The AI Debt Bubble: A Looming Risk to Financial Stability

The artificial intelligence sector is experiencing a debt-fueled frenzy that mirrors the speculative excesses of past financial bubbles. From 2023 to 2025, major tech companies have aggressively financed AI infrastructure through corporate bonds, private credit, and opaque financing structures. Goldman Sachs' AI equity basket companies alone issued $141 billion in corporate debt in 2025, surpassing the $127 billion raised in all of 2024. Oracle, a poster child for this trend, plans to increase its debt load from $100 billion in 2025 to $290 billion by 2028, with a debt-to-equity ratio of $4.50 for every dollar of equity-marking the most aggressive capital structure in the sector. This surge in leverage is not confined to giants: startups like CoreWeave and Nebius are borrowing billions to fund data centers, often with circular financing chains that obscure true demand according to research.

A Dot-Com Echo with Modern Nuances

The parallels to the dot-com bubble are striking. In the late 1990s, speculative investment drove valuations far beyond fundamentals, leading to a collapse that erased trillions in market value. Today's AI sector shares similar traits: rapid capital inflows, soaring valuations for unprofitable firms, and infrastructure projects (like data centers) built on speculative returns according to analysts. However, there are key differences. Unlike the dot-com era, where infrastructure was often disconnected from real use cases, today's AI demand is contractually guaranteed by long-term enterprise commitments according to Forbes. Microsoft, for instance, has secured multi-year deals with AI startups, ensuring revenue streams for infrastructure investments according to reports.

Yet the risks remain. Morgan Stanley analysts estimate AI-related data center debt could exceed $1 trillion by 2028, with companies like CoreWeave and Crusoe borrowing tens of billions for projects that may not generate the projected returns. The circular nature of AI financing-where Nvidia invests in OpenAI, which buys Nvidia chips, which are used in Oracle data centers-creates a fragile ecosystem where demand is artificially inflated according to industry analysis. As Andrew Odlyzko, a technology historian, warns, "Bubbles take a long time to build. But they burst very quickly" according to reports.

Energy Grid Strain and Supply Chain Bottlenecks

The AI boom is straining energy grids and supply chains, compounding systemic risks. U.S. data centers already consume 4% of total electricity, and AI-driven expansion could push this to 133% by 2030. In Virginia, data centers already account for 26% of electricity use, with projections of 35-50% by 2030. The Department of Energy warns that without significant grid upgrades, blackouts could increase 100-fold by 2030 due to retiring power sources and insufficient capacity. Microsoft's CEO, Satya Nadella, has called power the "primary bottleneck" for AI deployment, as even the most advanced chips are unusable without sufficient cooling and electricity according to industry experts.

Supply chain dependencies further amplify fragility. AI infrastructure and energy systems share overlapping needs for critical minerals like copper and gallium, with a potential 30% supply gap by 2035 according to industry reports. China's dominance in mineral refinement and the 15-year lead times required to expand mining capacity create energy security risks according to experts. Additionally, AI and energy sectors compete for skilled labor, exacerbating workforce shortages and delaying infrastructure projects according to industry analysis.

Systemic Risk Across Sectors

The interconnectedness of AI, finance, and energy creates a web of systemic risk. Private credit markets, already a shadow banking system, are financing AI infrastructure with opaque leverage structures reminiscent of the 2008 crisis. For example, $800 billion of AI infrastructure spending is expected to come from private credit, with firms like TeraWulf and CoreWeave issuing high-yield bonds according to research. If AI valuations decline, the ripple effects could destabilize financial institutions holding these debts according to analysts.

Energy firms, meanwhile, face counterparty risk as their revenue depends on AI companies' ability to service debt. A collapse in AI demand could lead to defaults, triggering a cascade of failures in energy providers and utilities according to industry experts. Central banks are also caught off guard: U.S. monetary policy shocks have heterogeneous effects on clean energy sectors, highlighting the need for regulators to integrate AI-driven energy dynamics into policy frameworks.

Conclusion: A Perfect Storm?

The AI debt bubble is not a simple repetition of the dot-com crash but a more complex, multi-sectoral risk. While today's AI leaders like Oracle and Microsoft show resilience and real revenue growth, their aggressive leverage and opaque financing structures create vulnerabilities according to Forbes. Energy grid strain, supply chain bottlenecks, and financial interdependencies mean a crisis could unfold rapidly and broadly. Investors and policymakers must act now to address these risks-before the next bubble bursts.