This message is from outside the City email system. Do not open links or attachments from untrusted sources. implicates the mandates and responsibilities of Santa Clara County Public Health, BAAQMD, and Valley Water. The impacts across multiple jurisdictions and the process must reflect a more collective and cohesive conversation.
Our tax dollars are not meant to subsidize speculative market bets. They are intended to secure our safety, well-being, and the essential public infrastructure required for our survival.
Attached, you will find my full testimony on the water, air, and electricity affordability impacts of hyperscale data center development as submitted to Little Hoover Commission Public Hearing on Data Centers and California Electricity Policy.
Masheika Allgood
Founder
AllAI Consulting, LLC | allai-us.com
Environmental Platform: Taps Run Dry
11/18/25, 8:38 AM Mail - Agendadesk - Outlook
https://outlook.office365.com/mail/[email protected]/inbox/id/AAQkADUxOWI4ZjE3LTRkNDEtNGUzMS04MjAwLTIzNzdiYTdkMjc5NAAQA… 2/2
November 17, 2025
Little Hoover Commission
Virtual Hearing on Data Centers and California Electricity Policy (Nov. 20)
WRITTEN TESTIMONY
Hyperscale data centers are often defined by their size - greater than 10,000 square feet. But in electricity terms, a hyperscale data center can be defined as any facility that is considered a large load development (>20MW). The current demand for these large load developments is having a significant impact on all facets of the electricity domain. From upending the capacity forecasting process and overwhelming equipment and component supply chains, to pushing out most vulnerable residential customers into making draconian financial choices.
PG&E’s data center pipeline grew from 5.5GW in February, to 8.7GW in May,1 and currently stands at 9.6MW.2 So any load forecasts or modeling3 that hasn’t considered the entirety of the current data center pipeline for all 80+ California Load Serving Entities4 is likely severely underestimating the demand. Given that California currently generates around 46GW of clean energy in-state,5 it will be difficult to build out the requested capacity quickly while still meeting our zero-carbon electricity goals. Expanding electricity generation for any source other than wind, geothermal, and hydropower will have a significant impact on our state’s water conservation goals. At the local level, the sheer size of the transmission equipment and scale of the build-out will call for easement expansion and some amount of takings to implement. Under the current regulatory system, the sharp increase in Electric Distribution Capital costs are solely borne by current rate payers - further impacting affordability for millions of Californians. At the facility level, data centers will host hundreds of on-site diesel generator backup installations that house thousands of industrial-sized diesel generators, which will significantly impact health outcomes in the immediate vicinity of the facilities and the air pollution will continue to impact communities for thousands of miles until it fully dissipates.
It is for these reasons that communities have started voicing serious concerns about where these data centers are built and the robustness of the processes for providing them access to core state and municipal resources. The key areas of concern I address in this testimony are environmental impacts and ratepayer protections.
5 CEC Electric Generation Capacity and Energy
4 CEC Electric Load-Serving Entities (LSEs) in California 3 CEC Data Center Load Forecasts, 2024 - 2040, see also California ISO Large Load 2 PG&E avoids ‘big bets’ as data center demand softens
1 PG&E Data Center Demand Pipeline Swells to 10 Gigawatts with Potential to Unlock Billions in Benefits for California
Environmental Impacts
There have been several studies written about the environmental impacts of large load data centers. The key concerns that are directly related to electricity are water consumption and air pollution. We’ll discuss each below.
Water Consumption
Large load data centers have implications for both scope 1 and scope 2 water use.6 In other words, water is consumed directly at the facility (scope 1), and indirectly during the power generation process (scope 2). The nexus between water and electricity is more apparent with scope 2 emissions, so we’ll start our discussion there.
Indirect Consumption
Indirect consumption is the water required to produce a data center's power. The process of generating electricity generally requires a lot of water. When generating power from coal,7 natural gas,8 or solar,9 water is used to create steam to turn the turbine. Nuclear power generation requires water both to create steam to turn the turbines and to cool the radioactive cores.10 Wind, geothermal, and hydropower are low-to-no water energy generation processes.
According to a landmark study by a US DOE laboratory, “14.7 gallons of cooling water must be withdrawn and 0.42 gallons consumed to generate one kilowatt-hour of electricity in the case of 91% of electricity production [in the US]”.11 The amount varies wildly depending on the local power mix. California’s energy mix is dominated by two sources of generation - natural gas (40%) and solar (23.44%), while large hydro (11.67%), nuclear (8.5%), wind (7.29%), and geothermal (4.84%), round out the mix.
The study listed the amount of water necessary to generate a kWh of energy for the sources in America’s energy mix. When we consider those numbers in relation to California’s energy mix, we get some idea of the amount of water we are currently using for energy generation:
11 Developing a Tool to Estimate Water Use in Electric Power Generation in the United States 10 Cooling Power Plants
9 Water Use Management
8 How it Works: Water for Natural Gas
7 Water Usage in Coal to Power Applications
6 How much water does AI consume? The public deserves to know 2 data center water consumption between 2019 and 2023.14 In 2023 alone, California consumed 47 billion gallons of water to generate 10.8 terawatts of electricity for data centers. The study estimates that California will consume between 74.54 and 116.63 billion gallons of water in indirect data center consumption in 2028.
Direct Consumption
Why do data centers use water? Traditional facilities have used air cooling for decades without incident. But modern data centers are fundamentally different. AI is powered by advanced GPU’s that are intended to run 24/7 at extremely high temperatures. For example, the chip that is powering Grok3 - NVIDIA’s H100 - runs at an average temperature of 188 degrees Fahrenheit and has a max operating temperature of 208F.15 The chips are structured to slow down their processing when they reach a 203F, and shut down when they reach 208F. If they run too long at the hottest temperatures, they start to wear and tear and don’t last as long. So in order to keep the chips running as efficiently as possible and get the most return on investment, modern data center operators are required to implement systems to cool the chips.
So how do you cool clusters of thousands to hundreds of thousands of these chips from 203F to 188F? Air conditioning generally can’t keep up at these levels of heat so modern data centers employ liquid cooling systems.16 The most common type of cooling system employed by modern data centers is the closed loop cooling system.
In a closed loop system, cold liquid is circulated past the server, the server heats it, the heated water is circulated through a heat exchanger where it is cooled, then the cold water is circulated past the server again.17 That loop is closed. No liquids gained, no liquids lost. But that isn’t the only loop. The key component of the closed loop system is the second loop, the heat exchange loop. That loop is open. The diagram below illustrates the process:18 18 The immersion cooling technology: Current and future development in energy saving 17 Understanding liquid cooling
16 When to move from air cooling to liquid cooling
15 NVIDIA H100 Product Brief
14 An Assessment of California Data Centers’ Environmental and Public Health Impacts 4 A complete closed loop liquid cooling system has 2 loops. The inner loop (1), that we discussed earlier, circulates cold liquid to the server and hot liquid to the heat exchanger. The system also has an outer loop, which is highlighted in green. The outer loop cycles cold water through the heat exchanger where it cools the inner loop, and it cycles hot water to the water cooling tower on the roof where it is cooled using an evaporative process. It is through this second loop, which is a mandatory part of every closed loop system, that data center facilities consume millions of gallons of water a year.19
There are facilities that use chillers, which are air cooling processes, for their heat exchange loops. But this is where the relationship between water and air becomes problematic for utility operators. Water and electricity have an inverse relationship when it comes to data center cooling. The less water you use (water towers), the more electricity you have to use (air chillers), and vis versa.20 So putting limitations on how much electricity a facility uses means requiring them to use more water, and putting limitations on how much water they can use means requiring them to use more electricity. Currently water cooling towers are favored for industrial-level cooling because of their energy efficiency,21 but this technology is coming under increasing scrutiny due to water scarcity concerns.
Water utility operators have few, if any real tools to reign in data center water usage in times of drought. There are currently no state or local laws or regulations requiring large load data center operators to slow or stop their operations in times of water stress. In fact, operators seem to rely on others’ water conservation efforts in times of drought to ensure uninterrupted operation:
21 Cooling Tower v. Chiller
20 The fallacy of closed loop cooling systems
19 Google Water Metrics Insights: 2025 Environmental Report v2 5 Through Valley Water’s implementation of conjunctive use programs, the Llagas groundwater subbasin has historically experienced well managed levels (City of Gilroy 2023). As a result of this management and ongoing efforts as part of Valley Water’s 2021 GWMP (Valley Water 2021a), the Llagas Subbasin is considered a reliable source of supply during water shortages. While pumping may exceed recharge during a drought, basin management practices would continue to prevent long-term adverse conditions (City of Gilroy 2023).22
Additionally, large load data center operators are increasingly relying on city and county wastewater treatment facilities to provide the millions of gallons of water they need for their closed loop systems. For reference, Google operates a 61MW facility in Lenoir, NC23 that consumed 327M gallons of the 351M gallons of water it withdrew last year.24 The City of San Jose has contracted with PG&E to fully build out a 250MW large load facility by Q2 2027,25 that will source it’s water exclusively from the San José-Santa Clara Regional Wastewater Facility.26 And while the facility currently only services commercial customers, the goal of its 2050 master plan is to “[m]eet 100 percent of annual water demand during non-drought years and at least 80 percent of demand in drought years.”27 The Los Angeles County Sanitation Districts, who own and operate one of the largest wastewater recycling programs in the world, are using their recycled water both for servicing customers and for groundwater recharge.28
So, while wastewater treatment facilities place a premium on serving industrial customers, they are also intended to play a critical role in ensuring that all Californians have access to clean, affordable water during our increasingly severe drought cycles. The onboarding of hundreds of large load data centers over the next 4 years is likely to exacerbate the tension between economic progress and water conservation, which raises the risk of operators being forced to shift back to less efficient air cooling systems. Transferring the strain of supporting data center cooling from the water utilities directly to the electricity utilities.
Air Pollution
Every large load data center produces PM2.5 pollution. Because all modern data centers use gas-powered generators for backup power.29 And as data center installations have grown, their backup generator installations have also grown, every watt of facility electricity must be backed up with a watt of diesel generation. And while these generators are not primary sources of facility power, they are run regularly.
29 The Critical Role of Generators in Commercial and Industrial Data Centers 28 Los Angeles County Sanitation Districts Water Reuse Program 27 Santa Clara Valley Water District Water Supply Master Plan 2050 26 Facebook: Mayor Matt Mahan's Post 11/11/2025
25 Implementation Agreement between Pacific Gas and Electric Company and City of San José 24 Google Water Metrics Insights: 2025 Environmental Report v2 23 Google Buys Solar Power for North Carolina Data Center 22 Amazon Data Center Final Gilroy Data Center Water Supply Assessment 6 Backup generator installations need to be run for hours every month to make sure they're working properly. California allows modern industrial sized diesel generators to be run 50 to 100 hours a year for maintenance and testing, with some facilities being allowed to run 100 hours a year.30 While older, more polluting models may run for 20 hours a year.31 Since large load facilities are so large, it is critical that they be encouraged to drop off of the grid in times of severe grid stress. In California, that usually means one of our increasingly long and frequent heat waves, where residential and office customers are using atypical amounts of power. Facility operators are allowed up to 50 hours of non-emergency backup power generation for these grid stress relieving purposes.32 Facility operators are also allowed to run their backup installations without limitation when the local utility performs a Public Safety Power Shutoff (PSPS).33
Public Health Impact
Large load data center backup generator installations don’t just run during emergencies, they run regularly. And that raises significant concerns around air quality, particularly from fine particle PM2.5 pollution. According to the EPA, the small particles that make up PM2.5 pollution can get deep into your lungs, and some may even get into your bloodstream.34 Once these fine particles enter your bloodstream they can attach themselves to your red blood cells.35 Which helps explain why the impacts of PM2.5 exposure aren’t limited to the lungs. PM2.5 exposure can cause a variety of health impacts including:36
● premature death in people with heart or lung disease
● nonfatal heart attacks
● irregular heartbeat
● aggravated asthma
● decreased lung function
● increased respiratory symptoms, such as irritation of the airways, coughing or difficulty breathing
There is no safe level of PM2.5 pollution and severe adverse health impacts can occur with 24 hours of exposure.37 According to the American Lung Association, the Bay Area is the 7th most polluted city in the country in terms of fine particle pollution.38 The Bay Area’ unique geography creates a near-daily inversion layer that serves to trap and concentrate pollutants close to the ground for large portions of the day.39 PM2.5 levels are also high in Los Angeles County, because heavy traffic, industrial emissions and the nature of the surrounding terrain, which 39 Why California Cities Rank Highest for Air Pollution 38 American Lung Association State of the AIR 2024 Report 37 Inhalable Particulate Matter and Health (PM2.5 and PM10) 36 Health and Environmental Effects of Particulate Matter (PM) 35 Researchers discover air pollution particles hitching a ride around the body on red blood cells 34 Health and Environmental Effects of Particulate Matter (PM) 33 Use of Back-up Engines for Electricity Generation During Public Safety Power Shutoff Events 32 EPA Clarifies Rules for Backup Generator Use
31 Use of Back-up Engines for Electricity Generation During Public Safety Power Shutoff Events 30 17 Cal. Code Regs. § 93115.6(a)(3)(A)
7 prevents pollutants from dispersing.40 For context, a single 99MW data center facility in San Jose’s Santa Teresa neighborhood will utilize thirty-nine diesel generators for backup power.41 The California Energy Commission found that running the 99MW backup system for 20 hours would emit as much pollution as 428 gas-powered cars driven for a year.42
The cumulative effect of the addition of hundreds of large load data centers hosting thousands of diesel generators, largely clustered near already heavily polluted areas, will have a devastating impact on public health.
Ecosystem Impact
Large load data center operators are increasingly buying farm land for their new facilities. While that has obvious implications for agricultural and livestock production, the location of these facilities near agricultural uses and fragile ecosystems raises a different risk - pollution. PM2.5 emissions aren’t just devastating for people, they have a severe adverse impact on land as well.
Specifically, when PM2.5 pollution settles on land or water, it can cause the following impacts:43
● making lakes and streams acidic
● changing the nutrient balance in coastal waters and large river basins ● depleting the nutrients in soil
● damaging sensitive forests and farm crops
● affecting the diversity of ecosystems
● contributing to acid rain effects.
The fine particles are carried on air currents and can travel thousands of miles, impacting vast swathes of land.44 But the impact on the local farmland, and neighboring parks and habitats will be the most significant. While California is home to only 4 percent of the country’s farms and ranches, we generate more than 10 percent of all U.S. agricultural value.45 Any property use that could undermine the effectiveness of California farmland must be carefully studied, and the tradeoffs publicly discussed.
Electricity Affordability
There is an assumption that all of the recent rate increases in the state have been due to data centers. That is fundamentally untrue. Wildfire risk was the largest driver of residential price 45 How California Agriculture has Shaped the State and the Country 44 The Unpaid Toll: Quantifying and Addressing the Public Health Impact of Data Centers 43 Health and Environmental Effects of Particulate Matter (PM) 42 California needs more data centers—but builders are using a legal loophole to power them with diesel 41 Great Oaks South Backup Generating Facility Small Power Plant Exemption 40 An Assessment of California Data Cneters’ Environmental and Public Health Impacts 8 increases through 2024.46 But that trend is beginning to change. We can see that change in PG&E’s latest Risk Spending Accountability Report. Electric Distribution Capital Costs drove nearly 60% of PG&E’s overall $1B cost increase from 2023 to 2024.47 According to Table 3-2, the Capital Comparison Table, while wildfire mitigation costs (lines 17 and 21b) increased marginally yoy ($8.18M), the costs for imminent hazard responses (line 27) and automation and protection (line 6), increased significantly yoy ($142.9M and $15.21M respectively).48 Wildfires will continue to be a major driver of residential price increases, but hyperscale data center activity is playing an increasing, but not fully recognized role.
According to PG&E, 95% of the 18 new large load data center projects that are currently in final engineering will enter service by 2030, with several beginning service as early as next year.49 Which means the procurement and build out of the distribution for many of these projects has already begun. Table 3-2 shows that nearly $325M of the yoy increase in Electric Distribution Capital Costs was driven largely by the kinds of capital expenses that are necessary to support data center capacity increases: Line 2 – Electric Distribution Line and Equipment Capacity ($62.46M), Line 3 – Electric Distribution Install/Replace OH Poles ($210.63M), Line 18 – Electric Distribution Substation Capacity ($32.97M), and Line 19 – Electric Distribution Substation Replace Other Equipment ($18.77M). While these activities have safety implications, they are also the exact expenditures required to support the capacity increases hyperscale data centers demand.
There is currently no reporting category for investment spending in the cost allocation process.
The introduction to the Report’s Electric Distribution Imputed Adopted v. Recorded Comparison Section provides the following description of the section’s information [emphasis added]:
This section includes the following information for the Electric Distribution functional area: a comparison of the total 2023 and 2024 imputed adopted spend vs. the actual spend as well as the required data points per program as defined and required in Decision (D.) 23-11-069.1 This section also includes, for programs that are related to safety, reliability, or maintenance, the Major Work Category (MWC)/Maintenance Activity Type (MAT) Code descriptions, imputed adopted vs. actual cost comparison details and variance explanations. As required by D.19-04-020,2 the MWC/MAT Code descriptions include a discussion of how each program/project relates to safety, reliability, or maintenance.
PG&E’s 2027-2030 Rate Case Application50 makes no mention of large load customers and the Risk Spending Accountability Report has no category for these kinds of invesments. So there is currently no way to identify how much of the proceeds from recent and future rate increase will be used for maintaining and improving the safety and reliability of the grid for current customers, and how much is being invested in building out the grid to support the large load data center pipeline.
Major data center hubs across the country are enacting rules and legislation to address this 50 2027 General Rate Case Application of Pacific Gas and Electric Company (U 39 M) 49 PG&E avoids ‘big bets’ as data center demand softens 48 See Appendix
47 Pacific Gas and Electric Company’s (U 39 M) 2024 Risk Spending Accountability Report 46 Factors influencing recent trends in retail electricity prices in the United States 9 concern. The utilities commissions in Michigan,51 Oregon,52 and Ohio53 recently passed rules to protect ratepayers from absorbing the initial investment of data center-driven capacity, transmission, and distribution upgrades. The initial text of SB57 was intended to accomplish the same goal.54 While the version of SB57 that was enacted was stripped of all ratepayer protection language, the concerns that drove its drafting remain. Given that Cost Allocation Filings are planned well in advance of securing a leaseholder for project,55 lack of clarity on the full purpose of the increase leaves rate payers exposed to significant risks. Some particularly significant risks are the abandonment or delay of a data center, and the overreliance on one or a few large customers.
Risk - Abandonment/delay
While overestimation of product demand is often raised as a key concern in ratepayer affordability discussions,56 the true risk of making long-term utility decisions based on business forecast is something much more comprehensive and opaque - shifting business priorities.
Companies delay or abandon infrastructure products for a host of reasons, and the impact on communities and municipal governments can be devastating. The most notable recent example is the planned Foxconn mega manufacturing plant in Wisconsin.57 The partnership for the development of a $10 billion dollar, 13,000 job, 20 million square foot manufacturing campus was announced in 2017. By 2019, the local government had spent $152 million to purchase 132 properties for the facility along with $7.9 million in relocation costs. In 2021, Foxconn restructured the deal to a $672 million dollar investment with 1,454 jobs. The local community was not refunded its investment and the tax incentives that initially made the deal extremely profitable for the government disappeared.
Foxconn didn’t overestimate product demand. The planned was upended by lack of proximity to crucial suppliers of the flat panel display parts that are required to run a flat panel display manufacturing plant, and local opposition to the scale of the municipal’s investment of local taxes, the extensive grant of water rights, and the grant of eminent domain authority.58 All of which are issues that are being raised by local opposition to large load data center development.
Businesses can, and will delay infrastructure development for a variety of reasons. Amazon delayed several large load data center leases earlier this year due to economic uncertainty.59 A
1.8 million square foot data center development in Kansas City has been delayed due to a lawsuit over community and environmental concerns.60 Two fully built 48MW data centers in Santa Clara may be delayed for years until Silicon Valley Power is able to provide them power.61
Businesses also abandon builds for business reasons. There is a risk of innovation making the 61 Data centers in Silicon Valley stand empty, awaiting power 60 Lawsuit delays $12B data center in Kansas City as community, environmental group voice concerns 59 Amazon has halted some data center leasing talks, Wells Fargo analysts say 58 Foxconn mostly abandons $10 billion Wisconsin project touted by Trump 57 Foxconn mostly abandons $10 billion Wisconsin project touted by Trump 56 Big Tech’s A.I. Data Centers Are Driving Up Electricity Bills for Everyone 55 Implementation Agreement between Pacific Gas and Electric Company and City of San José 54 SB-57 Electrical corporations: data centers: report. 53 PUCO orders AEP Ohio to create data center specific tariff 52 Data centers to pay new electricity rate under Oregon bill that has passed 51 Michigan Regulator Approves Plan to Shield Residential Customers from Data Center Rate Hikes 10 current glut of large load data centers obsolete. Deepseek proved that it is possible to create performant large language models while using less data center compute.62 Small language models, which require much less compute so can operate with a smaller data center footprint, are increasingly being seen as the next generation of enterprise AI.63 Innovations in chip design may allow for equal or greater performance with a smaller, less consumptive footprint.64
There’s also the spector of bad economics driving abandonment. A recent MIT study found that 95% of enterprise generative AI pilots, the exact products that large load data centers are built to support, fail.65 This result places the fundamental value of enterprise level generative AI applications, and the large load data centers that power them, at risk. Microsoft is shifting from new large load data center development to updating existing facilities.66 And data center financing, which has been largely a no-brainer in recent years, is beginning to come under greater scrutiny.67
Risk - Overreliance on a big customer/customers
Virginia, the home of the largest data center cluster in the country performed a detailed audit and review of data center electricity.68 One of the key areas of risk they flagged was co-ops and smaller LSE’s overreliance on a single or very few large customers. Specifically, the found that “delayed or disputed payments from a single large customer could create substantial financial liabilities.” California is home to 80+ California Load Serving Entities, including several co-ops and community choice aggregators.69 Virginia proposed a subsidiary company structure to protect co-ops from this risk, which may or may not be the best solution for California. But the risk needs to be discussed and mitigations developed.
As mentioned in the air pollution discussion, grid operators rely on large customers to engage in some sort of load shifting to alleviate the grid in times of extreme stress, like heatwaves.
Owner-operated data centers have the ability to load shift in these situations - move compute workloads from a grid-stressed facility to a facility that is not undergoing stress.70 However, load shifting isn’t always feasible. And load shifting isn’t a strategy that co-location operators can employ. Co-location data center facilities don’t own any of the process that go on in the data center. They own the facility itself and rent space to companies to run their operations. Which is why, in times of grid stress, co-location facilities resurt to running their diesel generator installations.71 So, in times of grid stress, overreliance on one or a few big customers may significantly limit the options a grid operator has to safely ensure reliability of the grid. If operator-owned customers can’t load shift, and co-location customers don’t have that option, 71 EPA Clarifies Rules for Backup Generator Use
70 Flexible Data Centers and the Grid: Lower Costs, Higher Emissions? 69 CEC Electric Load-Serving Entities (LSEs) in California 68 Virginia Joint Legislative Audit and Review Commission (JLARC) Data Center Electricity 67 Risks in data center lending: Development delays and SLA breaches 66 Microsoft Abandons More Data Center Projects, TD Cowen Says 65 The GenAI Divide: State of AI in Business 2025
64 NVIDIA Unveils Its Most Affordable Generative AI Supercomputer 63 The Case for Using Small Language Models
62 How Did DeepSeek Build Its A.I. With Less Money?
11 then grid operators have to consider the available generator hours their large customers have available as part of their outage prevention efforts.
But this overreliance on a few large customers is also a concern at the state level. Capacity forecasting and planning becomes difficult when infrastructure plans are directly tied to a single industry’s business. There are a variety of non-traditional business risks that have to be factored in. Risk that are not easily modeled, and that are wholly out of the utility’s control. Building hyperscale data center infrastructure isn't a tried and true investment strategy, it's a bet. A bet that the technology will continue to scale in its current form. A bet that the business value will meet forecast projections. A bet that innovation will not make large load data centers obsolete.
A bet that utilities are being asked to make with the lives and livelihood of every customer it services today and over the next 30 years.
Living in California is expensive. And in a state where lower-income Californians spend 80% of household expenses on basic necessities (food, housing, transportation, and health care), residential ratepayers are increasingly being forced into heat or eat choices in the winter, and sweat or starve choices in the summer. In an economic environment that is worsening for the least economically stable amongst us, residential ratepayers have been making tradeoffs between paying their rent and paying their utilities.72
Given the history of ratio utility billing in the Bay Area and Los Angeles, the additional impact of large load data centers may push some of our most vulnerable past the breaking point. A renaissance of RUBs in data center hotspots would cause costs go down in the aggregate by shifting costs from high usage to low usage users. While this would be a positive net outcome, it would come at the expense of the lowest ratepayers. Exacerbating the pain of our most vulnerable populations.
Conclusion
Every system in creation has constraints. It is an unalterable fact of life. Our state can only generate so much electricity. It only contains so much drinkable water. The key components for data center expansion are not unconstrained. Data center growth is inherently constrained by the availability of its key inputs. That is not bias, it is fact.
The goal of this Commission, and every other governmental entity that interacts with it, is to ensure a reasonable quality of life for all California residents. That means striking a balance between competing interests, and enacting restraints when corporate interests seek to overwhelm the public interest. The speed and scope of large load data center requests poses a threat to the water, air, and affordability of all Californians. It is for large load data center operators to provide proven, scalable, independently verifiable methods for operating their facilities within the limits of California’s resources and in harmony with the government’s responsibilities to its residents.
72 Rising Utility Costs Compound California's Housing Crisis 12 I conclude this testimony with some brief remarks in each area and recommendations for future research.
Water
Google withdrew 10 billion gallons of water to cool it’s global fleet of modern data centers last year, and consumed 8 billion of those gallons. And Google has access to the most advanced data center cooling methods and has full to control to granularly manage facility operations. Yet they’ve consistently consumed nearly 80% of the water they withdraw to cool their facilities.
The scale of direct and indirect water consumption by modern data centers is hard to fully grasp. The comprehensive impact of the rapid escalation of large load data centers on California’s water supply cannot be overstated.
Future Research
● Indirect water consumption of the specific power mix that supports existing California large load data centers.
● The specific environmental factors that make data center air cooling feasible and the average number of days per year that the climate in data center hotspots is conducive to data centers relying solely on air cooling systems.
Air
Communities cannot survive without clean air. But life is tradeoffs. So we allow noxious gas-spewing industrial sized diesel generators to operate as a backstop for critical services. The tradeoff between occasional pollution and a fully operational police force, hospitals, and fire stations is unquestioned. But there are serious questions about the tradeoff between uninterrupted modern data center operations and the scale of the regular pollution generated by their massive industrial diesel generator installations. That is a tradeoff that should be better understood and debated openly.
Future Research
● Concentrations of PM2.5 particles in communities within 20 miles of a large load data center between 2019 and 2024, with a specific focus on the speed and severity of changes that occurred in marginalized communities.
● Concentrations of PM2.5 particles in soil or waterways located within 20 miles of a large load data center between 2019 and 2024, with a specific focus on the speed and severity of changes that occurred on farmland or fragile ecosystems. ● Review how CEQA and local air quality management district rules regulating diesel generators can better address cumulative air pollution thresholds that integrate all currently planned facilities and the various ways operators are allowed to run their installations.
13 ● Consider the impact of adding requirements for air quality monitor installations to Diesel-Fueled CI Engine (>50 bhp) registration and small power plant exemption processes.
Affordability
Electric capacity is long lead infrastructure. Planned decades in advance due to the amount of time it takes to bring on new sources of energy generation. When a grid is 100% utilized, the city can no longer grow. It can't handle grid disruptions. A fully utilized grid is a fragile grid.
When we overutilize capacity today, we are constraining tomorrow's growth and opportunity.
This is why infrastructure is a quasi-governmental function. Infrastructure isn't about driving shareholder value. It's about survivability. Electricity is necessary for survival, so it must be provided to the rich and poor alike. Utilities have every right to turn a profit. But if people are investing in California infrastructure to get rich,73 we're doing something deeply, deeply wrong.
The revenue focus of shareholder-owned utilities must be balanced by the survivability focus of their government partners.
Future Research
● Considering how much non-data center capacity growth is forecast through 2040, determine the amount of investment necessary for hardening the grid against environmental threats.
● Review current regulations, policies, and procedures to determine how to develop a reporting regime that delineates between single industry investments and investments for broader grid reliability and safety improvements.
Respectfully Submitted,
Masheika Allgood
AI Ethicist, Founder
AllAI Consulting, LLC | allai-us.com
Environmental Platform: Taps Run Dry
73 PG&E Corporation 2025 First Quarter Earnings Presentation 14 Appendix
See chart on next page.
15 This message is from outside the City email system. Do not open links or attachments from untrusted sources.
I beg you, to reconsider these approving of item 8.2 approving the developer of Prologis from creating data centers. This will ruin not only the City of San Jose today for us, but for our future generations.
Thank you,
Clarissa
11/18/25, 8:39 AM Inbox - Agendadesk - Outlook
https://outlook.office365.com/mail/[email protected]/inbox/id/AAQkADUxOWI4ZjE3LTRkNDEtNGUzMS04MjAwLTIzNzdiYTdkMjc5NAAQA… 2/2
This message is from outside the City email system. Do not open links or attachments from untrusted sources. development and construction is complete, are few and far between. The risks compound beyond this - these data centers are powered by GPUs that are obsolete in 5 years or less, for an AI bubble that despite hundreds of billions of dollars of investment is showing no signs of profit or return on investment.
Fear Of Missing Out should not be the primary motivator of municipal planning.
It is disappointing, though perhaps unsurprising, that the Mayor chose to enter into these negotiations with Prologis, a company contributing to Chicago’s rapidly multiplying data center population (and rapidly worsening air quality), without involving the community who will bear the brunt of its effects. I appeal to my council member, Anthony Tordillos, to represent his constituents’ voices. During his campaign, he attacked his opponent for receiving donations from PG&E. Well, now he has his opportunity to demonstrate that this wasn’t just campaign talk. I implore him to take that opportunity and call for community discussions, hearings, and a vote on whether San Jose residents value their health and wellness over short-sighted corporate interests.
Sincerely,
Richard Saunders
11/18/25, 8:39 AM Inbox - Agendadesk - Outlook
https://outlook.office365.com/mail/[email protected]/inbox/id/AAQkADUxOWI4ZjE3LTRkNDEtNGUzMS04MjAwLTIzNzdiYTdkMjc5NAAQA… 2/2
PG&E is the monopoly transmission and distribution provider.
PG&E is funding staff who support these projects.
PG&E helped evaluate the competing developers.
PG&E benefits financially from whichever developer brings the biggest loads which happens to be Prologis.
PG&E serves as its own oversight body.
(2) A PG&E power commitment that is not guaranteed,
The primary justification for selecting a developer today is PG&E’s promise to deliver 250 MW of transmission-level power to this site in 30–36 months.
The timeline is explicitly described as an estimate, not a requirement. The City–PG&E Implementation Agreement contains multiple provisions that allow PG&E to delay or stall transmission-level power delivery to large-load projects—including the RWF buffer lands—without breaching the agreement or incurring penalties. The agreement contains no remedies, liquidated damages, or performance guarantees if PG&E’s transmission upgrades are late. Yet PG&E’s delivery depends on multiple conditions outside PG&E’s control, including:
CAISO approval (the biggest dependency) the California Independent System Operator must approve any new transmission-level service. The Implementation Agreement explicitly states that the 250 MW timeline requires: “Any required regulatory approvals, including concurrence by the CAISO.” If CAISO does not approve the interconnection as proposed—or requires upgrades—PG&E cannot deliver the power, regardless of intentions.
CEQA and local permits timeline assumes successful and timely CEQA and permit approvals.
Transmission interconnections of this scale typically take 5–10 years, not 30–36 months.
With PG&E funding staff, influencing selection, and benefiting from the project, the City’s oversight power is compromised. The City is being asked to select a developer whose entire plan depends on a power timeline that PG&E cannot guarantee without CAISO approval. Selecting a developer based on an uncertain power schedule exposes the City and its residences to major execution risk. (3) None of the developers have completed a hyperscale campus at this scale A 250+ MW hyperscale data-center campus, a type of project that requires: complex transmission engineering, multi-substation coordination, large-scale cooling, and years of environmental and utility compliance. The City’s staff memo states the preferred developer will build:
1.684 million sq ft of data center buildings,
785,000 sq ft of advanced manufacturing,
Five Electrical substations.
Prologis is a logistics developer, but their data-center experience is limited:
One 30 MW conversion, not a hyperscale build
One 600 MW campus under construction, not complete/operational They have never completed or operated a 250+ MW hyperscale data-center campus.
Catellus and Deca list data centers in their portfolios, but their only known hyperscale project— Quantum Frederick
—is still in development, facing delays and environmental issues. There is no publicly available evidence they have completed a multi-substation, 100+ MW campus.
All the shortlisted developers—Prologis, Catellus, and Deca—are new entrants to hyperscale data-center development. None have a verified, completed, operational project at the magnitude required for this site. Council is being asked to select a developer whose ability to deliver this project is unproven, while relying on a power commitment that PG&E cannot guarantee.
If either PG&E or the selected developer falls behind or fails to deliver, the City faces stranded or partially developed land. The City could be locked into an ENA while power timelines slip and construction stalls and loss of control over critical public lands. Additionally, Large-load projects—especially at 250+ MW—carry major implications for grid reliability and emissions. Before selecting a developer, Council and the community at large 11/18/25, 8:40 AM Inbox - Agendadesk - Outlook
https://outlook.office365.com/mail/[email protected]/inbox/id/AAQkADUxOWI4ZjE3LTRkNDEtNGUzMS04MjAwLTIzNzdiYTdkMjc5NAAQA… 2/4 should be allowed to see the full qualifications submitted by these shortlisted developers. The RWF lands represent one of San José’s most important public assets. Council should not commit this land to a developer until:
PG&E’s structural conflict of interest is addressed, by ensuring PG&E’s staff funding and advisory role do not influence developer selection or City oversight.
CAISO-dependent power deliverability is independently verified, and Additionally, neither the Implementation Agreement nor Solicitation of Qualifications mentions ANY ENVIRONMENTAL CLAUSES & CONDITIONS FOR HYPERSCALE DATA-CENTER DEVELOPMENT.
This
negligence must be remedied above all else to protect residents, ratepayers, water, air quality, grid reliability, and the regional environment for present and future generations.
I request Council and Staff to
look into the following:
1. Water Use Disclosure, Limits, and Drought Contingencies The developer must submit a full water-use profile for cooling, including maximum daily and annual water demand under peak AI-compute loads.
Full disclosure of planned water cooling, evaporative cooling, or liquid-to-air systems.
Mandatory water-use limits during drought years.
Prohibition on using potable drinking water for cooling during declared drought emergencies.
Installation of recycled water lines prior to occupancy.
Annual water-use reporting to ESD and the City Council.
2. Energy Source Transparency (Clean vs Fossil Electricity Mix) Before approval, the developer must disclose the expected energy mix they will consume, including what percentage will come from fossil-fueled generation vs renewables.
Annual disclosure of carbon emissions tied to electricity usage.
No marketing the site as “clean-powered” unless verified by third-party audits.
3. Air Quality & Backup Generator Restrictions
The developer must disclose the number, type, and fuel source of backup generators and demonstrate compliance with local air-quality district standards.
Limits on total diesel generator capacity.
4. Cooling System Heat Discharge Standards
Cooling system waste heat discharge must not raise local heat levels, impair air quality, or cause thermal pollution.
Install heat recovery or heat reuse systems where feasible.
No venting of heat exhaust toward residential neighborhoods.
5. Grid Impact & Load-Shedding Protections for Residents The City must ensure that data-center loads do not impair grid reliability for residents or critical services.
Hyperscale data centers would not cause increased outage frequency or duration for residential customers.
6. Noise Pollution Limits
Cooling towers, generators, and substation equipment must comply with strict noise limits at property lines.
Developer must design noise abatement strategies.
Continuous noise monitoring with public reporting.
No low-frequency or tonal noise impacts to adjacent neighborhoods. 11/18/25, 8:40 AM Inbox - Agendadesk - Outlook
https://outlook.office365.com/mail/[email protected]/inbox/id/AAQkADUxOWI4ZjE3LTRkNDEtNGUzMS04MjAwLTIzNzdiYTdkMjc5NAAQA… 3/4 This message is from outside the City email system. Do not open links or attachments from untrusted sources.
7. Environmental Justice (EJ) Impact Assessment
All hyperscale development on RWF lands must undergo an Environmental Justice Impact Assessment.
Identify which neighborhoods will bear air, noise, and traffic impacts.
Require mitigation actions if impacts are borne disproportionately by vulnerable communities.
Require direct community input in the assessment process.
8. Water Quality & Hazardous Materials Protection
Computer equipment, batteries, cooling fluids, and fuel storage pose contamination risks. Given proximity to the San José–Santa Clara Water Pollution Control Plant, the developer must demonstrate that no hazardous materials will jeopardize water quality.
Spill-prevention and containment systems.
Hazardous-material storage disclosures.
Independent third-party audit of chemical handling and storage plans.
9. Ongoing Annual Environmental Reporting
Developer must provide annual public environmental performance reports to the City Council. Reports must include:
Water use
Energy use and carbon intensity
Generator run hours
Noise monitoring results
Heat discharge data
Grid impact assessments
Any compliance violations and remedies
10. Environmental Performance Bond
A performance bond ensures the developer is financially accountable for environmental impacts or failures.
Developer must provide an Environmental Performance Bond held by the City, to be used if environmental mitigation fails or compliance is violated
P.S. - The City is approving Microsoft’s data center infrastructure on the RWF buffer lands (Item 8.1) while simultaneously selecting a developer for a much larger data center district on those same lands (Item 8.2), without analyzing the cumulative environmental impacts — this could be considered CEQA segmentation, which would be illegal.
11/18/25, 8:40 AM Inbox - Agendadesk - Outlook
https://outlook.office365.com/mail/[email protected]/inbox/id/AAQkADUxOWI4ZjE3LTRkNDEtNGUzMS04MjAwLTIzNzdiYTdkMjc5NAAQA… 4/4