https://publicenterprise.org/wp-content/uploads/Bubble-or-Nothing.pdf
The rapid buildout of artificial intelligence (AI)-driven data centers—referred to as the “AI boom” or “AO buildout”—has deeply intertwined effects across the tech sector, commercial real estate, banking, and regional economies. This proliferation has created a host of economic and systemic risks that could cascade through sectors should AI growth stall or a market correction occur.
Core Economic Fallout and Sectoral Vulnerabilities
The AI data center boom is driven by hyperscalers—tech giants such as Alphabet (Google), Meta, Microsoft, Amazon, OpenAI, and CoreWeave—which invest massive amounts of capital into infrastructure in anticipation of future demand for compute and AI services. The report reveals urgent weaknesses:
-
Cash Flow Mismatch: AI and related inference services currently generate far less revenue than is required to support the capital expenditures in the sector. Between 2024–2025, hyperscalers reportedly invested over $560 billion in AI/data centers, while only generating $35 billion in revenues, highlighting massive underperformance versus projections.
-
Debt and Circular Financing: The sector increasingly relies on complex, circular debt arrangements, in which major players finance each other’s expansions, creating high interdependency and concentrated risk. Debt, especially off-balance-sheet vehicles, is growing in prominence, including for non-investment grade firms.
-
Rising Operational Costs: The costs for AI model training and inference balloon with each new hardware and software cycle, often outpacing efficiency gains—resulting in persistent operating losses and continual need for fresh capital inflow.
Banking Exposure and Financial System Risks
As data centers blur the lines between real estate and infrastructure financing, U.S. banks and global financial institutions are exposed in multiple ways:
-
Asset-Liability Mismatches: Developers typically use short-term, “mini-perm” construction loans, then refinance into long-term loans secured against lease revenue. However, if tenant churn increases or lease cash flows falter, the underlying collateral becomes unstable, undermining refinancing and releasing stress into Commercial Mortgage-Backed Securities (CMBS) and bank balance sheets.
-
Concentration and Interconnectivity: The top 20 developers and operators control half the market, amplifying contagion risks. Moreover, regional and commercial banks providing liquidity become exposed to sector swings, potentially encountering sizable losses if vacancies rise or a cascade of lease defaults begins.
-
Ponzi Finance and Minsky Moments: Drawing on Minsky’s financial instability hypothesis, if cash inflows are insufficient to service sector-wide debt, some firms may resort to “Ponzi finance”—borrowing or liquidating assets just to meet payments. Once refinancing windows tighten or market confidence wavers, a fire sale of assets and a collapse in equity valuations can precipitate broader financial distress and credit contraction.
Cross-Industry Impacts and Regional Spillovers
-
Municipal Finance and Local Budgets: Regions that bet heavily on data center development—often lured by tax breaks and infrastructure spending—may find themselves saddled with stranded assets and diminished fiscal returns if demand weakens. Many states already lose significant tax revenue to exemptions, with little leverage to recoup losses if the boom turns to bust.
-
Energy Infrastructure: A collapse in data center demand jeopardizes power infrastructure investments, especially if new energy projects were commissioned specifically for tech tenants. Policymakers risk inheriting distressed energy assets unless proactive strategies for repurposing or public acquisition are set in advance.
Policy Warnings and Recommendations
-
Policymakers are urged not to overly commit public resources or fiscal future to the AI/data center boom. They should prioritize resilient investment strategies—such as acquiring or repurposing distressed infrastructure—and scrutinize the true, long-term viability of tax incentives linked to AI or data center growth.
-
Banking stakeholders should re-examine exposure to data center and tech sector debt instruments, stress-testing portfolios for lease churn, off-balance-sheet risks, and indirect exposures through asset-backed securities and regional lending channels.
Ultimately, the report underscores that a significant correction in the AI/data center sector could set off an economic domino effect—stranding specialized infrastructure, challenging bank solvency, distorting local and state budgets, and reverberating through the broader economy via diminished credit and financial confidence.