A one-year change in the assumed useful life of AI chips can move Big Tech earnings by $2–$3 billion. That's the accounting sensitivity nobody is talking about.
Everyone is focused on AI revenue. We're looking at the depreciation. The AI build-out is turning Big Tech from asset-light to capital-heavy – and the accounting treatment of those assets is becoming a material driver of reported earnings.
David Rosenberg
Mehmet Beceren
Exclusive research for institutional investors & advisors
Join 10,000+ institutional investors and advisors who read Rosenberg Research.
What You'll Find Inside
“Useful-life” assumptions on AI chips and servers are an accounting lever that can swing reported earnings by billions. When hyperscalers extend depreciation from three years to five, profits look better – until the next hardware cycle forces a write-down.
Big Tech built its premium on asset-light models. The AI capex wave inverts that narrative. Old valuation assumptions look less comfortable as capital intensity rises and earnings quality gets muddied by depreciation choices.
The uncertainty around AI payback strengthens the case for focusing upstream – chip designers, equipment makers, and energy infrastructure – where the earnings path is clearer and less dependent on accounting assumptions.
The Contrarian Angle
For two decades, the most valuable companies in the world built their dominance on capital-light business models – software margins, platform economics, and minimal physical infrastructure. The AI build-out is changing that equation fundamentally. Data center capex is measured in tens of billions, and the accounting treatment of those assets is becoming a material driver of reported earnings.
This represents a structural shift in how Big Tech's earnings should be evaluated. And yet the sell-side is largely treating it as a rounding error. In this report, Mehmet Beceren walks through the numbers, the accounting mechanics, and the investment implications that most analysts are glossing over.
“A one-year change in the assumed useful life of high-end AI chips and servers can move reported earnings by $2–$3 billion for the largest technology companies. That is the scale of accounting sensitivity now embedded in the AI build-out.”
– Mehmet Beceren, VP & Senior Market Strategist, Rosenberg Research
About Rosenberg Research
Rosenberg Research & Associates is an independent macro research firm founded by David Rosenberg, one of North America's most widely followed economists. The firm provides institutional-grade research to portfolio managers, RIAs, family offices, and sophisticated investors worldwide. Known for a data-driven, contrarian approach, Rosenberg Research delivers daily, weekly, and thematic analysis that cuts through consensus thinking.
Download the full “Fast Chips, Slow Books” briefing and see the accounting risk the Street is overlooking.
GET THE REPORT