AI Data Center Developers Use Bank Letters of Credit to Secure Grid Interconnection Rights
Miles Bennett
U.S. AI data-center developers are posting bank letters of credit to guarantee grid-upgrade costs to utilities, with total arrangements now reaching roughly $10 billion — power access, not chips or capital, is becoming the first bottleneck in AI infrastructure.
How do these letters of credit actually work?
A letter of credit — a bank-issued guarantee that says "if the developer can't pay, we will" — is a traditional tool in commodities and oil-and-gas. Data-center developers are now borrowing it at scale.
The logic: developers avoid tying up cash while waiting in the grid queue; if a project falls through, the utility draws on the bank line directly, so ordinary ratepayers don't foot the bill.
This means → the instrument solves two problems at once: developers conserve cash, and utilities get default protection.
Who is doing this, and at what scale?
Switch closed a $3.5 billion syndicated letter-of-credit facility with 15 lenders at a cost of about 2%, calling it a first-of-its-kind deal. Switch is controlled by digital-infrastructure investor DigitalBridge and is evaluating an IPO.
Blackstone-owned data-center operator QTS is in talks to expand a facility of roughly $2 billion; TeraWulf, a Bitcoin miner pivoting to data centers, is negotiating a similar arrangement.
In plain terms = the top-tier players have made "bank guarantee first, grid queue second" standard practice — and the numbers are still growing.
Why do utilities accept these guarantees?
Projects waiting for U.S. grid access now exceed 800 gigawatts — more than the country's average total electricity consumption — and many are speculative filings.
Utilities such as Dominion Energy in Virginia are culling the queue, and developers with bank guarantees have a clear advantage in the screening process.
Carson Kearl, senior analyst at energy research firm Enverus, put it bluntly: "Without a letter of credit, most utilities won't even evaluate your project."
Is policy pushing this forward or pulling it back?
The Trump administration is pressing data-center operators to cover their own power and infrastructure costs; regulators are directing regional grid operators to fast-track consumer-protection mechanisms.
This means → the policy direction is "whoever uses the power pays for it," forcing developers to put up real money or bank guarantees to prove they are not speculators.
That backdrop is directly driving up demand for letter-of-credit arrangements.
Where is the risk?
The AI sector faces a "heavy spending, thin monetization" squeeze, and bond-investor appetite for related assets has already cooled.
This reflects a deeper bet: Wall Street's accumulated exposure in these credit facilities is ultimately a wager that AI computing will deliver commercial returns on schedule.
Put simply = banks have vouched for the developers, but if AI's revenues never materialize, those guarantees could turn into real losses — that is the central test for whether this financing model can keep expanding.
Content is for reference only, not financial advice.