Collins is proposing a hypothetical future policy for the next Democratic administration: if AI data centers face debt crises, the federal government should forgive that debt in exchange for transferring the facilities to cities for conversion to community services. The underlying claim—that these centers will become debt traps "in a few years at most"—goes further than current evidence supports. While data centers are increasingly financed with debt (AI-related borrowers have issued around $170–$200 billion in data-center-related debt), and some analysts warn about potential overleverage and repayment risks, there's no consensus that data centers as a whole will become debt traps soon or that their financing structures are fundamentally different from other major infrastructure projects. Recent JPMorgan analysis shows the AI capex cycle remains profitable, though Bond Buyer reporting does flag infrastructure and public finance risks that states are beginning to address through policy adjustments. What's missing: any specific data on which data center projects are currently distressed, or broader analysis of whether these debt levels actually pose systemic risk.