Large loads as the entry point for understanding data center policy and the grid // Texas versus New York on regulating large loads // Work with me on my Foundations of the Grid class this fall!
Yessssssss I will let you know further details when they're ready. Also there is this interest form for my own counting: https://form.jotform.com/252336524090048
This is why I think we’re starting to ask different questions about AI.
The biggest questions are becoming less about what AI can do and more about who builds the infrastructure, who pays for it, who carries the risk and which institutions set the rules before it scales.
The recent grid disturbance in Northern Virginia showed that these are no longer theoretical policy questions. They’re operational ones…
I’ve been thinking a lot about this shift recently and wrote more about it in The End of Work’s Anchor, where I explore how AI, infrastructure, and democratic institutions are becoming increasingly inseparable.
This is why I think the AI conversation is changing.
It’s becoming less about the models and more about infrastructure, responsibility and the institutions that shape how AI scales.
The recent grid disturbance in Northern Virginia showed these aren’t theoretical questions anymore.
I’ve been exploring these questions in my essay The End of Work’s Anchor, about AI, infrastructure and the democratic institutions needed to govern both.
This is the clearest breakdown of the public policy side I've read.
On thing missing. The biggest players aren't trying to reform interconnection — they're bypassing it entirely.
Nvidia/SK's $500B commitment includes 2GW of dedicated generation. That's not a grid customer negotiating queue position. That's a private energy island that never touches the public grid.
Texas and New York are designing frameworks for large loads that want to interconnect. An increasing share don't.
This feels like an auto-response to the piece, and it is also wrong.
The piece is primarily about large load interconnection, but it explicitly discusses behind the meter generation. These two are often related in an ecosystem, rather than being fully firewalled.
Finally: interconnection reform as it is being done around the country, including TX and NY, includes policy about "dedicated generation."
My fault, went back and saw you did cover behind-the-meter generation, and the SB6 section on site control and financial commitment requirements. My apologies for that error.
Where I'm coming from is your frame is that dedicated generation sits inside the interconnection reform conversation which is a tool for grid customers to manage load.
I'm seeing the $500B+ commitments as something slightly different though. Loads that were never going to be grid customers in the first place.
That's less a critique of your piece and more of an observation that the policy frameworks in TX/NY may be optimizing for a category of load that's shrinking as a share of total AI infrastructure.
The interconnection queue is big but so is the growing share of capacity that never enters it.
I would love to join the class - what's the best way to enroll/apply?
Fill out this interest form!
https://form.jotform.com/252336524090048
You got it, thanks!
Would love to join this class!
Yessssssss I will let you know further details when they're ready. Also there is this interest form for my own counting: https://form.jotform.com/252336524090048
This is why I think we’re starting to ask different questions about AI.
The biggest questions are becoming less about what AI can do and more about who builds the infrastructure, who pays for it, who carries the risk and which institutions set the rules before it scales.
The recent grid disturbance in Northern Virginia showed that these are no longer theoretical policy questions. They’re operational ones…
I’ve been thinking a lot about this shift recently and wrote more about it in The End of Work’s Anchor, where I explore how AI, infrastructure, and democratic institutions are becoming increasingly inseparable.
https://saraeson.substack.com/p/the-end-of-works-anchor-ai-datacenters?r=1t83gn&utm_medium=ios
This is why I think the AI conversation is changing.
It’s becoming less about the models and more about infrastructure, responsibility and the institutions that shape how AI scales.
The recent grid disturbance in Northern Virginia showed these aren’t theoretical questions anymore.
I’ve been exploring these questions in my essay The End of Work’s Anchor, about AI, infrastructure and the democratic institutions needed to govern both.
https://saraeson.substack.com/p/the-end-of-works-anchor-ai-datacenters?r=1t83gn&utm_medium=ios
Hochul has been saying things about the Indian Point closure being a terrible mistake. Any nuclear angle in all this?
This was brilliant! Excited for what's to come!
This is the clearest breakdown of the public policy side I've read.
On thing missing. The biggest players aren't trying to reform interconnection — they're bypassing it entirely.
Nvidia/SK's $500B commitment includes 2GW of dedicated generation. That's not a grid customer negotiating queue position. That's a private energy island that never touches the public grid.
Texas and New York are designing frameworks for large loads that want to interconnect. An increasing share don't.
This feels like an auto-response to the piece, and it is also wrong.
The piece is primarily about large load interconnection, but it explicitly discusses behind the meter generation. These two are often related in an ecosystem, rather than being fully firewalled.
Finally: interconnection reform as it is being done around the country, including TX and NY, includes policy about "dedicated generation."
My fault, went back and saw you did cover behind-the-meter generation, and the SB6 section on site control and financial commitment requirements. My apologies for that error.
Where I'm coming from is your frame is that dedicated generation sits inside the interconnection reform conversation which is a tool for grid customers to manage load.
I'm seeing the $500B+ commitments as something slightly different though. Loads that were never going to be grid customers in the first place.
That's less a critique of your piece and more of an observation that the policy frameworks in TX/NY may be optimizing for a category of load that's shrinking as a share of total AI infrastructure.
The interconnection queue is big but so is the growing share of capacity that never enters it.
Either way, genuinely appreciate the depth here.