The North American Electric Reliability Corporation — the entity whose job is to prevent half a continent's lights going out — just issued its highest alert for the PJM Interconnection.
Level 3 is a big step. I wonder if the RTOs will make equally big steps and what that will look like. We know what it will look like in ERCOT, and, to some extent, in PJM. What do you think about the others?
The concentration framing is the most underappreciated part of this piece. Most of the policy debate treats this as a capacity problem — "build more generation, build more transmission" — but your point that the oscillation pattern of AI workloads is the real culprit reframes it entirely. More supply doesn't fix a synchronization problem.
One thing I'd push on: is distributed modular compute realistically incentivized under current market structures? Hyperscalers consolidate in PJM corridors partly because of talent density, fiber infrastructure, and interconnect agreements that took decades to build. The grid physics argument for distribution is compelling — but what's the forcing function that actually gets operators to build at the edge instead of in the same five counties in Virginia? Regulation, insurance pricing, or something else?
Level 3 is a big step. I wonder if the RTOs will make equally big steps and what that will look like. We know what it will look like in ERCOT, and, to some extent, in PJM. What do you think about the others?
The concentration framing is the most underappreciated part of this piece. Most of the policy debate treats this as a capacity problem — "build more generation, build more transmission" — but your point that the oscillation pattern of AI workloads is the real culprit reframes it entirely. More supply doesn't fix a synchronization problem.
One thing I'd push on: is distributed modular compute realistically incentivized under current market structures? Hyperscalers consolidate in PJM corridors partly because of talent density, fiber infrastructure, and interconnect agreements that took decades to build. The grid physics argument for distribution is compelling — but what's the forcing function that actually gets operators to build at the edge instead of in the same five counties in Virginia? Regulation, insurance pricing, or something else?