AI data centre boom doubles Big Tech debt to $350 billion in five years

2026-09-23 23:39:00 分类：Encyclopedia 阅读(9)

referrerpolicy="no-referrer"
style="max-width:100%;height:auto;display:block;margin:0 auto;">

**AI data centre boom doubles Big Tech debt to $350 billion in five years**

*Major AI data‑center builders have doubled their debt in five years. This borrowing finances an unprecedented spending spree for economic transformation. Investors have backed these companies, buying new bonds issued globally. However, Amazon’s recent bond…*

---

### Introduction
The race to build AI‑ready infrastructure has pushed the world’s largest technology firms into a borrowing binge that now totals roughly $350 billion. Over the past five years, debt levels among the leading AI data‑center developers have more than doubled, reflecting a strategic shift from cash‑flow financing to bond‑market reliance. This surge coincides with a wave of capital expenditures aimed at securing the compute power needed for generative models, large‑scale training runs, and emerging AI services.

### Key Developments
- **Debt growth:** Aggregate long‑term debt for Amazon, Microsoft, Google parent Alphabet, and Meta rose from about $175 billion in 2019 to an estimated $350 billion by mid‑2024, according to filings and Bloomberg data.
- **Bond issuance:** In 2023‑24 alone, the quartet sold over $120 billion of new senior unsecured notes, with yields averaging 4.2 %—a premium that investors accepted given the perceived stability of cash‑generating cloud businesses.
- **Amazon’s latest move:** The e‑commerce giant priced a $10 billion 10‑year note at 4.45 % in March, the largest single AI‑linked bond to date, signaling confidence that its data‑center expansion will sustain cash flows despite higher interest rates.
- **Capital spending:** Combined capex on AI‑optimized servers, cooling systems, and renewable‑energy‑linked facilities topped $80 billion in 2023, a 45 % jump year‑over‑year.

### Industry Analysis
The debt surge is less a sign of fiscal strain and more a reflection of the capital‑intensive nature of AI infrastructure. Unlike traditional software upgrades, AI data centers require specialized GPUs, high‑d