BIS and economists warn of AI infrastructure investment bubble as tech giants project $660–690B capex by 2026 against only $51B in direct AI revenue.
Macro & Markets ·
Five major technology companies—Amazon, Alphabet, Meta, Microsoft, and Oracle—are projected to spend between $660–690 billion on AI infrastructure by 2026, yet direct revenue from AI business is estimated at only $51 billion, according to reporting on the matter. This ratio—roughly ten dollars in spending for each dollar of revenue—has prompted the Bank for International Settlements and several economists to draw parallels with previous financial bubbles, including the dot-com crash.
The structural risk extends beyond the revenue-to-capex mismatch. Major technology corporations, chip makers, and AI labs operate through complex networks of private agreements, including circular financing arrangements in which large companies invest in AI developers that subsequently commit to purchasing computing capacity from those same investors. AI-related entities already account for about half of investment-grade bond placements and receive 87 percent of venture capital funding, concentrating systemic exposure within the sector.
What remains uncertain is whether current AI deployments will generate sufficient returns to justify the investments being made. The BIS has flagged the timing mismatch as the core concern: capital is being deployed now while profits necessary to recover those investments remain unguaranteed. The scale—with Goldman Sachs estimating annual AI infrastructure capex could reach $765 billion by 2026 and $1.6 trillion by 2031—means the outcome carries implications beyond individual company performance.