A data center for AI could nearly triple San Jose's energy use. Who will pay the bill?

Applications of AI


The county seat of Santa Clara is touting its partnership with Pacific Gas & Electric, claiming the city is “a premier destination for data center development on the West Coast.” The investor-owned utility currently estimates it has enough capacity in its planned pipeline to push the city's power usage nearly three times its current peak.

PG&E and city officials say the plans call for major power grid upgrades, but also raise questions about who will pay for them and whether the state will be able to keep the power clean.

Panelists at the CalMatters event in downtown San Jose clashed over key issues. They included local officials working with PG&E to build data centers in their cities, technology advocates urging California to seize the economic opportunity, energy experts at Stanford University pushing to modernize the grid, and utility watchdogs skeptical of AI's promised benefits.

Their discussions centered on how quickly California should move to meet new demands, what information the public should have a right to, and how to ensure customers don't bear the cost of infrastructure that may never be fully used.

Proposals to more tightly regulate data center development failed in Congress this year. Further discussions are expected in several state agencies and commissions, including the California Energy Commission, the Little Hoover Commission, and the California Public Utilities Commission.

How much energy will a new data center in California actually require?

The proliferation of AI is complicating regulators and utility companies' efforts to predict how fast data centers will grow and how much power they will require. Companies can propose large-scale facilities without committing to build them. The computing demands behind AI are changing rapidly, and cooling needs vary by state. These factors make it difficult to identify long-term energy needs.

The utility reports in its planning document that the data center requests 18.7 gigawatts of service capacity, according to the state's electricity demand forecast. That's enough to power about 18 million homes, compared to an estimated 14 to 15 million in California. Regulators do not expect all of these projects to be built, with projections that those that are built will come online gradually and at less than their requested capacity, amounting to 4 to 6 gigawatts by 2040.

Liang Ming, director of Stanford University's Bits & Watts Initiative and a CalMatters panel speaker, said predictions are especially difficult as companies deploy new AI apps, or in his words, “application layers,” at breakneck speed. This includes products such as ChatGPT that use large language models. No one knows which apps will become popular. That uncertain bet is creating huge demand on the power grid.

“We are really struggling right now,” Min said. “The risk is much higher at the application layer.”

The Office of the Public Advocate, an independent consumer watchdog within the California Public Utilities Commission, recently warned that the rapid growth of data centers could cost Californians billions of dollars in power grid upgrades if projects don't materialize or use far less electricity than promised.

“Ratepayers could end up paying for expensive infrastructure upgrades that may not be needed for years or at all,” the bureau said in a commentary.

Ming said data center load forecasting is a national challenge, but California will need better tools to control rates, meet clean energy goals and remain competitive with states competing to attract data centers and high-wage technology jobs.

Local authorities are also beginning to grapple with the uncertainty. San Jose energy officials say they are reluctant to buy additional power until they know which projects will actually be built. “We don't want to buy more power than we need,” said panelist Lori Mitchell, director of San Jose Clean Energy, the city's public utility. “That's job number one.”

What are the environmental concerns about the data center boom?

California's data center boom is raising a wave of environmental concerns that state officials are only beginning to understand. Those concerns center on water use, carbon emissions from increased energy demand, and air pollution from diesel-backed generators.

Air quality is of particular concern. Backup generators operate only intermittently, but their presence is concentrated in a few areas. In Santa Clara County, the local impact could be even greater simply because many facilities are located in close proximity to each other in a dense industrial area, with so much equipment crammed into a small space.

But states still have limited visibility into what their data centers are doing. Efforts to push for greater transparency stalled this year amid opposition from the tech industry. The only measure that became law gives regulators the power to determine whether data centers are driving up costs, but stops short of requiring environmental reporting.

Ahmad Thomas, CEO of Silicon Valley Leadership Group, and another panelist said his group opposes the electricity disclosure and water reporting measures because they make California less competitive.

“Without an approach to data centers that is at least somewhat competitive with other states, it is very difficult to see a world in which California is at the top of the AI ​​mountain,” he added.

Consumer advocacy groups say a lack of information leaves communities unprotected. “We certainly think there needs to be more transparency, and that's a good thing,” said panelist Mark Toney, executive director of the Utility Reform Network, a ratepayer advocacy group.

Will data centers slow California's switch to clean energy?

The rapid growth of data centers could slow the state's transition to clean energy if it remains tied to natural gas. And some of the carbon-free alternative energy sources that can meet their electricity needs are hotly debated among environmentalists.

Although the state has pledged to achieve 100% carbon-free electricity by 2045, it still relies heavily on natural gas plants for hot summer days. A recent report by environmental think tank Next 10 and UC Riverside estimates that data center carbon emissions nearly doubled from 2019 to 2023. This is primarily due to gas-fired power generation, highlighting how even relatively clean grids may struggle to absorb the load from AI without increasing emissions.

As demand for AI grows, national leaders are making policy shifts. California approved participation in the broader Western power market this year, a move driven in part by new demands on the power grid, including data centers. Critics have warned that the changes could expose the state to dirtier electricity from other states and weaken the state's control over clean energy rules.

Stanford University's Ming argues that California needs to resort to options that some environmentalists would rather avoid. This includes preserving existing resources like the Diablo Canyon nuclear power plant. In a recent report, Min argued that the state will also need more “clean, reliable” electricity, resources that can run 24 hours a day, such as geothermal energy and natural gas plants with carbon capture capabilities.

PG&E agrees. Spokeswoman Stephanie Magallon told CalMatters in an email that nuclear power, carbon capture systems, large-scale projects combining solar power and batteries are all being considered as options to power data centers in the region. But environmental justice critics in California oppose the technology, saying it is an unproven technology that risks expanding the use of fossil fuels.

Mitchell said the community's chosen aggregator will be able to manage the new data center load while keeping power clean and affordable. San Jose's electricity mix is ​​already 60% renewable, and the biggest opportunity is flexibility, she said, by shifting data center use from the hottest afternoons so the city can avoid purchasing additional power.

Will my electricity bill go up at a data center?

California's data center boom is reshaping the battle over electricity rates, exposing disagreements over whether these new customers will lower costs or raise costs for other customers.

PG&E argues that adding large users, such as data centers, can lower rates because the grid's fixed costs are spread over more customers. It also claims that the electricity grid is on average underutilized, operating at about 45% of capacity, but that some parts of the system are facing real loads during the hottest hours and routinely operating close to their limits. PG&E argues that connecting data centers where there is available capacity could spread costs without worsening congestion.

Another panelist, Toney, warned that California is moving forward with large infrastructure projects without knowing which data centers are real or how their costs will be reflected in customers' bills, and urged the state to slow down.

“I'm concerned that we're doing so-called faith-based policymaking,” he said. “The benefits are very speculative, but the costs are very real.”

Toney said some states are starting to tighten regulations around data center growth. An Oregon law requires data center grid costs to be kept out of household budgets. Minnesota's law gives very large data centers their own billing category, allowing regulators to separate their costs from other customers' electricity bills.

“This data center issue and the relationship between affordability and clean energy is a national concern, and California is really behind the curve on this,” Toney said. “There's a myth that California is always the leader.”

Alejandro Laso writes for CalMatters.



Source link