Tech giants take on record debt to finance the AI race (the-decoder.com)

Major tech players — Google, Meta, Microsoft, Amazon and others — are borrowing at record levels to fund a rapid AI buildout: roughly $112 billion deployed in the past three months alone. To underwrite sprawling new data centers and massive compute purchases, firms are using bonds secured by data centers, special-purpose vehicles that keep debt off corporate balance sheets, and large syndicated loans. Examples include Blackstone’s planned $3.46 billion data-center bond, Meta’s $30 billion SPV, banks underwriting an $18 billion loan for OpenAI’s Stargate site, and industry plans for an additional $38 billion of sites; Morgan Stanley estimates private lenders must supply about $800 billion over the next two years. The race is driven by prodigious compute commitments — OpenAI’s roughly $1 trillion in compute deals have already secured over 20 gigawatts of capacity — and public pledges of up to $1.4 trillion in investments. The scale and structure of this financing matter because they concentrate credit risk in both tech and financial markets. With only ~3% of consumers currently willing to pay for AI services, regulators and central banks warn that if hyperscalers can’t convert capacity into sustainable profits, losses could spill into broader credit markets. Circular capital flows — hyperscalers funding AI startups that in turn buy hyperscaler cloud and hardware — amplify valuations and create feedback loops that could worsen a downturn. The situation raises questions about lender exposure, off‑balance‑sheet leverage, and whether government support or tighter oversight will be needed to manage systemic risk as infrastructure spending accelerates.