Albany Agenda

Hochul’s ambitious AI plans may threaten state’s climate goals

New York state wants to become a leader in AI research and slash its use of fossil fuels. But can it do both?

Gov. Kathy Hochul shows off an nVidia RTX 3080 Ti graphics card, which is commonly used in artificial intelligence data centers. The governor announced a $275 million state investment into building an artificial intelligence computing center at the University of Buffalo.

Gov. Kathy Hochul shows off an nVidia RTX 3080 Ti graphics card, which is commonly used in artificial intelligence data centers. The governor announced a $275 million state investment into building an artificial intelligence computing center at the University of Buffalo. Susan Watts/Office of Governor Kathy Hochul

As part of the state budget, Gov. Kathy Hochul touted the inclusion of the Empire AI Consortium, her plan to bring artificial intelligence research to universities in New York and a supercomputing center to the University at Buffalo. The goal: to make the state a national and global leader in the world of AI research and development. 

While much of the public debate and discussion around artificial intelligence has focused on equitable and ethical applications of the emerging technology, the state would be wise to keep in mind the potential environmental cost of AI. With New York already lagging behind on hitting its ambitious environmental benchmarks – like reducing carbon emissions by 40% by 2030 – and major renewable energy projects stalled or scrapped, the introduction of a technology as energy-intensive as AI stands to further threaten the state’s climate goals.

The $275 million computing center in Western New York, as well as other aspects of the consortium, are still in their early phases. But climate activists are already trying to warn about the energy cost of something like Empire AI. Although different AI applications have different energy demands, a common comparison is between a Google search, which uses just 0.3 watt-hours of electricity, and a ChatGPT search, which uses nearly 3 watt-hours. Beyond the increased energy demand, massive AI data centers also consume large amounts of water used for cooling purposes and result in high levels of e-waste.

“From an energy consumption standpoint, it really depends on what they'll be doing there and how much energy they'll be using, and how much that's going to get in the way of our meeting our statutory mandates under the climate law,” said Mandy DeRoche, deputy managing attorney in the Clean Energy Program at Earthjustice. She said that as the state moves ahead with something that could be as energy intensive as AI, it’s important that whoever is doing the research is creating their own clean energy, rather than utilizing existing or planned clean energy infrastructure that would need to be diverted away from other sectors. “(Climate impact) should absolutely be at the forefront,” DeRoche said of expanding AI and associated policy. She likened it to the early days of cryptocurrency mining in New York – another high-profile, energy-intensive technology. “Because it's not intuitive, when you think about it, how much energy use this computing power needs.”

Environmentalists successfully lobbied for a two-year moratorium on fossil fuel-powered cryptocurrency mining in 2022. The law signed by Gov. Kathy Hochul that year was the first of its kind in the nation as climate advocates sought to hit pause on an industry pitched as having positive economic impacts, but would have significant environmental effects. Since then, lawmakers in New York and across the country have sought to better study the environmental impacts of cryptocurrency mining. Other localities have enacted their own moratoriums as well. The city of North Tonawanda just last month passed a two-year moratorium on bitcoin mining. 

Included in that moratorium was another, less high-profile provision: a two-year moratorium on AI data centers. Unlike cryptocurrency mining, the state and region don’t currently have any AI data center complexes in the pipeline. But the Western New York city is right next door to Buffalo, where the governor is hoping to create a supercomputing center, an endeavor that would affect nearby areas as the necessary data centers are sited.

For a glimpse into the possible effects of a proliferation of data centers in New York, one need only look at states like Virginia, which are already experiencing the energy drain associated with an expansion of AI data centers. Much has been written already about the environmental impacts of AI in northern Virginia, and lawmakers in the area are now attempting to enact mitigation methods after the consequences have become apparent. 

Tech-based economic development projects in New York are already responsible for significant burdens on the state’s grid. According to a July report from the New York Independent System Operator, which operates the state’s grid system, electricity demand is expected to increase from 50% to 90% over roughly the next two decades. That increased demand will be driven by two major sectors: the electrification of transportation and housing, and energy-intensive economic development projects.

Per the organization’s April Gold Book report, most of the large load projects in the coming years – projects that have put in requests due to the added stress they will apply to the grid – are related to major tech-related economic development. The most energy intensive project is the Micron chip manufacturing complex in Central New York, but several other projects are cryptocurrency mining operations. The Independent System Operator has not yet received any large load interconnection requests related to Empire AI or artificial intelligence data centers yet, so it’s still too soon to know for sure exactly what the energy demands of the projects will be. 

Specifics on how the state plans to mitigate the climate impacts of Empire AI are also still vague. The statute creating the consortium includes a provision on “energy efficiency,” requiring the initiative “ensure a reliable and sufficient clean energy supply for the institute, and minimize emissions and negative environmental impacts.” The law doesn’t go into any further specifics, but a spokesperson for the governor said she is committed to sustainable solutions. “Governor Hochul advanced Empire AI to harness the power of artificial intelligence and ensure the technology is being used for the public good, and the administration is working closely with consortium members and stakeholders to ensure New York is a model for responsible, sustainable and ethical use of AI,” the spokesperson said.

Assembly Member Anna Kelles, a legislative leader in the fight over cryptocurrency mining, is carefully watching how Empire AI begins unfolding. “When we talk about AI and computational power and e-waste – yes, do I have concerns? Absolutely,” Kelles said. “Am I interested in making sure that we have policies in place to minimize the potential environmental and negative impacts of launching massive AI initiatives in New York? Absolutely.” Kelles said that she has not yet introduced any specific bills related to AI due to the level of nuance it requires, but it is something she is actively exploring. She also echoed concerns raised by DeRoche about the use of clean energy. A major source of energy in Buffalo is hydroelectric power, which Kelles said contributed to the University at Buffalo being chosen as the location of the supercomputing center. “But we also have to remember that there are opportunity costs, because if we use the energy for AI centers and programs, then that renewable energy will not be available (for) other needs,” she said.

Kelles is also waiting to see what directions the AI research will go in, as she considers how that computational power is used inextricably tied with environmental considerations. She offered an example of whether AI is being used to improve social media interface or if researchers are utilizing it to create new predictive weather modeling to help mitigate climate change effects. “One of them uses a tremendous amount of energy and, you know, is used for play,” Kelles said. “And one of them ultimately helps mitigate the impact of climate change… Those things create inherent mitigation of the energy consumption nature of AI and waste of AI.”