California legislators considered dozens of bills related to artificial intelligence this year. That’s a number that’s climbed quickly over the last couple years as lawmakers grapple with the technology’s increasing presence — and possible negative impacts.
And one growing point of concern involves generative AI’s relationship with state resources as the technology becomes everyday life for Californians. Experts say generative AI is driving up energy and water demands at data centers.
But the question is: By how much?
Assemblymember Rebecca Bauer-Kahan authored a bill requiring data centers to report their energy usage. It failed to pass. But she said the work for the bill began with that question, and a desire for more transparency.
“As I started to ask questions about what kind of data was going into our understanding around the energy needs of the growing data center industry, it seemed like the answer was there wasn't great data,” Bauer-Kahan said.
Assemblymember Diane Papan authored a similar bill looking to track water use at data centers. It required centers to provide estimates of their expected water use when applying for a business license, and an annual water use report thereafter. These centers generally require large amounts of water to cool down servers and other equipment.
“I feel that the more information we have, it'll help us integrate the growth of data centers into that broader task of climate resilient resource management, particularly as it relates to water,” Papan said.
The bill passed through the legislature but was vetoed by Governor Gavin Newsom. In a statement, Newsom said he was “reluctant to impose rigid reporting requirements about operational details on this sector without understanding the full impact on businesses and the consumers of this technology.”
The decision surprised Papan, who originally described the bill as one that was sure to pass given legislators’ desire to optimize California’s water planning. She said she disagrees with his thinking.
“To not have that transparency and how much water they're using so that we can plan, isn't really helping the industry,” Papan said. “We've got to balance innovation with sustainability.”
Finding a balance
Data centers are places that contain the servers needed to provide essential services online, like web searches and video streaming. Large-scale data centers have been around for decades, first coming into existence about thirty years ago.
But with the more recent rise of generative AI — the kind used for chatbots like ChatGPT — the water and energy these centers demand has gone up.
Shaolei Ren, an associate professor of electrical and computer engineering at UC Riverside, said discussions about the environmental impacts associated with increasing AI use aren’t completely new. Artificial intelligence has been integrated in online platforms for years now — like with YouTube, for example, where AI is used to give tailored video recommendations.
But Ren said it wasn’t until after 2020, when the general public became more aware of the integration of generative AI online, that these conversations really took hold. And with the technology’s growth comes a greater demand on resources – in California and the rest of the country.
Citing a Lawrence Berkeley National Laboratory report, Ren said data centers accounted for 3% of the nation’s energy use in 2020. That’s projected to increase by up to 12% by 2028.
Ren said the way in which data centers consume these resources also plays a part in their impact on state resources. A data center may consume the same amount of water in a year as an office building, for example. But if most of that consumption happens during one hot summer month instead of evenly throughout the year, he said that could create a different kind of stress on water resources.
All of these factors, he said, could make for problems later if left unexamined.
“I think this will create a lot of tension because we don't have that many resources available,” he said. “And also because the users [are] starting to use these AI services directly.”
Ren said there are studies from other states that offer an idea of the water and energy use of some data centers. But he said most data centers that are sharing that information are massive, hyperscale centers run by companies like Google or Microsoft.
In California, however, Ren said you're more likely to find what’s called a “co-location” data center, which is a place where multiple independent businesses can lease out servers.
“This type of data center tends to be more conservative, more secretive, compared to Google or Microsoft, which are relatively more transparent,” Ren said.
But some voices in the industry say this kind of legislation could do more harm than good. Dan Diorio, vice president of state policy for the industry advocacy group Data Center Coalition, says the group opposed these bills largely due to concerns that publicizing their water and energy usage could endanger data centers’ proprietary information.
He also says making companies report energy and water use — and potentially adding efficiency standards on top of that — could encourage the growing AI industry to take its business elsewhere.
“It's going to create a whole separate set of efficiency standards for one industry rather than holistically across the board,” Diorio said. “This specifically makes it harder to develop data centers in the state of California.”
Shared interests
Ren sees bills like Bauer-Kahan’s and Papan’s as necessary to understand the resources data centers require, and the pressure of that demand, in California.
He mentioned another bill — one authored by California Senator Steve Padilla that did get signed into law — aiming to assess possible impacts data centers’ energy use has on ratepayers. This bill, he said, is an example of legislation that could shed some light.
“A lot of people speculate or suspect that these new data center loads would be impacting the utility rate,” he said. “But there's no concrete evidence because of the lack of the transparency of the power industry.”
Padilla said his bill aims to address exactly that.
“I think it is important to collect that data,” Padilla said. “It is important to get an assessment of what the impacts are because then we can stop arguing over some of the basic facts and think about how we're going to address them.”
Assemblymember Bauer-Kahan says that transparency will be key to making sure California’s energy and water resources are secure. She said she was surprised her bill faced the opposition it did.
“I really thought the data center industry would be on board with that, because they too want the energy to stay on at the data centers,” she said.
Now, she said she’s reassessing what it will look like to come back to this issue next year.
“All of us have a shared interest in having a grid that will stand up to the demands that are put on it,” she said. “In order for us to ensure that, we have to build a grid that is resilient enough, which involves a lot of planning.”
Papan also shared plans to return to her bill in the new year.
“We're going to bring this bill and this issue back because we've got to be prepared,” Papan said. “California always leads the way. Why wouldn't we do it with the water usage of data centers?”
