Big tech companies, including Microsoft, Oracle, Meta Platforms, and Alphabet, are projected to book over $680 billion in depreciation charges over the next four years, according to Morgan Stanley. This substantial figure is attributed to increased capital expenditures for AI infrastructure and a historical tendency to lengthen the assumed useful lives of their servers and network equipment. Michael Burry, known as "The Big Short," has criticized this practice, calling earnings inflation from lowballed depreciation charges "one of the most common frauds of the modern era."
This issue arises because these companies, historically asset-light, need to adapt their financial reporting to reflect their new capital-intensive business models. Morgan Stanley's analysis shows that Alphabet's depreciation expense could quadruple by 2028, and Oracle's 2025 depreciation charge of $4 billion might surge to $56 billion by 2029, potentially representing 28 percent of consensus revenue. The problem is exacerbated by the fact that depreciation is a non-cash cost, and changes to useful life assumptions, like Alphabet's $3 billion boost to earnings guidance in 2023 by extending data center equipment longevity, significantly impact GAAP income.
Analysts face challenges in accurately forecasting these impacts because tech companies rarely detail depreciation in their income statements. Morgan Stanley's Accounting & Tax desk takes a granular approach, splitting capex into AI and non-AI components and assuming GPUs have a useful life of up to six years and warehouses 15 years. However, this is complicated by Nvidia releasing new chip architectures roughly every three years, making Alphabet's six-year useful life assumption for equipment appear optimistic. The delay between capital outlay and depreciation, particularly for data centers taking years to build, further distorts financial reporting.
This looming depreciation challenge could significantly impact the operating margins of hyperscalers, which The Street generally expects to improve over the next four years (except for Oracle). Meta Platforms, for instance, projects its total expenses to be up to 44 percent higher this year, between $162 billion and $169 billion, and 2026 capex to increase by up to 94 percent, ranging from $115 billion to $135 billion. Microsoft also plans to roughly double its data center footprint in the next two years, indicating further substantial capital outlays. The Financial Accounting Standards Board has issued new rules on income statement expenses, effective in 2027, which may improve transparency.
The combined impact of these factors suggests that unless revenue forecasts increase significantly, depreciation will soon dominate total expenses for these tech giants. The uncertainty surrounding revenue opportunities and the durability of GPUs in the early stages of the AI investment cycle, coupled with the increasing use of finance leases for AI infrastructure, makes accurate forecasting difficult. The comparison between Morgan Stanley and Michael Burry's concerns points to potential market overvaluations that are not bullish.
