Raw gigawatt queues overstate financeable AI power demand; collateral, ownership, tariffs and energization convert headlines into investable capacity.
Collateral is turning data-center grid queues from demand headlines into financing tests
Analysis by Frank Locascio and TheBRRR Research
What happened
A Reuters review found more than 700 GW of very-large-load requests across parts of the Midwest, Mid-Atlantic and South, but highlighted duplicate and underfinanced requests. ERCOT separately reported more than 438 GW of requests, 89% from data centers. Exelon said stricter collateral reduced its high-probability data-center demand tally by about 40% to 11 GW.
Why it earned coverage
The post-cutoff investigation supplied a cross-region comparison and an observable queue-shrinkage receipt after stricter collateral.
Investment transmission
Deposits, site control, ownership disclosure and study fees force developers to demonstrate capital and project maturity. Queue attrition changes probability-weighted utility capex, generation need and supplier backlogs even if headline requested load remains enormous.
Affected exposures
Next observable receipt
ERCOT Batch Zero classifications, PJM/utility queue withdrawals, deposit forfeitures, disclosed owners, signed tariffs and energized megawatts.
What would invalidate it
Qualified queues remain near headline levels after collateral and ownership checks, projects reach energization on schedule and utilities disclose firm take-or-pay protection.