Wall Street Moves to Curb Exposure to $7 Trillion AI Data Center Boom

The race to supply computing power for generative AI is creating one of the largest infrastructure investment cycles on record, with global data center spending projected to approach $7 trillion by 2030. Much of the buildout depends on debt and carries risks beyond conventional real estate, including scarce grid capacity, construction delays, concentrated tenants and rapid chip obsolescence. If AI demand or rental income falls short, losses could spread through banks, private-credit funds, insurers and asset managers.

By May 2026, underwriters including JPMorgan Chase, Morgan Stanley and SMBC were seeking to sell loan exposures or use significant risk transfer transactions to stay within borrower and sector limits. JPMorgan and MUFG were among lenders that had spent more than six months distributing a $38 billion construction financing package for Oracle-leased data centers in Texas and Wisconsin. Some banks considered discounted sales to non-bank lenders, underscoring how the unprecedented deal sizes are straining balance sheets even as Wall Street tries to preserve capacity for further AI lending.