Open in App
  • Local
  • U.S.
  • Election
  • Politics
  • Crime
  • Sports
  • Lifestyle
  • Education
  • Real Estate
  • Newsletter
  • Goodyear Independent

    AI taxing Arizona power, water and residents

    2024-04-19

    https://img.particlenews.com/image.php?url=07gfsh_0sWXWfCD00

    Arizona is at risk of straining its power grid as AI servicing data centers proliferate around Maricopa County, leaving power companies rushing to meet developer demand, without compromising service and increasing consumer costs.

    According to the International Energy Agency, 2,700 data centers consumed more than 4% of the country’s total electricity in 2022. The agency projects this will reach 6% by 2026.

    Analysts believe the Phoenix area has been at the forefront of data demand, trailing only Ashburn, Virginia, for data center absorption . With the growing demand for AI, data center storage capacity nationally is expected to grow 98% from 10.1 zettabytes (ZB) in 2023 to 21.0 ZB in 2027.

    “Consumers and businesses are expected to generate twice as much data in the next five years as all the data created over the past 10 years,” suggested analysts at JLL, a global commercial real estate and investment management company.

    The growth of AI is a significant driver of the increased demand for power in Arizona.

    In 2023, Arizona experienced the hottest July in Arizona history, setting a new record for energy demand.

    Due to the significant growth in demand, APS said customers will require more than 13,000 megawatts (MW) of energy by 2031, which they are committed to delivering. APS also believes peak demand will grow by 40% from today’s levels in less than seven years.

    “While we are well prepared to manage Arizona’s existing above-average growth, like many other communities and utilities around the country, we are working to meet the extremely large resource needs of the data center sector,” said APS media rep Ann Porter.

    To put it in perspective, APS said it currently has requests for more than 10,000 megawatts (MW) of energy beyond what's already committed. Palo Verde Generating Station, the nation’s largest energy producer, generates 4,000 MW.

    “A large grocery store’s peak energy demand is about one MW," Porter said. “A large data center’s energy demand can range from 250 MW to 1,000 MW."

    A fourth of Palo Verde's energy can be considered significant, but so is the money generated from such developments.

    In a recent 529-acre data center development application with Maricopa County, developers suggested Maricopa County could generate $81 million a year in taxes or approximately $1.6 billion over 20 years.

    The current proposed application, however, has received some pushback from cities.

    “There is opposition from both the City of Buckeye and the City of Goodyear,” a Maricopa County rep stated. “APS may not have sufficient power for the proposed facility, but that has not been confirmed. At this time, Planning and Development are not supportive of this proposal.”

    Salt River Project (SRP) experiences unprecedented demand from the same culprit in the East Valley. SRP’s recent Integrated System Planning (ISP) process determined that the company will need to double or triple its current capacity in the next decade to meet projected growth in electric power demand while simultaneously becoming more sustainable.

    “This plan accounts for the significant, forecasted growth from high-tech manufacturing and industrial customers building data centers in our service territory,” said SRP media rep Jennifer Schuricht. “Data center electricity demand is relatively higher during the night than other types of customers. SRP’s system will grow in a manner that supports new and evolving customer needs throughout the day and night.  “

    SRP supports data center customers like Meta and Google in Mesa to help them offset their energy use with renewable resources.

    In 2023, SRP approved an average monthly customer rate increase of about $11.88.

    In February, APS increased rates after the state approved their request to “allow the company to continue critical investments.”  The rate increase resulted in an average consumer bill increase of roughly 8%, or about $10.50 monthly.

    Microsoft’s secretive Goodyear datacenter

    In 2018, Microsoft paid $48 million for a 259-acre tract in Goodyear near Phoenix Goodyear Airport. Microsoft recently claimed each building constructed on that campus represents more than $300 million in investment.

    According to a report from The Atlantic, Microsoft's Goodyear facility may be designed explicitly for generative AI and ChatGPT.

    If true, the center would be a significant asset in the AI arms race , with other countries and companies rushing to monopolize and corner the AI market by reaching an AI milestone or singularity. Analysts believe a cornered AI market could be worth trillions of dollars and will have significant national security and defense applications.

    According to estimates, the large Goodyear data center’s energy demand could be as high as 1,000 MW, or a fourth of the energy output from the Palo Verde Nuclear Generating Station.

    Microsoft's data centers in Goodyear allegedly use less water than traditional water-cooling data centers. Compared to the more efficient water cooling, air cooling potentially could transfer the cost and burden to energy consumers in various ways.

    Matt Majoli, founder and CEO of Alliance Optix, an IT consultancy specializing in data center operations, says that could be the case, but costs generally get absorbed.

    “The operational demands of large data centers can consume a great deal of power, which often leads to higher loads on local power grids,” he said.”This, in turn, often requires utility providers to upgrade or expand their capabilities. These changes could potentially increase energy costs; however, the increases are generally absorbed over wide customer bases and are frequently offset by efficiency gains and long-term infrastructural improvements.”

    Despite the diligent work APS is doing to meet incoming energy demands amid unprecedented heat, Goodyear and Microsoft worked to amend its agreement with Microsoft to allow more water to come into the complex.

    In 2023, according to city documents and reports, the Goodyear City Council approved an amendment allowing Microsoft to use 1.2 million gallons of water per day at the plant at full build-out. The wording in the agreement affords it to have up to 3 million a day delivered to the site.

    The Atlantic reported in March that once fully built out, the center may use 56 million gallons of water a year, but the city responded that it couldn’t disclose the amount due to proprietary concerns.

    The Goodyear Independent contacted Microsoft for data centers’ location, but the company said the information was proprietary.

    In its 2022 environmental report, Microsoft stated the company consumed 6.4 million cubic meters of water, up 36% from 4.7 million the previous year .

    Shaolei Ren, a researcher at the University of California, Riverside, believes generative AI’s increased demand contributed to that increase. According to his research , AI will consume 4.2 billion to 6.6 billion cubic meters of water a year globally by 2027, or four to six times the annual water withdrawal from Denmark.

    According to Ren's study, much water usage is tied to GPT-3 training.

    “Training GPT-3 in Microsoft’s state-of-the-art U.S. data centers can directly evaporate 700,000 liters of clean freshwater, but such information has been kept a secret,” the study claims.

    The Goodyear data center may be one of the most intensive in terms of computational demand. Ren’s recently published study analyzed the efficiency of water use in Microsoft AI processing and Arizona is a world leader in inefficiency.

    Arizona data centers only processed 15.1 inferences per 500 milliliters of water consumed, almost half the U.S. average at 29.6. Arizona data centers only trailed Washington in inefficiency (10.4). According to IBM Research, an AI Inference is “the process of running live data through a trained AI model to make a prediction or solve a task.”

    SRP and APS

    Despite drawing on massive amounts of resources and being subject to international AI scrutiny, in Microsoft’s case, both power providers maintain the power will be available and reliable come summer.

    A 2023 study suggested that up to 800,000 residents could need emergency medical care for heat stroke and other illnesses during a multi-day power failure.

    APS contends its power grid is redundant and extremely safe.

    “We won’t compromise reliability and affordability for the families and businesses that currently count on us for energy, and we are committed to working with the data center and AI industry, regional utilities and partners, our regulators, local municipalities and customers to innovate and find solutions for society’s growing energy needs while maintaining the reliable, affordable energy that our customers count on,” Porter said.

    SRP also said it has continuity plans to ensure safety.

    “While SRP transitions to more clean energy resources, we're adding proven resources like quick-start, flexible natural gas. This helps to keep Arizona residents and businesses safe from power outages and higher bills while we transition to renewable power sources,” Schuricht said. “SRP has partnered with data center customers including Meta and Google to help them offset their energy use with renewable resources.”

    Expand All
    Comments / 0
    Add a Comment
    YOU MAY ALSO LIKE
    Most Popular newsMost Popular
    Total Apex Sports & Entertainment2 days ago

    Comments / 0