Goldman Sachs Chief Economist Jan Hatzius warned that the aggressive spending on artificial intelligence infrastructure will eventually slow down, signaling a potential deceleration in demand for energy and raw materials tied to the sector. Speaking at the firm's Communacopia & Tech conference, Hatzius emphasized that while productivity gains remain the baseline assumption, the transition from an initial 'build-out' phase to an 'exploitation' phase will likely reduce capital outlays, directly impacting the commodity supply chains that have rallied on AI expectations.

Market Context

The warning comes amid projections that global investment in AI infrastructure will reach a record $31.6 trillion through 2050, according to PwC's Global Data Centre Outlook. Annual data center capital expenditures are forecast to surge from approximately $800 billion in 2026 to $1.8 trillion by 2050. This massive capital allocation involves major tech players like Meta (META), Google (GOOG), and Microsoft (MSFT), whose spending has been a primary driver for energy and industrial commodity demand. PwC identifies energy and real estate as critical components of this supply chain, making the trajectory of these capex cycles vital for commodities traders.

Analysis

Hatzius outlined two primary scenarios for the future of AI spending. The first is a downside risk where a significant portion of these investments proves unproductive. However, even in his baseline 'optimistic view'—where AI spending is sustainable and contributes to stronger productivity growth—Hatzius argues that a slowdown is inevitable. He described a natural cycle where an initial build-out phase sees investment volumes rise substantially, followed by an exploitation phase where the technology is utilized and investment levels decrease. For commodities markets, this transition suggests that the intense demand for energy and materials driven by new data center construction may peak and decline, challenging the narrative of perpetual growth that has supported recent rallies in the sector.

Key Numbers

- Projected global AI infrastructure investment through 2050: $31.6 trillion

- Annual data center capital expenditures in 2026: ~$800 billion

- Annual data center capital expenditures projected for 2050: $1.8 trillion

- PwC notes AI infrastructure investment is expected to accelerate as chips require upgrades every few years

What to Watch

Traders should monitor quarterly capital expenditure reports from major hyperscalers for signs of the transition from the 'build-out' to the 'exploitation' phase. Additionally, watch for shifts in energy and real estate sectors, which PwC identifies as critical components of the AI infrastructure supply chain. The speed of chip upgrade cycles and regulatory developments in financing will also influence the trajectory of this capital-intensive trend, with implications for long-term commodity demand forecasts.