The debate over Big Tech AI capex often starts with free cash flow pressure and debt issuance. A second issue may prove just as important for valuations: the lagged effect of depreciation on Return on Invested Capital (ROIC). Cash spent on servers and data centers is generally capitalized first, rather than expensed immediately. That means reported earnings can look more resilient during the buildout phase even as cash has already left the business.
The Accounting Lag Shielding Current Earnings
Alphabet offers a concrete example of how useful-life assumptions can shape reported profitability. In Alphabet Inc.’s Form 10-K for the fiscal year ended December 31, 2023, the company disclosed that it extended the estimated useful lives of its servers from four to six years, and its network equipment to six years.
That accounting change reduced Alphabet’s 2023 depreciation expense by $3.9 billion and increased net income by $3.0 billion, or $0.24 per diluted share. The economic point is not that the cost disappeared. It is that the timing of expense recognition changed. Longer useful lives spread hardware costs over a wider period, supporting near-term net income while the underlying capital intensity continues to build.
For investors tracking Big Tech AI Capex Free Cash Flow Pressure, this distinction matters. Free cash flow absorbs capital spending immediately. GAAP earnings absorb the same investment more gradually through depreciation, unless useful-life assumptions change again or assets are impaired.
How Depreciation Schedules Reshape ROIC
The shield is temporary because ROIC is affected on both sides of the equation. The invested-capital base expands as data centers, servers, and related infrastructure are added. Later, once assets move from construction to active service, depreciation weighs on operating profit and net operating profit after tax.
A June 2026 National Bureau of Economic Research working paper, NBER Working Paper No. 35290, examines technology capex accumulation, investor hurdle rates of return, and productivity lags in an open-economy asset-pricing model. Its relevance here is the timing problem: technological transitions can require large capital accumulation before long-run productivity benefits become visible.
That does not prove a precise earnings path for any hyperscaler. It does, however, frame the valuation risk. If invested capital rises before AI revenue and productivity gains are clearly proven, ROIC can compress even without a collapse in headline revenue growth.
The Hardware Obsolescence Scenario
The sharper downside case is hardware obsolescence. If AI hardware improves quickly enough that current-generation servers lose commercial relevance before their six-year accounting lives expire, companies could face accelerated depreciation or impairment charges.
That would reverse part of the earnings protection created by longer depreciation schedules. The risk is a mismatch between accounting life and commercial life: assets may remain on depreciation schedules even if their economic usefulness fades faster than expected. That is the core issue behind Big Tech AI Capex: The Hidden Depreciation Risk.
What Remains Unclear
The timing of sector-wide earnings dilution remains uncertain. Alphabet’s disclosure is company-specific and should not be generalized to every hyperscaler without comparable filing evidence. Each company has its own deployment cycle, useful-life assumptions, cloud mix, lease exposure, and margin structure.
It is also unclear whether future enterprise AI software revenue will scale quickly enough to offset depreciation pressure. The bullish case requires more than strong demand for AI tools. It requires those revenues to arrive fast enough, and at sufficient margins, to justify the capital base being built today.
Next Watchpoints for Tech Investors
The key disclosure area is the Property and Equipment footnotes in upcoming quarterly filings. Investors should watch for changes to server useful-life assumptions, depreciation expense relative to capital expenditures, and management commentary on utilization, capacity timing, and AI infrastructure returns.
The next confirmation point will be the Q3 2026 earnings season. The central question is whether depreciation remains a delayed accounting cost, or begins to show up as a visible drag on operating margins and ROIC.
***
*Disclaimer: This article is for informational purposes only and does not constitute financial, investment, tax, or legal advice. MarketPilotDaily does not guarantee specific price outcomes or investment returns. Readers should consult a licensed financial professional before making any investment decisions.*
Want the full institutional-style PDF version? Enter your email for the free PDF.