In my last job, whenever we had a month where we missed our targets, I often had to write memos that would explain why we missed, how much we missed by, and what our plan to get back to target included. When I was writing one of these memos, I generally had a vision in my mind of how I wanted this to look: a table here, a cleaner chart there, etc. When I got close to what I wanted it to look like, it was suddenly 45 minutes later.
We’ve all had that moment, where something small tends to take more time than you anticipate. Today, formatting can be a five minute task; I can use AI tools to help reformat a paragraph, organize my thoughts into a table, and more.
In the background, however, there are thousands of servers chugging along to help format documents, providing ideas for how to soothe a crying baby, and answering if you should put glue on pizza.
All of those questions that we put through AI seem somewhat weightless. But collectively, the massive physical footprint of a data center which includes thousands of computer servers housed inside large facilities is the physical embodiment of what powers all of these actions.
In the news, we often hear about the race to build the smartest AI model. Underpinning that race is the question: who can build the infrastructure to power that model?
This has led to companies spending a lot of money to build data centers. This leaves states and residents grappling with the question: Do we have enough power to continue supporting this technology?
New York is the first state to pause to seek an answer to that question.
Recently, Governor Hochul signed an executive order that placed a one year pause on new hyperscale data center approvals. This is one of the most significant attempts at regulating the infrastructure demands of artificial intelligence.
It is really important to clarify that this does not mean that New York banned all of its data centers or is turning its back on AI. Instead, this is a one year moratorium on permits for new hyperscale facilities. This means that existing data centers can continue to operate. It also means that smaller data centers are not affected, only very large facilities are paused. Specifically, it places a pause on new state environmental permits for hyperscale data centers, which are facilities that require at least 50 megawatts of electricity. To put that into context 50 megawatts of electricity can power about 50,000 homes.
It is an important distinction to make that this policy is not banning data centers and is not against AI. Rather it is a pause on the largest new projects so that New York can figure out how to manage the infrastructure and who should be paying for it.
Why Did New York Step In?
As I previously discussed in New York’s Data Center Dilemma, large language models require more computing power than other internet services. Companies are racing to build more facilities to support their AI ambitions. These data centers are extremely resource intensive and communities across the country have pushed back due to:
Increased energy strain due to the high amount of electricity required
Increased water consumption used to cool off the facilities
Rising electricity costs due to improper safe guards.
The rising electricity costs is probably the largest concern New Yorkers have because massive data centers require expensive upgrades. Without the right guardrails, these costs can be passed down to the residents.
Why Now?
New York already has several data centers and several are relatively small, ranging from 5-20 MWs. Now, there are several large data centers planned that would require 100+ MW. Several of the data centers will be located in upstate New York, where there is more land, existing transmission infrastructure, and cooler temperatures. Several upstate communities are worried about the amount of electricity and water these facilities would draw on after they are built. Additionally, the New York Independent System Operator has already warned that NY faces reliability strains due to rising electricity demand, delays in renewable energy projects and more.
Weeks earlier, the state legislature passed a separate bill, the Responsible Data Center Development Act, setting a lower 20 MW threshold. Thus, Hochul’s executive order isn’t a unilateral move by the Governor, it is a timely compromise. This executive order is important now because the next wave of data center projects are an entirely different scale of electricity, water and more.
So Why a Moratorium?
Hochul’s executive order places a temporary pause on state environmental permits for new hyperscale data centers. New York agencies will use this time to help develop rules around:
Energy Demand: The states wants to understand how these data centers will impact the electricity supply and specifically the grid’s reliability
Environmental Impacts: Regulators want to understand how to align future projects with New York’s climate goals
Community Benefits: The state is also exploring how the benefits of these data centers can be distributed to communities, not just the costs.
Essentially, Hochul recognizes the need for an effective playbook outlining how to govern AI’s infrastructure.
This executive order shows that the narrative behind the infrastructure is just as important as the one around the software. There have been several conversations on algorithms and chips but now the limiting factor is not just computing power, it is also whether states can generate and deliver enough electricity to support it.
Conclusion
AI runs on these massive data centers. I do appreciate this order because it does force us to think a bit more holistically about what powers all of our actions and take some time to create guiding principles around new data center projects. It is ensuring that residents aren’t left paying the bills or left in the dark.