Internet companies are in denial about getting fat. The advertising silos and data miners of 10 years ago are now infrastructure-heavy and capital-intensive, but their reporting has yet to adjust to the rapid weight gain.
This mismatch may become harder to ignore as the value of their spending is written off. Morgan Stanley forecasts that, collectively, Microsoft, Oracle, Meta Platforms and Alphabet could book more than $680bn in depreciation charges over the next four years:
Its estimates are even higher than those put forward by Michael “The Big Short” Burry, who calls earnings inflation by lowballed depreciation charges “one of the most common frauds of the modern era”.
The principle of depreciation is to spread the purchase cost of a physical asset over its useful life. Figuring out the practicalities of the principle is never easy, but it’s especially difficult when companies are expanding rapidly towards the unknown. As Morgan Stanley says:
Depreciation is challenging to forecast because it depends on the timing of data center construction completion and the useful life assumptions of both the shell and the chips. These assumptions are subject to higher uncertainty given the early stage of this emerging technology. Furthermore, major players, Alphabet, Meta Platforms, Microsoft, and Oracle are increasingly turning to finance leases to build out their AI infrastructure, also driving expenses higher. [ . . . ]
Because these companies have historically been asset-light businesses, their financial disclosures are not adequately set up for investors to see where depreciation is reported in the income statement, and most consensus models do not directly forecast adequate impact on earnings.
The first problem is hardware. Hyperscalers have since 2020 been lengthening the assumed useful lives for their servers and network equipment, with only Amazon in 2025 going the other way. The below chart simplifies things but shows the general direction of travel:
Though depreciation is a non-cash cost, the money having already been spent, any change to useful life assumptions has a big effect on GAAP income. For example, Google owner Alphabet raised earnings guidance in 2023 by $3bn by increasing the longevity of data centre equipment by a year or two.
Alphabet’s assumption that equipment has six years of useful life may appear at odds with Nvidia’s release of a new chip architecture approximately every three years. The reasons for its optimism are impossible to know because Alphabet, in common with peers, gives very little information about depreciation in 10-K disclosures.
A second problem also relates to timing. The value of a data centre under construction is held on a company’s balance sheet, but depreciation is only being applied after it becomes operational. Data centres take years to build, so the delay between a capital outlay and a net income deduction can be very long.
Also: capex adds to a company’s book value, showing up immediately in the cash flow statement as investments in property, plant and equipment, without a parallel increase in depreciation expense on the income statement. Comparing the former with the latter will exaggerate average asset longevity.
Measuring the size of the problem is hard. Microsoft doesn’t even give a figure for construction in progress. For the rest, grand data centre projects with uncertain switch-on dates make forecasting net income a crapshoot:
Because tech companies rarely break out depreciation in their income statements, analysts often take a blunt approach and model the effect through cash flow, then fiddle with their operating margin estimates until everything hangs together. During periods of high investment in assets whose useful lives are very different, namely GPUs and warehouses, the analyst has a very high chance of being wrong.
Morgan Stanley’s Accounting & Tax desk, led by analyst Todd Castagno, takes a more granular approach. The team splits hyperscaler capex into AI and non AI, smooths out construction-in-progress costs, and allocates the expected lease expenses to property. The team assumes GPUs have a useful life of up to six years and that warehouses will last for 15 years.
Based on Morgan Stanley’s calculations, Alphabet’s depreciation expense could quadruple by its 2028 year-end. Oracle’s 2025 depreciation charge of $4bn might balloon to $56bn by 2029, which would be equivalent to 28 per cent of revenue expected by the consensus:
Unless forecasts move significantly higher, depreciation will soon come to dominate total expenses:
Those sorts of numbers challenge The Street’s assumption that, with the exception of Oracle, hyperscaler operating margins will improve over the next four years.
To gauge the potential hit to profitability, Morgan Stanley compares consensus revenue forecasts against operating expenses excluding depreciation. Its figures suggest that, to deliver what’s expected, hyperscaler costs ex depreciation need to collapse:
So far, hyperscaler costs have been doing the opposite of collapsing. Meta said overnight that its total expenses would be up to 44 per cent higher this year at between $162bn and $169bn, and that 2026 capex would be higher by up to 94 per cent at between $115bn and $135bn. Microsoft also raised capex materially, saying it intends to “roughly double our total datacenter footprint over the next two years”.
Disclosure of where all this spending is going will have to improve eventually. The Financial Accounting Standards Board last year issued new rules around reporting income statement expenses that will come into force in 2027.
Until then, big tech can keep kicking capex recognition forward in the hope of an AGI-shaped miracle. Making the numbers add up “depends on both revenue opportunities and durability of GPUs, both of which are highly uncertain at this early stage of the AI investment cycle,” Morgan Stanley tells clients.
We’re not sure what it means when Michael Burry and Morgan Stanley are seeing the same market bubble, but it’s probably not bullish.