AI's $5T Data-Center Boom Will Dip into Every Debt Market: JPMorgan (www.bloomberg.com)

JPMorgan warns that the AI-driven construction and retrofitting of data centers will amount to roughly $5 trillion in investment over the coming years, and that financing this buildout will reach into virtually every corner of the debt markets. Because hyperscale facilities need specialized construction, massive power and cooling upgrades, and long-term leases to host expensive GPU and accelerator fleets, banks, corporate bond issuers, CMBS (commercial mortgage-backed securities), leveraged-loan markets, muni and infrastructure debt, and securitizations are all likely to become sources of capital — shifting credit supply, underwriting standards and investor appetite across sectors. For the AI/ML community this matters practically and strategically: more capital flowing into data-center capacity accelerates access to high-density compute, reduces training latency, and supports the larger models and bursty workloads researchers and companies demand. Technically, lenders will evaluate new risk profiles tied to energy consumption, stranded-asset risk for older facilities, supply-chain concentration around GPUs and networking, and contract structures that monetize predictable hosting revenue (e.g., long-term cloud or colocation commitments). The financing mix will shape where compute is built, who controls it, and how resilient and sustainable that infrastructure will be — with implications for model development pace, costs, and geographic concentration of AI compute.