AI Power Needs Spur Builders to Tap Billions in Bank Pledges

Power utilities across the US are already strained from AI data centers. But they’re increasingly facing an even bigger threat: that developers will be forced to walk away from projects, leaving households on the hook for massive infrastructure bills.

To minimize that risk and get utilities on board with projects, data center operators are tapping banks to guarantee the full costs of the AI buildout will be paid for.

They’re doing this through letter-of-credit facilities — a classic industrial backstop long used by oil and gas companies that’s surging in prominence across the data center industry. These ensure that if a developer can’t complete a site and doesn’t pay for upgrades that were needed to plug it into the grid, utilities can draw on the credit lines so that regular households are spared the costs.

TeraWulf Inc., a bitcoin miner-turned-data center developer, is in talks with bankers for such a facility, according to a person familiar with the matter. Blackstone Inc.-owned QTS is in discussions about expanding a potential $2 billion deal the operator was already negotiating with lenders, another person said. Switch Inc. recently got a host of banks to commit $3.5 billion to fulfill its grid obligations.

With the letters of credit in hand, developers can avoid tying up vital cash upfront when utilities demand billions of dollars for grid upgrades to keep them in the queue for power. And Wall Street is now syndicating these guarantees at an unprecedented scale to meet AI’s enormous spending needs.

“Securing abundant, reliable electrons now matters more than chips or capital,” said Arnaud Stevens, head of global trade for the Americas at Natixis SA, which co-led the Switch deal. “Raising large, syndicated letter-of-credit facilities is increasingly a key priority of large developers.”

The size of the commitments underscores the costs of fueling the AI revolution, which have propelled Wall Street firms to devise billions in novel debt deals.

See more: Beyond Moore’s Law: The Full Stack Driving AI in 2026