Corporate earnings in the S&P 500 are poised for strong growth, driven partly by significant investments in artificial intelligence (AI) infrastructure. Analysts project earnings gains exceeding 20 percent for a second consecutive quarter, buoyed by semiconductor manufacturers and other suppliers crucial to AI’s expansion. This dynamic has supported elevated stock market valuations while keeping price-to-earnings ratios within manageable ranges.

Key beneficiaries include companies supplying the hardware essential for hyperscale data centers, such as Meta Platforms, Microsoft, Alphabet, Amazon, and Oracle. These firms are investing heavily in facilities to power AI technologies, accelerating capital expenditures. For example, chipmaker Nvidia quickly records revenue and profits upon sale, but its customers treat these purchases as capital assets. Consequently, these customers defer recognizing related costs over several years through depreciation, leading to a lag between capital spending and its impact on reported earnings.

This mismatch means current earnings reflect strong revenue growth, while the costs of expanding AI infrastructure remain understated in income statements. Furthermore, depreciation expense forecasts for these hyperscalers vary widely, highlighting uncertainty about how these costs will affect future profit margins. Data compiled by Visible Alpha shows that while revenue projections for Meta through 2028 exhibit low dispersion, estimates for depreciation and amortization have substantially higher variance, underscoring investor challenges in modeling expenses accurately.

Several factors contribute to the difficulty in forecasting depreciation. Many hyperscalers only recently transitioned to capital-intensive models, limiting historical comparability. Disclosure on how depreciation is allocated across various expenses is minimal, and companies have latitude in assigning useful lives to assets, altering annual depreciation charges. Additional complexity arises from off-balance-sheet financing used in data center projects. As a result, some analysts group AI-related asset depreciation with broader totals, making precise evaluation challenging.

Data from S&P Global Market Intelligence indicates S&P 500 firms committed approximately $13.3 trillion in capital expenditures for 2025, including $412 billion among the five major hyperscalers. Projections for these companies’ capex in 2026 approach $760 billion, far outpacing expected depreciation expenses of around $21 billion. David Zion, founder of Zion Research Group, notes that analyst estimates for depreciation among these firms vary significantly.

This timing gap impacts free cash flow figures. For 2026, the combined free cash flow of the five hyperscalers is forecast to decline 91 percent to approximately $16 billion, even as net income is expected to rise by 25 percent to $60 billion. Amazon and Oracle anticipate negative free cash flow, while Meta is expected to generate only modestly positive levels. Analysts currently expect these companies’ earnings to grow at a roughly 20 percent annual rate through 2029, accompanied by a rebound in free cash flow reaching $185 billion in 2028 and $387 billion in 2029. This outlook assumes a slowdown in capital spending after 2026, with continued strong revenue growth enabling a V-shaped recovery in cash generation. However, uncertainty remains high.

For investors, the divergence between reported earnings and cash flow, driven by the timing of capitalizing AI investments, presents risks and complexities in valuation. The forward price-to-earnings ratio for the S&P 500, near 22 times projected earnings, exceeds historical norms even before accounting for upcoming increases in depreciation expenses. Much of the current profit surge rests on spending that will not fully impact income statements for years. How effectively the AI giants convert their infrastructure investments into sustainable revenue remains a critical question for market participants.