Subscribe now

In 2024 and 2025, the world's largest technology companies committed an estimated $700 billion to data center construction. Microsoft, Google, Amazon, and Meta are building facilities at a pace not seen since the fiber-optic boom of the late 1990s. Goldman Sachs called it "the largest infrastructure buildout since the industrial revolution." The Yale School of Management asked a more uncomfortable question: "Is this a bubble?"

The citizen created, Jerome County Energy Ordinance Reform Package doesn't claim to know the answer. What it does claim — with substantial evidence — is that the data center boom has all six structural characteristics of a speculative bubble, and that communities like Jerome County should be planning for the possibility that the music stops.

The Six Bubble Indicators

None of these indicators, taken alone, proves a bubble. Taken together, they describe a market environment where the downside risks are enormous and entirely borne by local communities — while the upside benefits flow to shareholders in San Francisco and Seattle.

Lessons from the Dot-Com Bust

If this comparison seems alarmist, consider what happened the last time the technology industry built massive infrastructure on speculative demand. In the late 1990s, telecommunications companies laid millions of miles of fiber-optic cable based on projections that internet traffic would grow exponentially forever. When the dot-com bubble burst, an estimated 95% of that fiber went dark. Companies like WorldCom and Global Crossing went bankrupt. The communities that had rezoned land, granted tax breaks, and built out infrastructure to attract these companies were left with nothing — empty buildings, broken promises, and no tax revenue to show for it.

The data center boom has an eerie parallel. The AI models driving current demand are extraordinarily expensive to train and operate. The revenue models are unproven. The efficiency gains in AI hardware are moving faster than the build cycle — meaning facilities being designed today may be partially obsolete before they’re completed. Goldman Sachs’s own research arm has published analyses questioning whether AI will generate enough revenue to justify the infrastructure investment.

The Moratorium Wave

Jerome County isn’t the only place asking these questions. Across the country, communities are hitting the pause button. Kootenai County imposed a 182-day moratorium in 2025 and is now considering a total ban. Jerome Township in Ohio — the ironic namesake — imposed a nine-month pause on new data center development in September 2025. Multiple councils across Georgia, Illinois, Kentucky, Ohio, Tennessee, Missouri, and Pennsylvania have introduced moratoriums or denied ordinance changes that would permit data center construction.

In Idaho specifically, the legislature introduced HB 895 to restrict data center water use and a bipartisan bill to limit the tax exemptions that data centers receive. The politics are shifting fast — and they’re shifting because communities are realizing that the “economic development” pitch doesn’t hold up when you look at the actual cost-benefit equation.

Why Decommissioning Bonds Are Non-Negotiable

The reform package’s answer to the bubble risk is simple and proven: require decommissioning bonds set at 125% of estimated restoration costs before any permit is issued. This isn’t a radical idea. It’s standard practice in the mining industry, the oil and gas industry, and nuclear energy. The principle is straightforward — if you’re going to alter land, you should guarantee you can put it back.

In the community survey, 90.8% of respondents rated decommissioning financial assurance as “Extremely Important.” That near-unanimity tells you something important: people in Jerome County understand risk. They’ve seen what happens when companies make promises without collateral. And they’re not willing to bet their county’s future on the assumption that the AI boom will last forever.

The bottom line: The data center boom may be transformative. It may also be a bubble. The reform package doesn't ask Jerome County to bet on either outcome. It asks the county to require that anyone who wants to build on our land guarantees they'll clean up after themselves if it doesn't work out. That's not anti-business. That's common sense.

Send this to your commissioners and legislators, let them know what you think!

Share

Subscribe now

The debate over hyperscale data center construction in the United States has largely been framed as an economic competition — states racing to offer the largest tax breaks, the cheapest power, and the most permissive land use approvals. That framing is wrong. It skips the only question that actually matters: what is the facility being built to do?

A hospital and a slaughterhouse are both large buildings. We don’t permit them the same way. We shouldn’t permit data centers the same way either.

The Current Framework Is Purposely Blind

Existing federal, state, and local regulatory frameworks treat hyperscale data centers as a class of industrial infrastructure — roughly equivalent to a warehouse or a manufacturing facility. Zoning approvals, utility interconnection agreements, tax incentive programs, and infrastructure grants almost never distinguish between a facility hosting domestic financial services and one hosting surveillance infrastructure for a foreign-linked corporation.

This isn’t an oversight. It’s the product of aggressive lobbying by an industry that benefits enormously from keeping regulators focused on square footage, employment projections, and capital investment rather than what the servers inside actually do.

The result is that American land, American water, American power grid capacity, and American taxpayer dollars are being used to build infrastructure whose purpose, ownership, and ultimate beneficiary are often deliberately obscured.

That has to end.

A Purpose-Based Framework for Construction Authorization

Congress and state legislatures should establish a mandatory purpose-disclosure and authorization framework that conditions construction approval, utility interconnection, and any form of public benefit on what a hyperscale data center is actually built to do. The framework should operate on a tiered basis:

Tier 1 — Presumptively Authorized

Facilities whose primary purpose is:

• Domestic commercial computing, cloud services, or enterprise IT

• Healthcare data processing and storage

• Financial services infrastructure

• Academic research and national scientific computing

These facilities may proceed through standard permitting with disclosure requirements and periodic auditing.

Tier 2 — Conditional Authorization Required

Facilities whose purpose includes:

• Media distribution and content hosting (social platforms, streaming)

• AI model training and inference at scale

• Telecommunications infrastructure

These require a disclosed use-case review, beneficial ownership transparency, and binding operational covenants before construction authorization.

Tier 3 — Presumptively Prohibited

Facilities whose primary or substantial purpose includes any of the following should not be authorized, full stop.

Three Categories That Should Disqualify a Project

1. Mass Surveillance Infrastructure

Hyperscale data centers purpose-built to aggregate, process, or store population-scale behavioral data — location tracking, biometric databases, communications metadata, financial transaction monitoring — represent a qualitatively different threat than ordinary commercial computing.

The distinction matters because scale is the product. A surveillance system that can process 10 million records is not merely larger than one that processes 10,000 — it is a different instrument entirely, one whose primary utility is coercive.

Domestic construction of such facilities should require a demonstrated lawful purpose, a constitutional nexus, and ongoing independent audit. Facilities that cannot satisfy that standard should not be built. This applies to private commercial surveillance operations — data brokers, behavioral advertising infrastructure, consumer profiling platforms — as much as it does to government-adjacent systems.

The legal foundation exists. The Supreme Court’s third-party doctrine has been substantially narrowed by Carpenter v. United States (2018). Congress has not caught up. A purpose-based construction authorization framework would be a meaningful step toward structural prevention rather than reactive litigation.

2. Facilities Primarily Supporting Pornography Distribution

This is the category most likely to generate reflexive first-amendment deflection, so it is worth being precise about what is and is not being proposed.

The proposal is not censorship of content. It is a refusal to extend public subsidy and public infrastructure to the industrial-scale distribution of pornography.

The distinction is critical. A publisher has a constitutional right to distribute material. That right does not include a constitutional entitlement to:

• Tax increment financing

• Below-market utility rates negotiated through public utility commissions

• State and local economic development grants

• Infrastructure upgrades to roads, power, and water systems funded by taxpayers

No serious legal scholar argues that the First Amendment compels affirmative government subsidy of speech. The government is not required to build the printing press.

Hyperscale data centers hosting pornography as a primary or substantial use case should receive no public benefit of any kind — no tax abatement, no infrastructure grant, no utility preference. They may operate under existing law. They may not do so on the public’s dime.

This position is consistent with longstanding obscenity and community standards jurisprudence and does not require Congress to resolve any contested constitutional questions. It requires only that subsidy programs specify eligible use cases — a routine exercise of legislative appropriations authority.

3. Facilities with Material Foreign Government or Corporate Access

This is arguably the highest-stakes category and the least adequately addressed by current law.

CFIUS reviews foreign acquisitions of existing domestic infrastructure. It does not adequately address greenfield construction of new infrastructure by entities with material foreign government ties, nor does it address operational relationships that give foreign governments effective access to data processed in nominally domestic facilities.

A purpose-based construction framework should prohibit authorization for any hyperscale facility where:

• A foreign government holds direct or indirect ownership of 5% or more

• Operational agreements, data processing contracts, or equipment supply chains create meaningful access rights for entities subject to foreign government legal compulsion (China’s National Intelligence Law being the most obvious example)

• The facility’s primary customer base includes foreign state-owned enterprises or government agencies

This is not a trade restriction — it is a sovereignty position. The People’s Republic of China does not permit American companies to build unrestricted data infrastructure on Chinese soil. Reciprocity is not protectionism.

No Public Subsidy, No Public Infrastructure, No Exceptions

The subsidy question deserves its own treatment because the numbers have become absurd.

State and local governments across the country have collectively offered tens of billions of dollars in tax abatements, infrastructure grants, utility rate preferences, and land deals to hyperscale data center operators. In many cases these concessions persist for decades. In virtually no cases are they conditioned on the nature of what the facility does.

The public policy rationale offered is jobs and tax base. The reality is that hyperscale data centers are among the most capital-intensive and least labor-intensive industrial facilities ever constructed. A one-million-square-foot hyperscale facility may employ fewer than 50 permanent workers. The jobs argument is almost always fictitious at the scale being claimed and it is debatable whether any subsidies ever offer an equal return, regardless of the industry.

Meanwhile the infrastructure costs are real and are borne by everyone else:

• Power: Hyperscale facilities are consuming grid capacity at a rate that is materially affecting electricity prices and reliability for residential and small commercial customers in affected markets. That is a direct wealth transfer from the general public to data center operators.

• Water: Cooling water consumption at hyperscale facilities is substantial and in arid western states represents a genuine resource conflict with agriculture and municipal supply.

• Roads: Construction traffic and ongoing logistics operations impose road maintenance costs that are rarely internalized by project developers.

• Transmission: Grid interconnection upgrades required by large loads are frequently socialized across ratepayers rather than charged to the benefiting customer.

The correct policy position is simple: no public subsidy of any kind for hyperscale data center construction, operation, or associated infrastructure. Not tax abatements. Not TIF districts. Not utility rate preferences. Not infrastructure grants. Not state economic development funds.

If the economic case for a facility requires public subsidy to pencil out, that is the market telling you the facility should not be built where it is being proposed. Listen to the market.

The Counterarguments and Why They Fail

“This will drive investment overseas.” Investment in surveillance infrastructure and foreign-accessible data centers going overseas is not a loss — it is the goal. We should not be competing to host it.

“You can’t define purpose with enough precision to regulate it.” We regulate purpose in land use law constantly. We distinguish industrial from commercial from residential uses. We distinguish uses within those categories. Defining permitted and prohibited data center purposes is not categorically more difficult than any other conditional use permitting exercise.

“Foreign investment creates jobs.” See above. Hyperscale data centers do not create meaningful employment relative to their capital base, public subsidy received, or infrastructure impact. The jobs argument is a lobby talking point, not an economic analysis.

What States Can Do Now

Federal action is preferable for foreign ownership restrictions, but states can move immediately on subsidy and land use questions:

1. Require purpose disclosure as a condition of any economic development incentive application for data center projects above a defined capacity threshold

2. Define ineligible uses in economic development statutes — surveillance aggregation, pornography distribution, and foreign-government-accessible facilities explicitly excluded from incentive eligibility

3. Require beneficial ownership transparency to the same standard as CFIUS for any facility receiving public benefit

4. Mandate utility cost internalization — require that any load above a defined threshold bear the full cost of grid interconnection and transmission upgrades without socialization to other ratepayers

5. Establish a community impact review process for hyperscale facilities that explicitly considers purpose, water consumption, road impact, and grid effects on existing residential and agricultural customers

Idaho has the additional leverage of water law. Consumptive water use rights for industrial facilities are subject to state appropriation review. Purpose should be a factor in that review.

Conclusion

The hyperscale data center industry has successfully convinced most state and local governments that the question of what a facility does is irrelevant — that a data center is a data center, and the only variables worth discussing are jobs and tax revenue.

That argument benefits exactly one party: the industry making it.

The rest of us — the ratepayers subsidizing grid upgrades, the farmers competing for water, the communities absorbing infrastructure costs, the citizens whose data is being processed in facilities we helped build — have a legitimate interest in what these facilities are for.

Purpose-based construction authorization is not a radical idea. It is the application of a principle we already accept in every other domain of land use and industrial permitting to a sector that has so far successfully evaded it.

The evasion should end.

Share

Jeff A. Pierson is the owner of Confidential Solutions LLC and a policy researcher. He writes on Idaho energy, land use, and technology policy at jeffapierson.substack.com.

Subscribe now

The New Layer Between Humans and Computers: How Windows 11 Signals the AI Operating System Era

For most of the history of personal computing, the relationship between humans and computers has been direct. A person issued commands through a keyboard, mouse, or touchscreen. The operating system executed those commands and passed them to applications. The computer was a tool that reacted to explicit instructions.

That architecture is beginning to change.

Windows 11 represents a transitional moment where an artificial intelligence layer is increasingly placed between the human user and the operating system itself. Instead of users directly manipulating the OS through menus, commands, or scripts, an AI intermediary interprets intent, observes activity, and performs actions on behalf of the user. This shift is not simply a software feature. It is an architectural change in how computing systems are designed. Understanding this shift explains why the AI industry is experiencing such explosive growth.

The Traditional Personal Computer Model

For decades the architecture of a personal computer followed a simple chain.

A user opened a program. The program interacted with the operating system. The operating system controlled the hardware.

Even when automation increased in the Windows 10 era through cloud services and telemetry, the structure remained largely the same. Software still required explicit commands. The system did not interpret human behavior in a continuous way.

The Emerging AI Mediation Layer

Windows 11 introduces the early form of a new model.

In this structure the AI layer sits between the user and the operating system. It observes behavior, interprets intent, and then issues commands to the system. Instead of telling the computer exactly what to do, users increasingly describe goals.

Examples include asking the system to summarize documents, locate previously viewed information, automate system configuration, or perform multi-step actions through natural language. The operating system no longer acts solely as a passive executor of commands. It becomes part of an intelligent system that interprets context.

Continuous Observation as a Feature

For AI systems to work effectively they must observe user behavior. This requires a computer to monitor activity such as:

• Application usage

• Documents viewed

• Web pages accessed

• On-screen content

• User workflows

These observations are processed by machine learning models that create searchable context and predictive assistance. The controversial element of this design is that monitoring is not incidental. It is fundamental to the architecture.

Without continuous observation, the AI layer cannot provide contextual assistance. This architectural reality is one reason AI computing requires far more processing power than traditional software systems.

The Hardware Shift: Neural Processing Units

The AI mediation model also requires new hardware. Traditional CPUs are optimized for sequential logic operations. GPUs accelerate graphics and parallel workloads.

AI systems rely heavily on matrix math and neural network inference. To support this workload efficiently, new computers increasingly include neural processing units, or NPUs. NPUs allow machine learning models to run locally on a device instead of constantly relying on remote cloud servers.

This shift allows an operating system to continuously analyze user behavior while maintaining reasonable performance and reducing network dependence. The result is a new class of personal computers designed specifically for AI workloads.

Why This Architecture Is Fueling the AI Boom

The transition to AI-mediated operating systems has enormous implications for the technology industry. Every layer of computing is affected. AI models must be developed and trained. Specialized hardware must be manufactured. Cloud infrastructure must scale to support model training. Software ecosystems must adapt to AI-driven interfaces.

If the operating system becomes an AI platform rather than just a software foundation, then AI is no longer a niche tool. It becomes the central mechanism through which computing works. This reality is driving massive investment in:

• AI semiconductor design

• Large language models

• Cloud compute infrastructure

• Enterprise AI software

• AI-optimized operating systems

The AI boom is not only about chatbots or generative art. It is about a fundamental change in the architecture of computing itself.

The Long-Term Implications

If the AI layer continues to grow in capability, future operating systems may behave less like tools and more like assistants. Computers will increasingly anticipate actions, manage workflows, and organize information automatically.

The role of the user shifts from operator to supervisor. This raises legitimate questions about privacy, control, and system transparency. A system that continuously observes behavior to provide assistance also possesses unprecedented visibility into user activity. At the same time, the productivity potential is enormous. Complex multi-step tasks could be reduced to simple conversational requests.

A New Computing Paradigm

The rise of AI in operating systems marks the beginning of a new computing paradigm. The personal computer is evolving from a command-driven machine into an intent-driven system. Instead of telling the computer exactly what to do, users increasingly describe what they want to accomplish. The system determines how to do it.

Windows 11 represents one of the earliest mainstream examples of this transformation. As operating systems continue to integrate AI more deeply, the boundary between user, software, and machine intelligence will continue to blur.

This architectural shift, more than any single product, is what is fueling the global surge in artificial intelligence development.

Now ask yourself…

Share

Related:

AI Privacy Risks

Stanford Human-Centered AI Institute
Protecting Personal Data in the AI Era
https://hai.stanford.edu/news/privacy-ai-era-how-do-we-protect-our-personal-information

IBM Research Brief
AI and Privacy Concerns
https://www.ibm.com/think/insights/ai-privacy

National Institutes of Health (NIH) / PubMed
Ethical Issues in Artificial Intelligence and Privacy
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9044077

ScienceDirect
Artificial Intelligence and Data Privacy Risks
https://www.sciencedirect.com/science/article/pii/S2352250X24000423

Electronic Frontier Foundation
Artificial Intelligence and Privacy
https://www.eff.org/issues/ai

AI Surveillance and Monitoring

Brookings Institution
How Artificial Intelligence Enables Public Surveillance
https://www.brookings.edu/articles/how-ai-can-enable-public-surveillance

ScienceDirect
Artificial Intelligence–Powered Surveillance Systems
https://www.sciencedirect.com/science/article/pii/S0160791X22002780

arXiv Research Paper
AI-Enabled Smart Video Surveillance Systems
https://arxiv.org/abs/2302.04310

Carnegie Endowment for International Peace
The Global Expansion of AI Surveillance
https://carnegieendowment.org/research/2019/09/the-global-expansion-of-ai-surveillance

American Civil Liberties Union (ACLU)
Facial Recognition Technology and Civil Liberties
https://www.aclu.org/issues/privacy-technology/surveillance-technologies/face-recognition-technology

AI-Generated Pornography and Deepfakes

University of Pennsylvania Annenberg School
What Deepfake Porn Is and Why It Is Growing
https://www.asc.upenn.edu/news-events/news/what-deepfake-porn-and-why-it-thriving-age-ai

Columbia University Institute for Global Politics
The Rise of Deepfake Pornography
https://igp.sipa.columbia.edu/news/rise-deepfake-pornography

Wired Magazine
Deepfake Porn Is Out of Control
https://www.wired.com/story/deepfake-porn-is-out-of-control

Euronews Technology
Generative AI Fueling the Spread of Deepfake Pornography
https://www.euronews.com/next/2023/10/20/generative-ai-fueling-spread-of-deepfake-pornography-across-the-internet

Utah Legislature Technology Policy Brief
Deepfakes, AI, and Intimate Images
https://le.utah.gov/interim/2025/pdf/Deepfakes_AI_and_Intimate_Images.pdf

Government and Policy Reports on AI Abuse

U.S. Department of Homeland Security
Increasing Threats of Deepfake Identities
https://www.dhs.gov/sites/default/files/publications/increasing_threats_of_deepfake_identities_0.pdf

European Commission
Artificial Intelligence Act and Privacy Concerns
https://digital-strategy.ec.europa.eu/en/policies/european-approach-artificial-intelligence

UK Parliament Research Briefing
AI, Deepfakes, and Digital Harm
https://researchbriefings.files.parliament.uk/documents/POST-PN-0697/POST-PN-0697.pdf

OECD
Artificial Intelligence, Privacy, and Data Governance
https://www.oecd.org/digital/artificial-intelligence

Key Statistics Often Cited in Policy Discussions

University of Amsterdam / Deeptrace Labs Research
https://sensity.ai/reports/deepfake-report

University of Pennsylvania analysis of deepfake content
https://www.asc.upenn.edu/news-events/news/what-deepfake-porn-and-why-it-thriving-age-ai

Subscribe now

Subscribe now

A New Concept in Gaming (The Public)

Boom and Bust is a strategic infrastructure economics board game that simulates the political, financial, and regulatory forces driving large-scale data centers, renewable energy buildouts, transmission expansion, and public subsidy regimes. Designed as both a competitive strategy game and an economic simulation, it exposes how incentives, leverage, zoning authority, and subsidy stacking can generate rapid growth followed by systemic instability.

The game models capital flows, regulatory discretion, ratepayer risk, community backlash, grid stress, water constraints, and the moral hazard embedded in public-private infrastructure partnerships. Players assume asymmetric roles within the system, each pursuing distinct objectives under shifting macroeconomic conditions. Boom phases reward expansion and leverage. Bust phases punish overexposure, regulatory shortcuts, and unsustainable debt structures.

Boom and Bust functions as a playable economic model. It is suitable for tabletop deployment, educational use, civic dialogue, and scalable digital simulation.

Overview

Boom and Bust combines the accessibility of a recognizable board layout with the depth of a modern asymmetric strategy game. It sits at the intersection of Monopoly-style movement mechanics, negotiation-driven political games, and macroeconomic simulation.

Target audiences include strategy gamers, policy enthusiasts, energy market professionals, law and public administration programs, and players interested in real-world systems modeling. The theme is contemporary and highly relevant: artificial intelligence-driven data center expansion, renewable energy mandates, transmission congestion, subsidy stacking, and municipal decision-making under capital pressure.

Replayability is driven by shifting macro conditions, event decks, asymmetric incentives, and dynamic system metrics. No two games unfold the same way. One session may reward aggressive leverage. Another may punish it abruptly through interest rate shocks or community backlash.

The design is expandable. Future modules can incorporate drought cycles, federal stimulus packages, AI demand spikes, nuclear additions, or supply chain disruptions without altering the base architecture. The system is modular, extensible, and digitally adaptable.

BOOM AND BUST

Infrastructure, Influence, and the Collapse of Overreach

I. Game concept and design philosophy

Boom and Bust is a strategic political economy board game modeling infrastructure expansion driven by capital markets, public subsidies, regulatory discretion, and technological hype. It simulates data center buildouts, renewable energy overexpansion, transmission bottlenecks, debt structuring, regulatory capture, zoning leverage, and the destabilization that can follow unrestrained growth.

The core thesis is incentive driven. When capital is abundant and regulatory frameworks are permissive, growth accelerates beyond real demand and local capacity. When the cycle reverses, costs are socialized, assets are stranded, and instability radiates outward.

Players do not merely accumulate property. They interact across asymmetric power structures. Each role represents an incentive position within a system. The game models moral hazard, capture risk, public backlash dynamics, and capital flight.

The tone is serious with restrained satire. The simulation is designed to feel realistic and occasionally uncomfortable.

II. Player roles

Each role has asymmetric powers and a distinct victory condition.

Hedge Fund / Private Equity

• Starting capital: 500 units.

• Power: Leverage multiplier up to 3x per project through bond issuance.

• Special ability: Early Exit. Liquidate one sector during Boom Phase at +20 percent valuation.

• Risk: Bankruptcy if leverage exceeds 300 percent when Bust triggers.

• Win condition: Exit before Bust with the highest retained capital.

Data Center Developer

• Starting capital: 300 units.

• Power: Demand Projection tokens increase regional capacity price by 10 percent per token.

• Special ability: Anchor Tenant card once per game guarantees a 5-turn fixed PPA.

• Risk: Water shortfall or transmission denial reduces revenue by 50 percent.

• Win condition: Deploy maximum operating capacity at time of Bust without insolvency.

Renewable Energy Developer

• Starting capital: 250 units.

• Power: Stack up to three subsidy certificates per project.

• Special ability: Production Tax Credit doubles early return during Boom Phase.

• Risk: Curtailment reduces effective revenue.

• Win condition: Highest installed capacity before Grid Stability collapses.

Regulated Utility

• Starting capital: 200 units.

• Power: Rate adjustment authority based on Ratepayer Burden Index.

• Special ability: Cost Recovery Rider shifts 50 percent of cost overruns to ratepayers.

• Risk: Political backlash if Ratepayer Index exceeds threshold 80.

• Win condition: Maintain Grid Stability above 60 and avoid Public Backlash above 70.

County Commissioner / Zoning Authority

Starts with 5 Political Influence tokens.

• Power: Approve or deny zoning.

• Special ability: Fast Track reduces project delay by 2 turns at cost of 10 Public Trust.

• Risk: Ethics Investigation event.

• Win condition: Maximize tax base while keeping Public Backlash under 60.

Ratepayer Advocate

Starts with 5 Legal Challenge cards.

• Power: Trigger regulatory review.

• Special ability: File injunction once per game to freeze a sector.

• Win condition: Keep Ratepayer Burden below 70 and prevent Utility bankruptcy.

Transmission Operator

• Starts with 100 capital and 3 Corridor rights.

• Power: Expand transmission lines at 80 cost each.

• Revenue tied to congestion index.

• Risk: Overbuild leads to stranded corridor penalty.

• Win condition: Maintain congestion fee income without triggering overcapacity collapse.

III. Game board design

The board is circular like Monopoly but sector based.

Sectors include:

Rural Farmland. Industrial Zone. Suburban Expansion Edge. Aquifer Sensitive Basin. Enterprise Tax District. Decommissioned Coal Plant Site. Urban Core. Transmission Corridor. State Regulatory Commission. Federal Subsidy Office.

Tracks along the outer perimeter:

• Boom Index 0–100.

• Grid Stability 0–100.

• Ratepayer Burden 0–100.

• Public Backlash 0–100.

• Interest Rate Track 1–10 percent.

Bust triggers when Boom Index exceeds 85 and one secondary condition exceeds its threshold.

IV. Core economic mechanics

Turn structure

1. Capital Allocation.

2. Political Negotiation.

3. Zoning / Regulatory Decision.

4. Build Phase.

5. Revenue Collection.

6. Stability Adjustment.

Project cost formula

Base project cost.

• Data Center = 120.

• Solar = 80.

• Wind = 90.

• Transmission tie-in = 40.

• Water rights = 20.

ROI formula per turn

ROI = (Demand Index × Capacity Deployed − Curtailment Penalty − Debt Service).

Debt Service = Principal × Interest Rate.

Leverage rule

Debt allowed up to 3× equity.

If Interest Rate Track reaches 7 or above, leveraged projects suffer -20 percent yield.

Subsidy stacking

Each subsidy certificate reduces effective cost by 15 percent but increases Boom Index by +3.

Curtailment

If installed renewable capacity exceeds grid capacity by 20 percent, renewable income reduces by 30 percent.

V. Boom and bust cycle mechanic

Boom increases when:

• Subsidies issued: +2 each.

• New data center: +3.

• Transmission expansion: +2.

• Debt issuance over 200 percent leverage: +5.

Bust triggers if:

• Boom Index ≥ 85 and at least one of the following is true.

• Interest Rate ≥ 7.

• Public Backlash ≥ 70.

• Grid Stability ≤ 50.

Bust consequences

• Asset values reduced by 40 percent.

• Leverage penalties double.

• Ratepayer Burden +15.

• Public Backlash +10.

• Transmission overcapacity forces corridor devaluation.

• Losses are applied immediately.

VI. Game pieces

Currencies

• Capital tokens.

• Political Influence chips.

• Public Trust markers.

• Grid Capacity cubes.

• Water allocation markers.

Project pieces

• Data center complexes.

• Solar fields.

• Wind farms.

• Transmission lines.

• Substations.

Certificates

• Tax credit cards.

• Bond certificates.

• Rate adjustment permits.

• Environmental challenge cards.

Event decks

• Federal Policy deck.

• Interest Rate deck.

• Community Referendum deck.

• Litigation deck.

VII. Negotiation and capture mechanics

Lobbying

Players may transfer capital to the County Commissioner for Influence tokens at 2 capital per token. The Commissioner gains a personal Influence pool. Public Trust decreases by 3 per undisclosed transfer.

Revolving door

At any time, the Commissioner may convert 3 Influence tokens into a board seat at the Utility, gaining revenue share. Public Backlash increases by 10.

Public records request

If drawn, forces disclosure of one hidden negotiation.

Ethics investigation

Triggers if Influence exceeds 15. Penalty includes Public Trust collapse.

VIII. Win conditions

• Each player wins individually based on their defined metric.

• Optional cooperative mode allows players to collectively attempt to avoid Bust for 15 turns.

• In adversarial mode, Bust is expected and central to play.

IX. Educational application

Designed to teach:

Moral hazard. Regulatory capture. Concentrated benefit versus dispersed cost. Boom-bust cycles. Local government constraints. Debt-driven expansion economics. Infrastructure externalities.

A facilitator can freeze the game at Bust and conduct a debrief.

X. Board version versus online version

Board version uses fixed Boom multipliers and card decks.

Online version introduces dynamic demand simulation, real-time interest rate modeling, an AI regulator, variable consumer demand, map-based regional water modeling, and multi-county expansion.

Expansions

• AI Demand Surge expansion.

• Federal Green Stimulus expansion.

• Supply Chain Crisis expansion.

• Small Modular Nuclear add-on.

• Drought and Water Crisis pack.

XI. Balance principles

No role has full control.

Developers depend on zoning. Utility depends on rate approval. Hedge funds depend on timing. Commissioners depend on trust thresholds. Transmission depends on real demand. System instability is a balancing mechanism.

XII. Ethical framework

The game explores tension between private reward and public risk, short-term growth and long-term stability, subsidy dependency and market discipline, local land use and national demand, and capital efficiency and community cost.

XIII. Art and visual style

Industrial utility map aesthetic. Subdued palette. Grid overlays and zoning demarcations. No cartoon caricatures. Subtle realism.

XIV. Prototype rulebook outline

Setup

• Each player selects a role.

• Place Boom Index at 20.

• Interest Rate at 3 percent.

• Grid Stability at 90.

• Ratepayer Burden at 20.

• Public Backlash at 10.

Round flow

1. Roll movement die.

2. Resolve the space landed.

3. Enter the negotiation window.

4. Build or expand if approved.

5. Collect revenue.

6. Adjust system metrics.

Example sequence

Turn 1: Renewable Developer builds Solar in Rural Farmland and stacks 2 subsidy credits. Cost reduced from 80 to 56. Boom Index increases from 20 to 26. Grid Stability drops from 90 to 85 due to intermittency.

Turn 3: Hedge Fund finances 2 data centers at 3× leverage. Boom Index jumps to 60. An Interest Rate card increases the track to 6 percent.

Turn 5: Public Backlash reaches 70 after a water shortfall event. Bust triggers. Asset values collapse. The Hedge Fund goes bankrupt. The Utility increases rates. The Ratepayer Advocate wins under its metric.

Share

BOOM AND BUST™

Copyright © 2026 Jeff A. Pierson. All Rights Reserved.

Subscribe now

“Tiers and Beyond: Understanding the Multiple Ways to Classify Modern Data Centers” explains that modern data centers can no longer be understood by a single rating system. The Uptime Institute’s Tier I through Tier IV classifications still matter, but they now sit alongside multiple other frameworks that describe efficiency, sustainability, reliability, and operational maturity. The article stresses that data centers are too complex and too varied for one label to capture their full impact or performance.

The Uptime Tiers remain the industry’s most recognized system. Tier I describes basic infrastructure with minimal redundancy. Tier IV describes a fault tolerant facility designed to stay online even in the face of major failures. The article notes that companies increasingly combine this Tier system with other frameworks such as the EN 50600 standard in Europe, the ISO 30134 PUE energy efficiency metric, and sustainability evaluations tied to power sources and carbon reduction goals.

The main point is that classification is now multi dimensional. Operators, investors, local governments, and communities should not look at any single rating and assume they know a facility’s true demands on land, water, energy, and local infrastructure. The article highlights that each system measures different things and that modern data centers are being built so quickly that classification alone does not fully reveal their real world impacts.

How This Applies to Idaho

Idaho is now being targeted for hyperscale and large scale data center expansion. The central warning of the article applies directly to counties evaluating permits. Local officials regularly rely on simple labels such as “Tier III facility” or “high efficiency data center.” The article makes it clear that these descriptions are not sufficient for understanding the actual load placed on Idaho’s aquifer, electricity, transmission networks, and land use systems.

Idaho’s counties often lack the technical staff to evaluate multi layered data center classifications. The result is a real risk of approving facilities whose operational demands far exceed what the label suggests. A Tier III or Tier IV data center can still require millions of gallons of cooling water, dozens of megawatts of continuous electricity, and round the clock deliveries of equipment and fuel. The Uptime Tier alone does not inform any county about water requirements, energy consumption, grid impacts, cooling method, wastewater discharge, or long term environmental risk.

Idaho must also consider that investors use these classifications as part of their marketing. A facility advertised as “high tier” or “efficient” may still impose burdens that local infrastructure cannot support. This matters in places like Jerome County, Gooding, Ada, and the Magic Valley where water supplies, substation capacity, and broadband backhaul are finite. Idaho Power’s generation and transmission constraints mean that the true rating of a data center does not guarantee that the grid can sustain it.

The central lesson is simple. Idaho cannot rely on tier labels, efficiency acronyms, or promotional classifications. Counties must demand full disclosure of electric load, water demand, cooling method, wastewater discharge, backup generator cycles, broadband transport requirements, and lifecycle expansion plans. The article demonstrates that the industry has moved beyond single ratings. Idaho must do the same or risk approving projects whose infrastructure demands exceed what communities can absorb.

Share

Subscribe now

Hyperscale data centers are the largest, most power-hungry, and most resource-intensive digital facilities in the country. They are built by tech giants to run cloud platforms, AI systems, and massive online services. Their impact on American communities is direct, long-lasting, and often underestimated.

Understanding that impact starts with understanding how many of these facilities exist today and how many more are planned.

How Many Hyperscale Data Centers Are in the United States

As of 2025, industry reports identify approximately 152 hyperscale data centers operating inside the United States. These are the mega-facilities owned by companies such as Amazon, Google, Microsoft, Meta, Oracle, and Apple. They represent the largest share of hyperscale installations in any single country.

How Many Are Planned or Under Development

Across the United States, an estimated 30 to 50 additional hyperscale data centers are in various stages of planning, permitting, or early construction.

This number is based on:

• announced hyperscale campuses

• multi-building expansions inside existing campuses

• land acquisition filings

• utility-upgrade applications tied to hyperscale customers

• public development disclosures from hyperscale operators

Because hyperscale companies often build in phases, a single new “campus” may contain four to ten buildings. That means the true long-term count will be much higher than the number of initial filings suggests.

The Impact: What Hyperscale Development Means for Communities

Hyperscale data centers do not behave like typical industrial buildings. Their impact on local resources is outsized and continuous.

1. Electricity Demand at City-Scale

A hyperscale data center can draw hundreds of megawatts of electricity.
For comparison, a single large hyperscale campus can use more power than:

• an airport

• a steel mill

• or an entire mid-sized American city

This demand requires new substations, new transmission lines, and expensive upgrades. Utilities recover those costs from ratepayers. Communities frequently see electric rate increases tied directly to data-center expansion.

2. Large-Scale Water Use

Most hyperscale facilities rely on evaporative cooling or hybrid cooling systems that require steady access to water.

Annual consumption for one campus can reach hundreds of millions of gallons.
In regions with tight water budgets, this becomes a direct competition with:

• agriculture

• municipal drinking-water systems

• groundwater protection

• long-term aquifer stability

Communities often do not fully understand the water implications until after approval.

3. Land Conversion and Permanent Change

A hyperscale campus typically requires dozens to hundreds of acres.

Once approved, the land is permanently converted into:

• server buildings

• substations

• cooling systems

• fiber corridors

• security perimeters

• multi-phase expansion pads

Farmland, open space, and rural landscapes are replaced by industrial complexes that operate 24 hours a day.

4. Heavy Impact on Roads and Infrastructure

Construction lasts for years.
Trucks, concrete mixers, fiber trenching, power-line construction, and water-line installation all place significant strain on local roads.

Rural counties often lack the staff, the budget, or the engineering capacity to manage the scale of these projects.

5. No Guaranteed Local Benefit

Most hyperscale centers create:

• minimal permanent jobs

• private, not public, fiber

• very limited local access to power improvements

• no direct service benefit for local residents

The benefits accrue to the hyperscale operator.
The burdens — water, power, land, and rising utility rates — fall on the community.

Why These Numbers Matter

With 152 hyperscale data centers already operating in the United States and 30–50 more under active planning, the trend is clear.

Hyperscale growth is accelerating, and it is moving into smaller, cheaper, rural counties — exactly the kind of communities least able to absorb the infrastructure strain.

Approving one hyperscale campus is not approving one building.
It is approving:

• a long-term resource commitment

• a permanent industrial footprint

• and a rising share of local power and water obligations

Communities have every reason to review these projects with full information and full transparency.

Resources

• Data center growth drives locals to fight for more say (Nov 2025)

• Global data center expansion and human health: A call for empirical research (May 2025)

• What is a hyperscale data center? | IBM

• The future of US hyperscale data centers | McKinsey

• Mass. prepares for first hyperscale data center

• What is a Hyperscale Data Center and How Does It Work?

Share

Article Link: https://www.npr.org/2025/11/23/nx-s1-5615410/ai-bubble-nvidia-openai-revenue-bust-data-centers

Subscribe now

Review of the NPR Article on the AI Bubble

The article lays out a simple but uncomfortable truth. The entire AI infrastructure boom rests on a fragile assumption that demand for AI services will grow forever and at an enormous scale. That assumption is underwriting trillions of dollars in activity, from GPU production to data center construction to massive energy and land commitments. The problem is obvious. If demand does not keep pace with these expectations, the industry will be left with staggering levels of debt, unused capacity, and stranded assets with almost no alternative use.

The article hints at this, but the deeper problem is the kind of financing behind the boom. Much of the money is coming from private credit structures and special-purpose vehicles. These are lightly regulated, opaque, and increasingly leveraged. This behavior is not innovation. It is financial engineering designed to stretch risk across a large number of investors who do not fully understand what they are exposed to. The American Prospect has already warned that this could mirror patterns seen in past bubbles, where the absence of oversight magnified the damage once expectations snapped.

The risk is not limited to AI startups. The fallout would spill into large technology companies, the banking sector, private credit funds, pension systems, and any investor who bought into the hype without understanding the weak economic foundation underneath it. Today’s excitement over “AI infrastructure” could easily become tomorrow’s financial wreckage. History has seen this before in the dot-com collapse and several real estate cycles. The parallels are obvious.

The lesson is not complicated. Regulators and policymakers should already be paying attention. The scale and speed of these investments do not reflect organic technological evolution. They look more like a leveraged bet on a future that may not materialize. If the assumptions crack, the losses will spread far beyond Silicon Valley. The article raises a warning. The underlying trend necessitates the warning.

Share

Subscribe now

Author’s Note

These descriptions are general and are meant to help the public understand the broad differences among the major types of data centers. Technical terms and architectural details are kept to a minimum because the goal is clarity, not engineering depth. Every project has its own requirements, infrastructure demands, and local impacts, but the categories outlined here give a reliable sense of scale. They show why some facilities blend into existing development while others overwhelm small towns with power, water, and land pressures. This guide is not a technical manual. It is a plain-language framework to help communities recognize what kind of project they are actually being asked to accept.

The Different Types of Data Centers

Explained by Ownership, Purpose, Size, and Community Impact

People hear the word “data center” and imagine a giant warehouse full of mysterious computers. The truth is simpler. There are several types of data centers, and each behaves differently depending on who owns it, why it exists, how big it is, and what it does to the community around it.

Enterprise Data Centers

• Ownership: A private company, hospital, bank, or government agency owns it outright.

• Purpose: To run its own internal systems. These are not built for outsiders or customers.

• Size: Usually mid-sized. Not small, but nowhere near the mega-facilities built for big tech.

• Community Impact: Low. These facilities are normally tucked into existing office campuses. They use steady amounts of electricity and water, but nothing like the industrial-scale AI farms. They rarely strain local infrastructure.

Colocation Centers

• Ownership: A company builds and runs the building, then rents space inside it. Businesses bring their own servers and pay for power, cooling, and security.

• Purpose: To give many companies a shared facility without each one needing to build its own.

• Size: Small to large. Depends on the market.

• Community Impact: Moderate. They pull significant power, may need upgrades to substations, and can add traffic and construction noise, but they are not usually tied to massive industrial water use.

Hyperscale Data Centers

• Ownership: Big tech companies such as Amazon, Google, Meta, and Microsoft. These firms own the buildings, the land, the servers, and everything inside.

• Purpose: To run huge cloud platforms and AI systems that serve millions of users.

• Size: Gigantic. Thousands of servers. May include multiple buildings per campus.

• Community Impact: High. These are the data centers that can reshape towns. They demand huge electrical capacity, often more than entire cities. Many require large volumes of water for cooling. They can force utility rate hikes, strain power grids, and change land values overnight. These are the ones that rural communities argue about for good reason.

Edge Data Centers

• Ownership: Telecom companies, cloud companies, or specialized operators.

• Purpose: To put small clusters of servers close to customers so websites, apps, and services load faster.

• Size: Small. Sometimes, it is only a single room or a small building.

• Community Impact: Low. They use little power and water and slip into existing commercial spaces. Most people never notice them.

Modular or Micro Data Centers

• Ownership: Varies. Anyone from a utility to a private business may deploy them.

• Purpose: To add quick computing capacity without building a full facility. Often used at industrial sites, in remote areas, or as temporary expansions.

• Size: Very small. Sometimes the size of a shipping container or a large shed.

• Community Impact: Minimal unless deployed in large numbers. They are compact, quiet, and self-contained.

What This Means for Communities

When people say “a data center is coming,” the type matters more than the label. An enterprise or edge facility barely registers in daily life. A colocation center is noticeable but manageable. A hyperscale data center changes everything: land use, roads, water systems, and especially power grids.

Ownership determines who controls it. Purpose determines why it is built. Size determines whether it fits the landscape. And all of that together determines whether the community feels the benefits or pays the price.

Resources

• Data center growth drives locals to fight for more say (Nov 2025)

• Global data center expansion and human health: A call for empirical research (May 2025)

• What is a hyperscale data center? | IBM

• The future of US hyperscale data centers | McKinsey

• Mass. prepares for first hyperscale data center

• What is a Hyperscale Data Center and How Does It Work?

Share

Article Link: Beyond the Bubble

Subscribe now

KKR argues that the AI boom is not a temporary bubble but the start of a long infrastructure cycle. They frame AI data centers, power development, and grid expansion as the modern equivalent of railroads and electrification. Their core claim is straightforward. Even if speculative companies fail, the physical infrastructure built during the frenzy will remain and will power economic growth for decades. They present the bubble as a driver of essential national buildout rather than a threat to long-term stability.

The weakness in the article is the assumption that the AI buildout behaves like past infrastructure cycles. KKR treats data centers as if they share the same public benefit and permanence as fiber, roads, or power lines. They do not. AI infrastructure consumes permanent megawatt-scale loads, requires nonstop hardware replacement, and offers no general public utility when the operator fails. If a railroad fails, the tracks remain. If a fiber company fails, the lines remain in place. If a data center fails, it becomes a stranded box, leaving the local grid with fixed costs and the public with higher rates. The article never acknowledges that AI infrastructure can drain local resources rather than add long-term public value.

KKR is correct that AI demand is real and growing. The blind spot is the belief that all infrastructure created during the boom is inherently valuable. In states like Idaho with limited grid capacity and tight water resources, that assumption becomes risky.

Share

Article Link: “AI Data Center Frenzy Is Pushing Up Your Electric Bill. Here’s Why. “

Subscribe now

Why This Matters for Idaho

The article highlights a national surge in electricity demand driven by the rapid expansion of AI and data centers. This is not a distant problem. It has direct consequences for Idaho.

Data centers already consume a significant share of the nation’s electricity, and that demand is rising fast. Idaho’s low cost, reliable power makes the state an attractive target for developers looking for cheap energy and open land. That combination puts pressure on the grid, affects water supplies, and raises long term costs for residents.

When large data centers move into a region, they do more than open buildings. They place heavy and constant demand on the electrical system. Utilities often respond by expanding transmission lines, adding substations, or upgrading generation capacity. These upgrades are expensive, and the cost is often passed on to every ratepayer, not just the corporations that created the demand. The result is higher power bills for households, farmers, and small businesses.

There is also the water problem. Modern AI data centers require enormous amounts of cooling. Many rely on continuous water use to manage the heat produced by their servers. Idaho communities depend on aquifers, irrigation systems, and carefully managed water rights. A single industrial scale data facility can stress that balance and push local systems beyond their limits.

There is another issue that deserves attention. If regulators give preferential treatment to these companies or fail to require full cost recovery, the financial burden shifts to the public. That becomes an indirect subsidy from ordinary Idahoans to some of the wealthiest corporations in the world. Other states have already learned this lesson the hard way.

What Idahoans Should Watch

Communities should insist on complete transparency from utilities, developers, and state agencies. Residents deserve to know exactly how much power a project will consume, how much water it requires, and who will pay for the infrastructure it needs.

Lawmakers should ensure that data center developers pay their fair share. Cost shifting onto households must not become part of Idaho’s economic development strategy.

Local leaders need to weigh short term tax revenue against long term pressure on the grid, rising consumer rates, and the strain on limited water resources.

The AI data center boom is not just a national trend. It is headed for Idaho because the state offers cheap power and available land. Without active oversight, the benefits flow to corporations while the costs fall on families and communities.

Share

Subscribe now

Review

Article Link: Pew Research Center https://www.pewresearch.org/short-reads/2025/10/24/what-we-know-about-energy-use-at-us-data-centers-amid-the-ai-boom

The Pew Research Center’s report on data center energy use is a blunt snapshot of an industry that has grown faster than the public understands and far faster than most states can regulate. The article compiles the most reliable national data on the number of data centers, their electricity demand, and their impact on water use and ratepayer costs. It confirms that the AI boom is not a virtual revolution. It is a physical one. It consumes land, water, and grid capacity at levels rivaling those of major industrial sectors.

Pew draws from the International Energy Agency and industry maps to outline a simple truth. The United States is now home to more than four thousand data centers. One third are concentrated in Virginia, Texas, and California. The clustering effect intensifies regional stress. In Virginia, data centers already use more than a quarter of the state’s entire electricity supply. The implications are obvious. Local grids are being re-engineered around the needs of a single industry.

Energy use is rising faster than in any other major sector. Data centers consumed about 183 terawatt-hours in 2024. Pew notes that this amounts to more than 4% of all U.S. electricity demand. That is the electricity use of a nation the size of Pakistan. Under the IEA baseline scenario, demand will exceed 426 terawatt-hours by 2030. That is a 133 percent increase in six years. This entire surge is driven by high-intensity computation, especially AI. The servers needed for AI draw two to four times the power of standard systems. Water demand rises alongside the power curve. Data centers consumed 17 billion gallons of water in 2023, and hyperscale data centers alone are projected to consume up to 33 billion gallons per year by 2028.

Pew also confirms the financial reality that local residents pay for the infrastructure these companies require. Grid upgrades are expensive. Utilities shift those costs to households. In states within the PJM grid, such as Ohio and western Maryland, residents may see monthly increases of 16 to 18 dollars next year. National projections suggest an eight percent rise in average bills by 2030. In high-density areas, the increase may exceed twenty-five percent. The average bill in the United States has already jumped to 142 dollars per month, a 25% increase over the past decade.

The article serves as a neutral report, but its implications are not neutral. The AI buildout demands fuel. More than forty percent of the electricity powering data centers comes from natural gas. Renewables account for about twenty-four percent. Nuclear accounts for twenty percent. Coal fills the remainder. Despite corporate marketing claims, the backbone of AI is still fossil energy. Pew also points to states that are debating whether to require data centers to use some level of renewable energy and to publicly report their energy and water use.

The meaning is plain. AI is not weightless. Every new model, every server farm, and every hyperscale build comes with a measurable cost borne by citizens who did not approve these projects and cannot opt out of paying for them. Local governments are already facing the same crossroads that heavy industrial regions faced in earlier eras. The choice is between planning growth with transparency and honest numbers or letting speculative demand dictate the shape of entire communities.

The Pew report does not argue policy. It presents the data. The lesson is straightforward. The AI boom is transforming the physical and economic landscape of the United States. It draws down electricity at a historic rate, consumes billions of gallons of water, strains local grids, and raises household energy bills. Any state or county that invites this industry must confront these realities in advance. Those who do not will inherit the consequences.

Share

Source: Pew Research Center
https://www.pewresearch.org/short-reads/2025/10/24/what-we-know-about-energy-use-at-us-data-centers-amid-the-ai-boom

Subscribe now

Before Continuing…

While reading the statistics below, consider how Idaho incentivizes data center construction and what the actual cost to Idaho may be.

Data Center Count and Location

• More than 4,000 U.S. data centers exist or are planned.

• Virginia, Texas, and California host about one third of all U.S. data centers.

• 643 are in Virginia, 395 in Texas, 319 in California.

• Roughly half of all new builds are in regions already saturated with data centers.

Electricity Demand

• U.S. data centers consumed 183 terawatt-hours (TWh) in 2024.

• That equals over 4 percent of all U.S. electricity use.

• Consumption equals the entire electricity use of a nation the size of Pakistan.

• Projected to reach 426 TWh by 2030, a 133 percent increase.

• In Virginia, data centers already consume 26 percent of statewide electricity.

• In ND, NE, IA, OR, data centers consume 11 to 15 percent of statewide electricity.

Internal Energy Use Breakdown

• 60 percent of electricity powers servers.

• AI-optimized servers consume 2 to 4 times the power of traditional servers.

• Cooling uses 7 percent of energy in efficient hyperscale centers.

• Cooling can exceed 30 percent in less efficient enterprise centers.

Water Consumption

• Data centers used 17 billion gallons of water in 2023.

• 84 percent of all water consumption comes from hyperscale and colocation centers.

• By 2028, hyperscale facilities alone will use 16 to 33 billion gallons annually.

Energy Sources

• Natural gas supplies 40+ percent of data center electricity.

• Renewables supply 24 percent.

• Nuclear supplies 20 percent.

• Coal supplies 15 percent.

Electric Bill Impacts

• PJM region expects $9.3 billion in increased capacity costs in 2025–26 due to data centers.

• Residential bills in that region may rise $16–18 per month.

• Nationally, data centers and crypto mining may push bills 8 percent higher by 2030.

• High-density regions could see increases of 25 percent or more.

• Average U.S. household electric bill in 2024 was $142/month, up 25 percent from 2014.

Resources

• Data center growth drives locals to fight for more say (Nov 2025)

• Global data center expansion and human health: A call for empirical research (May 2025)

• What is a hyperscale data center? | IBM

• The future of US hyperscale data centers | McKinsey

• Mass. prepares for first hyperscale data center

• What is a Hyperscale Data Center and How Does It Work?

Share

Subscribe now

What is an AI or Data Center Bubble?

An AI and data center bubble is a period of inflated investment, speculative growth, and large-scale infrastructure expansion centered around artificial intelligence and the massive computing facilities that support it. The core problem is a mismatch between the hype-fueled expectations and the actual utility or demand. Companies, governments, and investors pour billions into AI models and server farms with the belief that they are building the future. But the scale, speed, and justification for these investments are often untethered from reality.

What are the Defining Traits of a Bubble?

1. Overbuilding without demand
   Corporations and utilities are constructing enormous data centers and AI infrastructure that far exceed current public, commercial, or strategic needs. Much of it is driven by forecasts and investor pressure rather than grounded analysis.

2. Speculative capital and inflated valuations
   Venture funds and major tech firms are throwing money into AI startups, chip manufacturers, and data infrastructure projects. These investments are often based on narratives and branding rather than proven returns.

3. Resource extraction disguised as innovation
   Projects consume vast amounts of land, power, and water, often in rural or economically vulnerable areas. These centers rarely deliver promised jobs or community benefits. They function as industrial resource hogs behind the marketing of futuristic progress.

4. Techno-utopian marketing shields scrutiny
   The idea that AI will transform everything justifies nearly any scale of project. Politicians and planners are told to get out of the way because “the future is coming.” This narrative silences legitimate questions about environmental impact, civic utility, or long-term consequences.

5. Hollow applications dominate
   Much of today’s AI is being used for surveillance, advertising, automation of basic tasks, and monetizing user behavior. It does not replace critical thinking, solve existential problems, or elevate public well-being. Most consumers do not see direct benefits proportionate to the scale of infrastructure being built.

6. Risk of collapse and stranded assets
   If demand fails to materialize or regulatory and public backlash halts expansion, the result is empty buildings, drained resources, and broken promises. Some facilities will be repurposed or consolidated. Many will sit idle, relics of overconfidence.

Historic Parallels
This follows the pattern of the dot-com bubble, where companies built fast without profit models. It echoes the crypto rush, where technology was sold as liberation but became a playground for speculation. It mirrors the housing bubble, where the assumption of perpetual growth justified unsustainable development.

In this case, the physical infrastructure — energy-intensive, water-dependent, and land-consuming — makes the consequences of failure far more tangible. If this is a bubble, its burst will not be just economic. It will leave scars on landscapes, utilities, local budgets, and trust in digital progress.

What Does an AI Bubble Look Like?

It would look like a rush to build capacity first and justify it later. Counties would approve huge campuses tied to ten or twenty year power reservations, water allotments, and tax breaks, even though no specific customer workloads are identified. Utilities would promise “growth” while filing for rate increases to finance new substations and transmission for facilities that produce few permanent jobs. Real estate prices near industrial zones would spike on rumor, not revenue. Politicians would pose at groundbreakings. The spreadsheets would lean on forward curves and marketing decks, not signed offtake.

It would look like resource strain dressed up as innovation. Megawatts disappear into private contracts while local manufacturers face curtailments or higher tariffs. Wells and aquifers feel the draw for cooling and dust control. Road bonds and sewer expansions get accelerated for “critical infrastructure,” then households are told to conserve. The public is told that national security or competitiveness requires silence, while NDAs and PILOT agreements keep the numbers out of sight.

It would look like hollow use cases propping up heavy concrete. Most compute would chase ad targeting, model training for marginal product features, or data hoarding with no clear civic value. Operators would talk about medicine and education but deliver inference for marketing and surveillance. Headcount on site would be a security crew, a skeleton ops team, and rotating contractors. The payroll promise would shrink after the abatement is locked in.

It would look like financial engineering outpacing engineering. Chip orders would surge, then secondary markets would fill with barely used GPUs. Developers would flip entitled land and interconnect queues for quick gains. Special purpose vehicles would refinance campuses on rosy utilization forecasts. When interest costs bite, projects would be “paused,” “sequenced,” or “optimized,” which means delayed, downsized, or sold.

It would look like political backlash and legal resistance. Neighborhoods and farm districts would organize around water rights, noise, diesel backup emissions, and line siting. Lawsuits would target defective environmental reviews, rushed annexations, and sweetheart abatements. Public records fights would multiply because basic documents are withheld as proprietary. The story would pivot from “jobs” to “grid stability,” then to “national interest,” as the benefits fail to show up.

It would end with stranded capacity and quiet sales. Some halls would go dark. Others would be consolidated by stronger operators at a discount. Counties would learn the abatement sunsets after the high load leaves. Utilities would seek to roll unrecovered costs into general rates. Officials would insist the assets can be repurposed, but the power contracts, location, and cooling design would limit alternatives.

If you want a quick field test, ask five questions.

1. Who is the anchor tenant, in writing?

2. What percent of contracted power serves public services or in-state industry?

3. How much water is consumed annually and from which source?

4. When do tax benefits turn net positive for the county?

5. What is the decommissioning and remediation plan with posted security?

If those answers are vague, redacted, or “competitive,” you are looking at a bubble.

Proof the Bubble has Formed

Construction Outpacing Capacity

Northern Virginia’s projections are the clearest warning. Dominion says peak demand from data centers could rise to 13.3 gigawatts by 2038, up from about 2.8 gigawatts in 2022, while PJM and Virginia groups note that nearly all projected demand growth through the 2040s is attributable to data centers. Meanwhile, finished facilities in Santa Clara, California, are literally built but dark, with nearly 100 megawatts of capacity unable to energize until the local utility completes grid upgrades, a delay that could last years. Reuters and state coverage tie Mid-Atlantic bill pressure and system upgrades to the same buildout dynamics. This is not demand following customers. It is construction sprinting ahead of verified load.

Speculative capital and inflated valuations

Venture capital is flooding AI. 2025 set new records, with about $193 billion directed to AI startups and more than half of global VC dollars chasing the sector. Public markets mirror the frenzy. Nvidia briefly crossed 4 to 5 trillion dollars in market value, as hyperscalers devoted roughly 60 percent of operating cash flow to capital expenditures and bankers forecast three to four trillion dollars of AI infrastructure spending by 2030. Even bullish analysts concede that spending is running far ahead of proven monetization.

Resource extraction disguised as innovation

Where these campuses land, the resource draw is blunt. In The Dalles, Oregon, Google’s water consumption data only became public after a thirteen-month legal fight, underscoring how basic withdrawal numbers are routinely hidden. In Virginia, watchdogs and planners document multi-gigawatt campus proposals whose power needs rival or exceed nuclear output, while PJM and state groups warn that meeting this load growth keeps fossil plants and new gas capacity on the table. These are industrial footprints with heavy externalities, packaged as inevitable progress.

Techno-utopian marketing shields scrutiny

The sales pitch demands deference. Coverage from Broadband Breakfast and the Piedmont Environmental Council shows utilities and developers using sweeping economic and competitiveness claims to justify extraordinary expansions and grid upgrades, with costs socialized across all customers. In Louisiana, regulators approved new gas turbines and a $550 million transmission line to power Meta’s 4-million-square-foot campus amid public criticism over rushed approvals and thin job guarantees. When the narrative is destiny, scrutiny is treated as a nuisance.

Hollow applications dominate

Independent market work shows where the money actually goes. McKinsey reports that the most commonly reported AI revenue gains are in marketing and sales, and that generative AI’s largest near-term lift is optimizing advertising and customer acquisition. Trade forecasts show digital ad spend accelerating, with industry groups noting rapid adoption of generative tools to crank out ad variants at scale. That is not medicine or public safety. It is targeted persuasion that demands enormous compute without commensurate civic value.

Risk of collapse and stranded assets

We are already seeing stranded concrete. In Silicon Valley, two fully built halls may sit empty for years awaiting power. Bloomberg and the Los Angeles Times detail multi-year energization delays, while Tom’s Hardware reports nearly 100 megawatts of completed but idle capacity. In PJM, Reuters ties rising bills and auction spikes to a buildout racing ahead of generation and wires. Once subsidies sunset or rates bite, weaker projects will pause, sell at a discount, or go dark.

Jobs: thin headcount after the ribbon cutting

Operators and staffing firms acknowledge the numbers. A 12 megawatt facility might run with about 20 full-time staff, and even 40 to 100 megawatt sites often employ only a few dozen. Watchdogs and recent investigations find public costs that routinely exceed a million or two per permanent job, with Ohio deals crossing the two million per job mark. That is a poor trade once construction ends.

Subsidies and abatements carry the load

A fifty-state survey by NAIOP shows at least 36 states offering exemptions or abatements, often for long periods and with modest job requirements that are hard to meet because modern sites are automated. Good Jobs First documents weak disclosure and large public costs, and Business Insider’s Ohio work finds incentives commonly topping two million dollars per permanent job. These are not market wins. They are policy-engineered outcomes.

Hardware signals of overbuying

A secondary market for top-tier accelerators is emerging. Reports show used A100 and even H100 GPUs being stripped and redeployed after policy shocks, while data center trade press now advises operators on how to resell data center GPUs to recover 60 to 80 percent of original cost. Deepening resale channels are exactly what you would expect when procurement outruns workloads.

The Boom is Real, What Happens Next?

Put together, this is not a theoretical risk. It is a visible pattern in U.S. markets right now. Multi-gigawatt forecasts and public bills, water secrecy, thin permanent headcount, rich abatements, idle finished space, and a nascent GPU resale wave are the telltale signs of a bubble forming in steel, concrete, and grid interconnects, not just on a spreadsheet.

The bust will come, what does that mean for your community and the nation?

Resources

• Data center growth drives locals to fight for more say (Nov 2025)

• Global data center expansion and human health: A call for empirical research (May 2025)

• What is a hyperscale data center? | IBM

• The future of US hyperscale data centers | McKinsey

• Mass. prepares for first hyperscale data center

• What is a Hyperscale Data Center and How Does It Work?

Share