Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
74 67 80 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
62 43 81 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 1 promise resolved, 5 needed for a score.
All types combined: 73 (likely 66 to 79, high confidence), from 72 claims: 66 factual claims, 6 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Michael Burry's claims
14 claims
“This caused additional expense across 9 months of $677 million.”
Michael Burry, Substack author 9 July 2026 Short Thoughts July 8, 2026 - NVDA, Neos, Hyperscalers, Jevons Paradox, and Compression (Cassandra Unchained)
Amazon's shortening of server lives in early 2025 caused additional expense of $677 million across 9 months.
Mixed Partly true and partly false in roughly equal measure.
What would count as true or false
True: Amazon's SEC filings (a Form 10-Q or 10-K, as amended or restated if applicable) disclose that the early-2025 shortening of estimated server useful lives raised depreciation or expense by $677 million (within rounding) over a nine-month period. Mostly true: the disclosed nine-month effect is within about 5% of $677 million, or the figure matches but the change took effect at a slightly different time or the figure is net of a related item. Mixed: Amazon disclosed an effect from the useful-life change, but $677 million refers to a different period (for example, a quarter or a full year) or a different effect (for example, net income rather than expense). Mostly false: there was a real useful-life change, but the stated nine-month figure is off by more than about 5% and up to about 25%, or it misattributes a different charge. False: Amazon made no such early-2025 shortening of server lives, or the disclosed effect differs from $677 million by more than about 25%.
Evidence
- amzn-20250930 (Amazon.com, Inc. Form 10-Q for the quarter ended September 30, 2025) U.S. Securities and Exchange Commission (filed by Amazon.com, Inc.)
“and those acquired during the nine months ended September 30, 2025, was an increase in depreciation and amortization expense of $889 million and a reduction in net income of $677 million,”
- AI: How Long Does a Chip Last? Token Runner · 3 October 2026
“That cost it $889 million more depreciation in the first nine months of 2025, $392 million of it in the third quarter alone.”
- Train, Deploy, Write Down AI Realist (airealist.ai) · 7 April 2026
“The nine-month impact through September 2025 was an increase in depreciation expense of $889 million.”
- Amazon Extended Server Life Twice, Then Reversed It. The Reason They Gave Was AI. ($AMZN) Hudson Labs · 30 July 2026
“was an increase in depreciation and amortization expense of $1.4 billion and a reduction in net income of $1.0 billion, or $0.10 per basic share and $0.10 per diluted share, which primarily impacted our AWS segment.”
Reasoning
I rated this mixed because the $677 million figure is real but describes a different effect than the claim says. Amazon's 10-Q for the quarter ended September 30, 2025 [E1] puts the nine-month effect of the useful-life change at an $889 million increase in depreciation and amortization expense and a $677 million reduction in net income. The claim calls the $677 million additional expense, but the filing ties that figure to net income. Two secondary sources [E2, E3] agree that the nine-month increase in depreciation expense was $889 million. The resolution criteria count a matching figure that measures net income rather than expense as mixed.
Cite this claim
Michael Burry, “Amazon's shortening of server lives in early 2025 caused additional expense of $677 million across 9 months.” (said 9 July 2026). Verdict: mixed. Reliability API, claim c_6514151b7695, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_6514151b7695
Claims without a verdict
Is a verdict here wrong, or do you speak for Michael Burry? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.