Goldman Sachs' warning about the long-term earnings impact of massive AI infrastructure spending highlights a growing risk that investors may be overlooking amid enthusiasm for artificial intelligence.

Key Highlights

  • Technology companies are committing roughly $770 billion to AI-related capital expenditure.
    • Depreciation expenses from data centres and AI hardware will become a larger drag on earnings.
    • Previous technology infrastructure booms often delivered returns more slowly than expected.
    • Revenue growth alone may not offset the financial impact of rising capital costs.

AI Spending Has Become a Capital Cycle

The artificial intelligence investment cycle has become one of the largest capital deployment programmes in technology history. Major cloud providers and platform companies are spending aggressively on data centres, networking equipment, and advanced processors to secure a leading position in the AI economy.

Investors have largely focused on revenue opportunities. A less discussed financial reality is now becoming more important.

Depreciation Is the Hidden Cost

Every dollar spent on AI infrastructure eventually appears on income statements through depreciation. Data centres, servers, networking systems, and specialised AI hardware have finite useful lives, meaning their costs are gradually recognised as expenses over several years.

For companies investing hundreds of billions of dollars, these charges can become substantial. The issue is particularly relevant for hyperscalers such as Amazon and Microsoft, which are leading the infrastructure buildout.

The Earnings Drag Arrives Early

Their investment programmes are designed to support future AI demand, but the accounting impact arrives long before the full commercial benefits are realised. As infrastructure assets enter service, depreciation expenses rise regardless of how quickly AI revenues develop.

Even strong growth in cloud and AI services may initially be offset by the cost of maintaining one of the largest technology investment cycles ever undertaken.

History Shows the Risk

Past technology cycles offer useful parallels. During the fibre-optic expansion of the late 1990s and the telecommunications infrastructure boom that followed, companies invested heavily based on expectations of future demand growth.

The underlying technologies ultimately transformed the economy, but many investors underestimated the time required for infrastructure spending to generate attractive returns.

A similar pattern appeared during earlier cloud computing and mobile network expansions. Infrastructure providers often shouldered enormous upfront costs while customers captured a significant portion of the economic value through lower prices and improved services.

Strategic Necessity Is Not Immediate Profit

None of this means AI spending is misplaced. The strategic rationale remains compelling because companies risk losing market share if they fail to build sufficient computing capacity.

The challenge is that strategic necessity does not automatically translate into near-term earnings growth. Investors currently reward AI leaders based largely on expectations of accelerating revenue, but earnings are determined by both revenue growth and expenses.

The Next Metric Investors Will Watch

Valuation models often assume infrastructure spending will convert efficiently into future profits. If monetisation takes longer than expected, earnings forecasts may prove too optimistic.

AI may ultimately become one of the most transformative technologies in modern economic history. However, the path from capital expenditure to shareholder returns is rarely immediate.