Six of the largest US technology companies — Alphabet, Microsoft, Amazon, Nvidia, Oracle and Meta Platforms — have accumulated close to $1.5 trillion in purchase commitments tied to AI infrastructure, chips, data-center capacity and energy, according to a Financial Times analysis. Goldman Sachs pegs another roughly $1.5 trillion in lease obligations sitting alongside those contracts.
Commitments Are Not Capex
Purchase and lease commitments do not show up on the balance sheet like debt. They are future cash calls tied to multi-year GPU allocations, wafer supply, power purchase agreements and hyperscale real estate. FT flagged Alphabet's disclosures as a case study: purchase commitments jumped sharply between Q1 and Q2 as the company locked in long-duration compute and energy. That is capacity insurance in a shortage market — and future obligations if AI monetisation falls short.
Heavy-Industry Economics Meet AI
The buildup means AI companies now look more like utilities or heavy industry than software peers: enormous fixed obligations, multi-year contracts and revenue models that must eventually service those commitments. Investors trying to compare Big Tech AI exposure to any single quarter's capex line are missing the load-bearing part of the risk stack.
What It Rhymes With
The disclosure lands into a market that has spent 2026 racing to fund the compute layer with new financial instruments. Compare with our coverage of Nvidia's $500B financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman and KKR, Databricks closing $5B at $190B, and Thrive Holdings' $2B rollup vehicle at $12B. The through-line: AI's real balance sheet is being built off the balance sheet.
Reporting based on coverage from Financial Times and Goldman Sachs research summarised by TechStartups.