Saad Ullah
⬤ Alphabet's cloud division sparked fresh controversy after Friedberg challenged claims about depreciation practices at hyperscaler data centers. Friedberg pointed out that older GPUs and TPUs continue running at full capacity, supporting the longer depreciation timelines companies like $GOOGL currently use.
⬤ Hardware that's 7-8 years old still operates at 100% utilization, Friedberg noted, suggesting extended depreciation reflects actual usage rather than accounting tricks. Google's Cloud GM backs this view, confirming that older-generation chips remain essential in large-scale operations. This ongoing reliance shows hyperscalers keep extracting real value from hardware well beyond initial deployment.
⬤ The discussion sits within broader questions about how hyperscalers handle capital-heavy AI infrastructure, where depreciation directly impacts reported earnings. Burry worried companies might inflate profits by stretching asset lifespans. Friedberg countered that full utilization of aging GPUs and TPUs mirrors genuine usage patterns at cloud providers like $GOOGL, not financial manipulation.
⬤ This exchange reveals how data center accounting shapes understanding of hyperscaler economics. For $GOOGL, the debate highlights how AI workloads and hardware longevity influence depreciation policies. It also reflects mounting scrutiny over cloud spending, earnings quality, and how tech giants balance infrastructure costs with platform demand.
Saad Ullah
