5% of profit, not 20-plus. The accusation points at a real lever. It overstates the size of it.

Why it still matters (and where it doesn’t)

Depreciation is non-cash, so none of this changes free cash flow, and these companies are valued largely on cash. So why care? Because reported earnings and margins drive multiples, headlines, and comparability, and right now the five are not comparable: they are assuming different lifespans for the same hardware. The open question is not whether anyone is lying. It is which assumption is right when one of them, Amazon, is already telling you in its own filings that the hardware ages faster.

Every figure is drawn from the companies’ own 10-K filings (property-and-equipment useful-life disclosures and the disclosed earnings impact of each change). Useful lives are assumptions, not facts about hardware; this page documents what each company assumes and discloses, and does not assert a single correct lifespan. See the methodology for sourcing standards.

Stay informed

We check the AI boom’s reported reality against the filings.

Sourced, structured research on what companies actually disclose about AI, before it makes the headlines.

By subscribing, you agree to our Privacy Policy.