THE CLOUD HAS A POWER BILL. WHO GETS STUCK WITH IT?

VAULT · 09 EARTH

Published by Langhorne P. Breedlove

MONDAY, NOVEMBER 9, 2026 · 2:50 AM ET

Tags: Data Centers · Electricity · Grid Reliability · Water · Utility Rates · Environmental Justice

Dear fellow American proletariat: Someone told us our photographs, bank accounts, medical records and artificial-intelligence experiments live in the cloud. Charming. The cloud is a building full of computers, wires, cooling equipment, backup generators, and an unusually serious electric bill. Sometimes that building also needs enough water to make the local utility ask whether anybody checked the reservoir before signing the development agreement.

The question is not whether America should have data centers. Modern communications, hospitals, banks, laboratories and public services rely on computing. The question is whether extraordinary new electricity and water demand is disclosed, planned, financed and governed without leaving existing households and communities to absorb undisclosed costs. American Proletariat's related Justice policy proposes accountable large-load planning; this Vault essay examines why the problem exists and where the evidence leads.

ROOTS · HOW THE CLOUD BECAME AN INDUSTRY

Corporate server rooms evolved into large colocation facilities, cloud platforms and hyperscale campuses serving many customers and increasingly intensive artificial-intelligence workloads. Efficiency improvements once moderated the sector's growth in electricity use, but demand for accelerated computing has changed that trajectory. Lawrence Berkeley National Laboratory estimated 176 terawatt-hours of U.S. data-center consumption in 2023, around 4.4% of national electricity use. That is historical modeled consumption, not a claim about the share measured in November 2026. (Source 01)

The laboratory's June 2026 update projected a 2030 central estimate of 11.8% of U.S. electricity consumption, with a 9.5%–15.3% scenario range. These percentages depend on future equipment shipments, utilization, efficiency, cooling, and national electricity demand. They are not reserved quantities of energy or proof that each proposed data-center project will be constructed. (Source 02)

THE ELECTRICITY QUESTION

A BIG LOAD DOES NOT FIT INTO A SMALL FORECAST

Large campuses can request hundreds of megawatts, creating requirements for generation, substations, transmission and distribution infrastructure. Grid planners must distinguish firm commitments from speculative applications: counting the same possible project in multiple development queues can inflate forecasts, while underestimating firm loads can jeopardize reliability. A computer campus cannot obtain continuous grid power merely by promising that innovation is important. (Sources 03, 04)

In May 2026, the Energy Information Administration explained that its Annual Energy Outlook 2026 projected server electricity consumption reaching between 446 and 818 billion kilowatthours in 2050 under different cases. That measure concerns server consumption, not necessarily the full facility's electricity demand, and the long-run scenarios should not be misrepresented as precise predictions. (Source 05)

WHO PAYS FOR THE WIRES?

Suppose a utility upgrades transmission for a very large customer, but the customer cancels the project. The infrastructure may remain, and its cost must be allocated somewhere. Federal and state regulators determine different parts of that allocation, subject to applicable tariffs and law. A transparent agreement can require deposits, contracted minimum payments, or reimbursements for dedicated facilities so that customers already on the grid are not unfairly exposed to stranded costs.

On June 18, 2026, the Federal Energy Regulatory Commission directed six regional grid operators to justify or reform their large-load interconnection tariffs under Section 206 of the Federal Power Act. The proceedings address customer protections, cost allocation, reliability, transparency and efficient connections. They are live regulatory proceedings—not final proof that all proposed protections have been implemented or that every customer has paid its share. (Source 04)

THE WATER QUESTION

WITHDRAWAL IS NOT THE SAME AS CONSUMPTION

Water withdrawal is the amount taken from a source; consumption generally refers to water not promptly returned to that source, such as water evaporated in cooling. Discharge and water quality are separate concerns. Air cooling, evaporative cooling, closed-loop liquid systems and hybrid configurations have different electricity and water consequences. There is no defensible universal figure for water used by one AI request, regardless of where or how it was processed. (Source 06)

The Congressional Research Service reported in July 2026 that direct data-center water use is challenging to measure comprehensively, and that the relative impact depends on regional supply, seasonality, other users, drought conditions and cooling design. It cited an International Energy Agency illustration in which a 100-megawatt U.S. facility may directly consume water comparable to roughly 2,600 households each day. That is a modeled example, not a meter reading from every 100-megawatt center. (Source 06)

Indirect water consumption can arise from generating the electricity a facility uses. Moving from evaporative onsite cooling to a more electricity-intensive system may reduce water consumed at the building without necessarily eliminating water used elsewhere in the energy system. Conversely, some electricity sources use little operational water. A serious comparison must identify boundaries and location rather than add incompatible numbers into an alarming national total. (Sources 01, 06)

UTAH · A CONCRETE POLICY EXPERIMENT

Utah's 2025 S.B. 132 established rules for large electrical loads, including responsibility for just and reasonable incremental costs of generation, distribution and, where federally permissible, transmission improvements. The law attempts to prevent expenses attributable to a large customer from being passed to other utility customers without appropriate allocation. Its effectiveness depends on contracts, regulatory implementation and enforcement. (Source 07)

Utah's 2026 H.B. 76 created reporting duties for certain new data centers larger than 10,000 square feet that intend to withdraw at least 75 acre-feet of water annually. Effective July 1, 2026, the law provides for preconstruction and annual reporting through the Division of Water Rights. Seventy-five acre-feet is a reporting threshold, not a water allowance, a statewide ban or a finding that every facility below it has no effect. (Source 08)

Utah's situation illustrates a broader issue for arid regions: water rights, municipal supply capacity, local development approvals and environmental consequences cannot be resolved solely by a developer's promise to use an efficient cooling system. The Great Salt Lake and other western watersheds already face multiple demands, and each proposed use should be assessed within its actual source basin.

WHAT SHOULD COMMUNITIES KNOW BEFORE APPROVAL?

THE REAL POWER REQUIREMENT

A large-load application should disclose projected maximum megawatts, annual megawatt-hours, commissioning schedule, backup supply, expected demand flexibility and grid upgrades. Independent reviews should separate confirmed commitments from speculative proposals. Proprietary computer designs and cybersecurity details need not be exposed to disclose fundamental resource demands.

THE ACTUAL WATER SOURCE

Publish expected and measured withdrawals, consumption, discharge, water source, seasonal peak, drought response and cooling method. A permitted withdrawal does not demonstrate that the source can sustainably meet every competing future need. State water law, water-quality rules, Tribal rights and local permits must be considered in their own jurisdictions.

THE PUBLIC FINANCIAL EXPOSURE

Before public incentives are approved, disclose construction and operating employment separately, local tax effects, tax exemptions, utility investments, developer guarantees, water-system expansion and likely consequences for residential bills. A large private capital investment is not equivalent to an equally large public fiscal benefit. Jobs promised during construction are not the same as permanent positions.

THE ENVIRONMENTAL & RELIABILITY EFFECTS

Evaluate electricity-generation emissions, diesel backup operation, air permits, transmission impacts, waste heat, equipment replacement and emergency curtailment. A renewable-electricity procurement contract may be valuable, but it does not establish that the regional grid has zero incremental emissions in every hour. Reliable electricity for hospitals and existing neighborhoods must be part of a large-load planning decision.

THE GOVERNMENT'S ACTUAL POWERS

Federal law governs relevant interstate transmission rates and wholesale electricity matters through FERC; state commissions oversee many retail utility issues; states administer water rights under their applicable legal regimes; local authorities handle significant zoning decisions; and federal or state environmental laws can apply to emissions or discharges. Congress could expand national reporting within constitutional limits, but a proposed single nationwide approval process cannot erase existing state or Tribal jurisdiction.

A sensible reform would identify who measures and verifies each element, who makes a permit decision, who handles appeals, who is responsible if a project fails, and where the public can find actual—not merely projected—outcomes. Approvals should be consistent, timely and reviewable without treating delay as the only form of accountability.

THE ONE-YEAR & TEN-YEAR LEDGER

This is a problem about public spending and private cost allocation. A facility's billion-dollar construction budget is not automatically a federal appropriation. A utility-financed transmission line is not necessarily a Treasury expenditure. A tax exemption is a fiscal concession, while a developer-paid grid upgrade may cost ratepayers little if the contract is enforceable. These should appear in separate columns.

For an illustrative national transparency pilot, assume 100 specialist analysts, engineers, water scientists, auditors and legal staff at a fully loaded annual average of $225,000. Personnel would cost $22.5 million per year. Add a hypothetical $17.5 million annually for public data systems, independent sampling, outreach and evaluation, yielding $40 million recurring. A one-time $30 million setup budget would bring first-year gross public spending to $70 million. Holding operating expenses flat for ten years produces a $430 million undiscounted total including startup. These are planning assumptions, not a Congressional Budget Office estimate or a proposal to finance private construction.

A separate utility cost-recovery example makes the stakes visible: if a $300 million dedicated grid improvement is contractually attributable to one new large load, the essential question is how much that customer pays, when, and with what financial security. A hypothetical 100% enforceable developer reimbursement would shift that identified cost away from existing ratepayers; a 50% reimbursement would leave $150 million to allocate elsewhere under applicable rules. These are examples of allocation, not proof that every grid improvement benefits only one user or that any specific project has this cost.

The mandatory ten-year fiscal model must include agency administration, utility investments, private contributions, financing and interest, canceled-project exposure, water infrastructure, incentives, potential environmental mitigation, tax receipts, employment effects and uncertainties. It must compare the current-law baseline with alternative disclosure and cost-allocation regimes. Reduced water use, cleaner energy or higher economic activity should be counted only when verified or credibly modeled.

WHAT COULD GO WRONG?

A strict system may slow useful computing infrastructure, increase development costs or encourage companies to build in jurisdictions with weaker safeguards. Excessive reporting can burden smaller sites or reveal sensitive business information. On the other hand, loose utility commitments can shift stranded infrastructure costs to households, inaccurate water projections can burden communities, and rushed siting can create reliability or pollution problems.

There is also a real engineering tradeoff: cooling methods that save direct water can require additional electricity; systems that conserve electricity may consume more water in some climates. A drought-sensitive desert project and a cooler-region project should not be judged with a single nationwide water-per-server number. Regulators should compare location-specific alternatives and record the assumptions.

MEASURING SUCCESS

Publish forecast versus actual peak load and annual electricity use; grid-upgrade costs attributed and recovered; abandoned-load exposure; changes in residential and small-business rates; source-specific water withdrawal, consumption and discharge; facility emissions and outage events; permanent employment; incentives provided; and permit compliance. Use basin-level analysis and appropriate protections for proprietary security details, sensitive infrastructure and private records.

THE POINT

The country needs computers. It also needs affordable electricity, dependable water supplies, and accountable decisions about infrastructure. Large data centers can be economically useful while imposing genuine local costs. The public deserves facts about both, not a caricature in which every server farm is a villain or every investment announcement is a public blessing.

The cloud is made of metal, land, electricity and water. A business that wants to build one should know exactly which resources it needs—and the people already living there should know exactly what they are being asked to share.

SOURCES & FURTHER READING

01 · Lawrence Berkeley National Laboratory — 2024 United States Data Center Energy Usage Report, December 19, 2024.

https://eta.lbl.gov/publications/2024-lbnl-data-center-energy-usage-report

02 · Lawrence Berkeley National Laboratory — United States Data Center Energy Usage Report: 2025 Update, published June 2026.

https://bies.lbl.gov/publications/united-states-data-center-energy-2025

03 · Congressional Research Service — Data Centers and Their Energy Consumption: Frequently Asked Questions, updated January 23, 2026.

https://www.everycrsreport.com/reports/R48646.html

04 · Federal Energy Regulatory Commission — Large-Load Interconnection Show-Cause Orders, June 18, 2026.

https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration

05 · U.S. Energy Information Administration — Data Center Server Energy Use Grows Across the Commercial Building Stock, May 19, 2026.

https://www.eia.gov/todayinenergy/detail.php?id=67704

06 · Congressional Research Service — Data Centers and Water: Frequently Asked Questions, July 31, 2026.

https://www.everycrsreport.com/reports/R49057.html

07 · Utah Legislature — S.B. 132, Large Load Electric Service Amendments, enrolled 2025.

https://le.utah.gov/Session/2025/bills/enrolled/SB0132.pdf

08 · Utah Division of Water Rights — Large Data Center Water Use Reporting, effective July 1, 2026.

https://waterrights.utah.gov/data-centers/

10 RECOMMENDED ARTICLES, ESSAYS & POLICIES

→ Explore Accountable Data Centers & Grid Planning

→ Explore Grid Modernization, Water Resilience & Natural Disasters

→ Explore Great Salt Lake & Western Water

→ Explore Ten-Year Fiscal Accountability

→ Explore Universal Public Broadband & Internet

→ Explore Greens

→ Explore Bourbons

→ Explore Elite & Industrial Power

→ Explore Reformers, Abolitionists & Civil Rights

→ Explore Right to Repair & Seed Ownership

WE ARE ONE.  ·  E PLURIBUS UNUM.