Let’s start with the numbers: capacity auction prices for the largest regional power grid in the U.S. nearly jumped ninefold in a year. The story didn’t stop at electricity prices—it followed the cables straight into AI’s books. $XAU
Capacity prices for PJM Interconnection’s 2025/2026 delivery year were nearly nine times higher than the year before. By the 2026/2027 delivery year, they had hit the cap approved by the Federal Energy Regulatory Commission: $329.17 per megawatt-day. There isn’t enough power, and new connections are hard to get—creating a more formidable bottleneck than chips. $CL
The demand side is just as staggering. Reuters calculated on August 4 that Microsoft, Meta, Oracle, Amazon, and Google’s parent company had a combined roughly $1.09 trillion in outstanding lease payment commitments, most of them tied to AI data centers. By May this year, the four major cloud providers had disclosed nuclear power purchase commitments totaling about 9.8 gigawatts, using self-built power sources to bypass the public grid’s connection queue. $BNB
The chip supply situation hasn’t eased either. Delivery times for some high-end chips have stretched to 36 to 52 weeks. In May, the North American Electric Reliability Corporation issued its highest-level reliability alert to date, noting that loads of this kind had experienced unexpected disconnections exceeding 1,000 megawatts in a single event—disrupting grid frequency faster than operators could respond.
There’s another, more tangled layer: money. The International Monetary Fund has calculated that AI could push up U.S. electricity prices if transmission and renewable energy fail to keep pace. The Bank of England, meanwhile, has warned that interconnected corporate financing structures—where companies issue debt to fund data centers—could amplify credit risk when markets turn. The longer and more tangled the financing chain becomes, the more a regulator’s intervention at any one link could make the whole structure wobble.
So “tightening” doesn’t necessarily mean an administrative shutdown. More often, it means power, money, and chips all getting squeezed at once. Whichever bottleneck eases first will determine whether computing capacity can actually be brought online.
These are personal views, not investment advice. You are responsible for any gains or losses.
#U.S. plans tighter data center regulation, potentially adding new constraints to AI expansion
Capacity prices for PJM Interconnection’s 2025/2026 delivery year were nearly nine times higher than the year before. By the 2026/2027 delivery year, they had hit the cap approved by the Federal Energy Regulatory Commission: $329.17 per megawatt-day. There isn’t enough power, and new connections are hard to get—creating a more formidable bottleneck than chips. $CL
The demand side is just as staggering. Reuters calculated on August 4 that Microsoft, Meta, Oracle, Amazon, and Google’s parent company had a combined roughly $1.09 trillion in outstanding lease payment commitments, most of them tied to AI data centers. By May this year, the four major cloud providers had disclosed nuclear power purchase commitments totaling about 9.8 gigawatts, using self-built power sources to bypass the public grid’s connection queue. $BNB
The chip supply situation hasn’t eased either. Delivery times for some high-end chips have stretched to 36 to 52 weeks. In May, the North American Electric Reliability Corporation issued its highest-level reliability alert to date, noting that loads of this kind had experienced unexpected disconnections exceeding 1,000 megawatts in a single event—disrupting grid frequency faster than operators could respond.
There’s another, more tangled layer: money. The International Monetary Fund has calculated that AI could push up U.S. electricity prices if transmission and renewable energy fail to keep pace. The Bank of England, meanwhile, has warned that interconnected corporate financing structures—where companies issue debt to fund data centers—could amplify credit risk when markets turn. The longer and more tangled the financing chain becomes, the more a regulator’s intervention at any one link could make the whole structure wobble.
So “tightening” doesn’t necessarily mean an administrative shutdown. More often, it means power, money, and chips all getting squeezed at once. Whichever bottleneck eases first will determine whether computing capacity can actually be brought online.
These are personal views, not investment advice. You are responsible for any gains or losses.
#U.S. plans tighter data center regulation, potentially adding new constraints to AI expansion
