THE NUCLEAR BARGAIN: HOW BIG TECH'S AI HUNGER IS QUIETLY RESTARTING THE ATOMIC AGE (AND WHY YOUR NEIGHBOR SHOULD BE WORRIED)
A deeply researched guide to the most consequential — and least discussed — infrastructure gamble of the 21st century
Welcome to the future, where the same Silicon Valley visionaries who gave your teenager a crippling social media addiction and your uncle a conspiracy theory rabbit hole are now pivoting to split atoms in your backyard. All so that a chatbot can write your grocery list faster. Progress, baby. 🚀
The story of AI data centers and nuclear power isn't just a tech story or an energy story. It's a story about who pays, who profits, who glows in the dark, and who gets left holding the bill — literally and figuratively. Buckle up, because this particular can of worms has a half-life of about 10,000 years.
Part One: The Hunger Games — AI's Insatiable Appetite for Power
Here's the fundamental problem nobody in the glossy investor decks wants to talk about plainly: AI is a glutton.
Traditional data centers — the ones quietly hosting your cat videos and cloud spreadsheets — are the polite dinner guests of the electricity world. They eat steadily, predictably, and go home at a reasonable hour. Generative AI data centers are the guests who show up uninvited, eat everything in the fridge at 3 a.m., and then ask if you have a second fridge.
The Raw Numbers Are Staggering
A standard server rack pulls 5 to 10 kilowatts of power. A modern AI rack packed with Nvidia GPUs pulls 40 to 100+ kilowatts — up to twenty times more — and it does so 24 hours a day, 365 days a year, without pause, without weekends, without holidays.
Scale that up to a hyperscale data center campus running tens of thousands of racks, and you're looking at facilities that individually consume as much electricity as a mid-sized American city.
The projections are not subtle:
- U.S. data centers consumed roughly 4% of total national electricity in 2023.
- By 2030, credible projections put that figure at 8% to 9% — a near-doubling in under a decade.
- A single large AI campus can require 100 MW to 500+ MW of dedicated power.
- The entire city of Buffalo, New York uses about 500 MW. Let that sink in.
The Grid Doesn't Care About Your Launch Date
Here's where the comedy — or tragedy, depending on your electricity bill — really begins. Tech companies can design, permit, and construct a massive data center in under two years. The electrical grid infrastructure needed to feed it — transmission lines, substations, transformer upgrades — takes 5 to 10 years to build.
The result is a perfect storm of corporate impatience meeting physical reality:
The bottleneck isn't generating electricity. It's getting it from point A to point B without frying the neighborhood transformer — or more precisely, without making your neighborhood pay for the privilege of having a $10 billion corporate campus move in next door.
Who Pays for All This?
This is the question that utility regulators, consumer advocates, and angry homeowners in Northern Virginia are screaming into the void. When a utility company spends billions of dollars upgrading grid infrastructure to serve a hyperscale data center, the cost has to go somewhere. And historically, "somewhere" has a disturbing tendency to mean your monthly electricity bill.
The policy debate is fierce: should corporate tech giants pay for the dedicated grid upgrades their facilities require, or should those costs be socialized across all ratepayers? Consumer advocates argue the former. Utilities, who love having giant anchor customers, often prefer the latter. The residents of places like Loudoun County, Virginia — already living next to the largest concentration of data centers on Earth — are increasingly discovering the answer is "a little of both, and also the noise and the water usage are your problem too."
Part Two: The Current Nightmare — What's Already Happening to Your Neighborhood
Before we get to the glowing future, let's spend a moment appreciating the present-tense problems that communities near existing AI data centers are already dealing with. Because the nuclear chapter is coming, but the data center chapter is already here.
Water: The Invisible Consumption
Most large data centers use evaporative cooling — essentially giant industrial swamp coolers — to keep their servers from melting. A single large facility can consume millions of gallons of water per day, drawing from local aquifers, rivers, and municipal supplies.
In regions already stressed by drought — looking at you, Arizona and the American Southwest — this is not a minor footnote. It's a genuine crisis. Communities that have managed water rights for generations are watching corporate campuses arrive and drain local supplies with the casual confidence of someone who has very good lawyers.
Noise: The Hum That Never Stops
Industrial cooling towers, backup diesel generators, and the constant mechanical roar of HVAC systems create a persistent low-frequency noise that residents near data centers describe as maddening. Unlike a construction project that eventually ends, this noise is permanent — 24/7/365, as reliable as the servers it cools. Property values near large data centers have shown measurable declines in multiple studies, a fact that the industry's PR materials somehow never mention.
Heat Islands and Land Use
Large data center campuses consume hundreds of acres of land, generate significant waste heat that affects local microclimates, and often arrive with tax incentive packages so generous that local governments find themselves having subsidized the very infrastructure that's now straining their schools, roads, and emergency services.
The Electricity Bill Problem
Here's the kicker that ties it all together: as utilities invest in grid upgrades to serve data centers, and as those data centers consume ever-larger shares of regional electricity capacity, residential and small business electricity rates rise. This is already documented in Virginia, Georgia, and parts of the Midwest. The people least able to absorb higher utility costs are subsidizing the infrastructure of the world's most profitable companies. If you find that arrangement troubling, you are not alone — and you haven't even heard the nuclear part yet.
Part Three: The Nuclear Pivot — "Don't Worry, We've Thought of Everything"
So the grid can't keep up. Renewable energy — as genuinely wonderful as solar and wind are — has a fatal flaw for AI workloads: intermittency. The sun sets. The wind stops. AI servers do not.
Battery storage technology is improving rapidly, but storing the kind of power a 500 MW data center needs through a three-day cloudy, windless stretch requires battery infrastructure that doesn't yet exist at commercial scale and would cost more than the data center itself.
Enter nuclear power — specifically, the shiny new concept of Small Modular Reactors (SMRs) — stage left, wearing a hard hat and carrying a PowerPoint deck full of optimistic projections.
The pitch from the TechBro Oligarchy™ is elegant in its simplicity: nuclear power is carbon-free, runs 24/7, doesn't depend on weather, and the new generation of reactors is totally different from the old ones. Smaller, safer, factory-built, cheaper (eventually), and deployable in clusters wherever you need them.
What could possibly go wrong?
(Narrator: Several things.)
What Exactly IS a Small Modular Reactor?
Before we get to the problems, let's understand what's actually being proposed, because the technology landscape is genuinely fascinating — and genuinely varied.
SMRs are broadly defined as nuclear reactors producing 10 MW to 300 MW of electricity per module, compared to the 1,000+ MW of a traditional gigawatt-scale plant. The "modular" part means components are factory-built and shipped to site — theoretically reducing construction costs and timelines.
But "SMR" is not one technology. It's a category containing several very different approaches:
| Reactor Type | Coolant | Key Example | Key Feature |
|---|
| Advanced Light Water | Water | Holtec SMR-300 | Lowest regulatory risk; proven fuel |
| High-Temp Gas-Cooled | Helium | X-energy Xe-100 | Extreme efficiency; industrial heat |
| Sodium-Cooled Fast | Liquid Sodium | TerraPower Natrium | Burns nuclear waste as fuel |
| Molten Salt | Fluoride Salts | Kairos Power KP-FHR | Low pressure; no explosive rupture risk |
| Microreactors | Varies | Oklo Aurora | Truck-deployable; 1–20 MW |
| Deep Borehole | Water | Deep Fission | Underground; 15 MW; genuinely weird |
Each of these technologies has different safety profiles, different waste outputs, different fuel requirements, and different timelines to commercial deployment. Lumping them all together as "SMRs" is a bit like calling a bicycle, a motorcycle, and a semi-truck all "vehicles" — technically accurate, practically misleading.
Part Four: The Big Tech Nuclear Shopping Spree
The deals being struck right now represent the largest private investment in nuclear energy since the 1970s. The hyperscalers aren't just buying electricity — they're buying entire reactors, funding startups, and in some cases, essentially becoming nuclear energy companies with a side hustle in cloud computing.
Here's the current scoreboard:
| Tech Company | Nuclear Partner | Deal Structure | Scale | Target Date |
|---|
| Microsoft | Constellation Energy | 20-year PPA — Three Mile Island restart | 835 MW | 2028 |
| Amazon (AWS) | X-energy / Energy Northwest | SMR fleet + $500M equity investment | 320–960 MW | Early 2030s |
| Google | Kairos Power | World's first multi-reactor corporate SMR deal | 500 MW (6–7 units) | 2030–2035 |
| Meta | Constellation / Oklo | Clinton Plant PPA + 1.2 GW SMR campus | 1.1 GW + 1.2 GW | 2030s |
| Oracle | Undisclosed SMR developers | Three on-site SMRs for dedicated AI campus | ~1 GW | TBD |
The geographic footprint of these deployments is already taking shape across specific American communities:
Where the Reactors Are Actually Going
Oak Ridge, Tennessee — Google and Kairos Power are breaking ground on the Hermes 2 fluoride-salt-cooled reactor. Target: 50 MW by 2030, feeding into the TVA grid. Oak Ridge has a long nuclear history, which is either reassuring or a sign that the town has simply run out of reasons to object.
Richland, Washington — Amazon and X-energy are building the Cascade Advanced Energy Facility near the existing Columbia Generating Station. Four pebble-bed gas-cooled reactors in Phase 1 (320 MW), scaling to 12 units and 960 MW. Richland sits adjacent to the Hanford Site — the most contaminated nuclear site in the Western Hemisphere — which adds a certain je ne sais quoi to the local real estate market.
Kemmerer, Wyoming — Bill Gates' TerraPower is constructing the flagship Natrium sodium-cooled fast reactor here. 345 MW base capacity, with molten salt storage that can surge to 500 MW. Meta has a pipeline agreement for up to eight more globally. Kemmerer has a population of about 2,600 people, which means the reactor will likely employ more engineers than the town has residents.
Covert Township, Michigan — Holtec International is deploying two SMR units alongside the historic Palisades plant on the shores of Lake Michigan. Target: 2030. The name "Covert Township" is doing a lot of unintentional work here.
Eielson Air Force Base, Alaska — Oklo's Aurora microreactor (1–5 MW, sodium-cooled) is scheduled to go live by 2027, making it the first operational commercial microreactor in U.S. history. The Air Force's logic: if the grid goes down, the base needs power. Hard to argue with that.
Piketon, Ohio / Texas / Utah / Kansas — Various early-stage deployments and borehole reactor explorations. Deep Fission's approach — dropping a 15 MW pressurized water reactor a mile underground inside a 30-inch borehole — is either brilliant engineering or the setup to a very specific kind of disaster movie.
Part Five: The Dangers — What the Press Releases Don't Mention
Here's where we stop being politely enthusiastic and start being rigorously honest. Because the dangers of this nuclear buildout are real, documented, and being actively minimized by the people with the most financial interest in minimizing them.
🔴 Danger #1: Weapons Proliferation — The HALEU Problem
Many advanced SMR designs — particularly the high-temperature gas-cooled and fast-fission varieties — require HALEU: High-Assay Low-Enriched Uranium, enriched to between 5% and 20% concentration.
Standard reactor fuel is enriched to about 3–5%. Weapons-grade uranium is enriched to 90%+. HALEU sits uncomfortably in the middle — not weapons-grade, but significantly closer to it than conventional fuel.
The proliferation concern is straightforward: the more HALEU circulating in the world, the more opportunities exist for it to be intercepted, stolen, or diverted. And here's the supply chain problem that nobody in the SMR pitch deck mentions prominently: the primary commercial source of HALEU outside the United States is currently Russia.
The U.S. is scrambling to build domestic HALEU enrichment capacity, but it doesn't exist at commercial scale yet. The ADVANCE Act and DOE programs are attempting to address this, but the timeline gap between "we need HALEU now" and "we can produce HALEU domestically at scale" is measured in years — years during which the supply chain remains geopolitically fragile and proliferation-adjacent.
🔴 Danger #2: Exotic Coolants and Untested Engineering
The advanced SMR designs use coolants that are genuinely exotic by the standards of conventional nuclear engineering:
Liquid Sodium (TerraPower Natrium, Oklo Aurora):
- Excellent heat transfer properties
- Enables "fast fission" that can burn nuclear waste
- Reacts violently and explosively with water or air
- Requires entirely separate secondary cooling loops to prevent sodium-water contact
- The world has limited operational experience with sodium-cooled reactors at commercial scale
Molten Fluoride Salts (Kairos Power):
- Operates at low pressure, eliminating explosive rupture risk
- Highly corrosive to most conventional materials
- Requires specialized alloys and containment systems
- Long-term corrosion behavior in commercial operation is not well-characterized
- The Molten Salt Reactor Experiment at Oak Ridge ran in the 1960s — for four years, then was shut down
Helium Gas (X-energy pebble-bed):
- High-temperature operation enables excellent efficiency
- Pebble fuel handling at commercial scale is complex
- Germany's THTR-300 pebble-bed reactor had operational problems and was shut down after 423 days of full-power operation
None of this means these technologies can't work. It means they haven't been proven at commercial scale over decades of operation, and the SMR industry is essentially asking communities to be the test case.
🔴 Danger #3: The Waste Problem — More of It, Not Less
Here's the counterintuitive finding that SMR developers really don't enjoy discussing: some SMR designs produce significantly more radioactive waste per unit of energy than conventional large reactors.
A landmark study from Stanford University and the University of British Columbia found that certain SMR designs could generate 2 to 30 times more low-and-intermediate-level radioactive waste per megawatt-hour than traditional plants.
The physics reason is elegant and unfortunate: smaller reactors have a higher surface-area-to-volume ratio. This means more of the surrounding structural materials — steel, concrete, shielding — are exposed to neutron bombardment, becoming radioactively "activated" over time. When the reactor is eventually decommissioned, all of that activated structural material is radioactive waste.
SMR developers dispute this characterization, arguing that modular designs can mitigate the effect. The debate is ongoing. The waste, however, will not be.
🔴 Danger #4: The Waste Storage Crisis — Nowhere to Put It
The United States does not have a permanent repository for high-level nuclear waste. Yucca Mountain in Nevada was designated as the national repository, licensed, and then politically killed in 2010. It has been in regulatory and political limbo ever since.
Currently, spent nuclear fuel from America's 96 operating reactors sits in:
- Spent fuel pools on-site at reactor facilities (water-cooled storage)
- Dry cask storage on-site (passive, concrete-and-steel cylinders)
- Consolidated Interim Storage facilities (proposed; facing fierce legal opposition)
The proposed interim storage facilities — Holtec's HI-STORE in New Mexico and Waste Control Specialists in Texas — are facing legal challenges from state governments, environmental groups, and Indigenous Nations whose lands border the proposed sites.
This is not a minor logistical detail. It is a fundamental unresolved problem. Adding dozens of new SMRs to the national fleet will generate additional spent fuel that has nowhere permanent to go, joining the existing backlog of waste that has been accumulating since the 1950s.
The fast-fission designs (TerraPower, Oklo) offer a partial answer: they can burn legacy spent fuel as their primary fuel source, reducing the radioactivity lifespan of remaining waste from hundreds of thousands of years to a few hundred years. This is genuinely impressive technology — if it works at commercial scale, which remains to be demonstrated.
🔴 Danger #5: The Security Multiplication Problem
Traditional large nuclear plants are heavily secured facilities with elite physical security forces, elaborate emergency planning zones, and decades of operational security protocols. They are hard targets.
The SMR business model involves deploying dozens of smaller reactors across dispersed locations — rural grids, corporate campuses, military bases, remote communities. Each of these sites contains nuclear material. Each requires physical security. Each requires emergency planning.
Spreading active nuclear infrastructure across scores of sites doesn't just multiply the security requirement linearly — it multiplies the attack surface and the operational complexity of maintaining consistent security standards across a geographically dispersed fleet. The NRC and DOE are developing new security frameworks for this reality, but the frameworks are newer than the deployment timelines.
🔴 Danger #6: The Financial Track Record Is Not Inspiring
Let's talk about money, because the financial history of nuclear power in America is a horror story wearing a hard hat.
NuScale Power — once the darling of the SMR industry, the first company to receive NRC design certification for an SMR — saw its flagship project, the Carbon Free Power Project in Utah, completely canceled in 2023 after projected costs exploded past $9 billion for a 462 MW facility. The cost per megawatt made it economically uncompetitive with virtually every other energy source.
Vogtle Units 3 and 4 in Georgia — the most recent large conventional nuclear plant completed in the U.S. — came in at roughly $35 billion, more than double the original estimate, and 7 years late. Georgia Power ratepayers are still absorbing the cost overruns.
The SMR industry's counter-argument is that factory fabrication will drive costs down through manufacturing scale — the same learning curve that made solar panels 90% cheaper over two decades. This is a reasonable argument. It is also, at this point, entirely theoretical. The first-of-a-kind cost premium for novel nuclear technology is real, documented, and historically severe.
🛑 Part Six: Who's Pushing Back — and Why
Opposition to the SMR buildout is not monolithic, and it's not simply "people who are afraid of atoms." It comes from distinct groups with distinct, well-reasoned concerns:
Environmental Groups
The Union of Concerned Scientists, Food & Water Watch, and the NRDC make a pointed argument: SMRs won't be commercially operational until the 2030s at the earliest. The climate crisis is happening now. Every dollar of federal subsidy going to unproven SMR technology is a dollar not going to immediately deployable solar, wind, and battery storage. They call SMRs a "climate distraction" — a shiny future solution that delays present-tense action.
Ratepayer Advocates
Consumer utility boards across multiple states are raising alarms about who bears the financial risk when SMR projects go over budget — as nuclear projects historically do. The NuScale cancellation is Exhibit A. When a utility signs a long-term power purchase agreement with an SMR developer and the project fails or costs triple, ratepayers typically absorb the difference through higher monthly bills.
Local Communities (NIMBY, But Make It Reasonable)
Communities near proposed SMR sites are raising concerns about:
- Localized radioactive waste storage pads that will exist for decades
- Property value impacts
- Zoning of nuclear facilities near municipal water supplies
- Emergency planning zone requirements
- The simple fact that nobody asked them
Indigenous Nations
This is perhaps the most serious and least-covered dimension of the nuclear buildout. Uranium mining, HALEU enrichment, and proposed interim waste storage sites disproportionately impact Indigenous lands. Holtec's proposed HI-STORE facility in New Mexico borders tribal lands. Indigenous leaders are filing legal challenges arguing they are being asked to bear the environmental burden of the nuclear fuel cycle without meaningful consent or benefit. This is not a new pattern in American nuclear history — it is a continuation of one.
Part Seven: The Regulatory Landscape — Faster, Whether You Like It or Not
The federal government has made a clear choice: accelerate nuclear deployment. The ADVANCE Act (Accelerating Deployment of Versatile, Advanced Nuclear for Clean Energy), signed into law in 2024, directed the NRC to:
- Cut standard multi-year review cycles to an 18-month cap for approved SMR designs
- Create streamlined licensing pathways (Part 53 and Part 57 rules) for advanced and microreactor designs
- Reduce licensing fees for certain applicants
- Hire more NRC staff to process the incoming wave of applications
The NRC has also approved the first SMR design certification (NuScale, though that project was canceled) and is actively reviewing designs from Kairos, TerraPower, Oklo, and others.
FERC (the Federal Energy Regulatory Commission) is simultaneously wrestling with the "co-location" problem — the question of whether a tech company can plug a data center directly into a nuclear plant's output, bypassing the public grid. FERC rejected the Amazon-Talen Energy co-location deal in 2024, citing concerns about shifting grid costs to regular consumers. The policy fight is ongoing.
Part Eight: The Financial Architecture — Who's Really Paying
Let's follow the money, because the financial structure of this nuclear buildout is more complex than "tech companies pay for reactors."
Tech companies are signing Power Purchase Agreements (PPAs) — long-term contracts to buy electricity at fixed prices. They are also making equity investments in SMR developers (Amazon's $500M in X-energy; Sam Altman's backing of Oklo). But they are generally not taking on construction risk directly.
Utilities are taking on construction risk, often with ratepayer backstops — meaning if the project goes over budget, they can petition regulators to recover costs through higher electricity rates.
Federal taxpayers are funding the DOE's Advanced Reactor Demonstration Program, the Loan Programs Office (which has extended multi-billion-dollar loan guarantees to projects like Vogtle and Palisades), and various research and enrichment infrastructure programs. The HALEU enrichment buildout alone will require billions in federal investment.
State governments are providing tax incentives, regulatory fast-tracking, and in some cases, ratepayer guarantees to attract SMR deployments.
The result is a financial architecture where the upside (cheap, reliable clean energy; tech company profits; shareholder returns) flows primarily to corporations and investors, while the downside risk (cost overruns, stranded assets, waste management, grid disruption) is distributed across taxpayers, ratepayers, and communities.
This is not unique to nuclear power. It is the standard American model for large infrastructure. It is worth naming clearly.
Part Nine: The Genuine Possibilities — Because It's Not All Doom
Here's the thing about this whole nuclear-AI-industrial complex: the underlying technology, if it works as advertised, offers genuinely transformative possibilities. Let's be fair.
♻️ Burning the Old Waste
TerraPower's Natrium and Oklo's Aurora reactors can run on recycled spent nuclear fuel — the legacy waste sitting in dry casks at reactor sites across America. By "burning" this material as fuel, they can reduce its radioactive lifespan from hundreds of thousands of years to a few hundred years. This is not science fiction. It is physics that has been demonstrated at smaller scales. If it works commercially, it transforms an existing environmental liability into a clean energy resource.
🌡️ Industrial Decarbonization
High-temperature SMRs (particularly the gas-cooled varieties) can produce process heat at temperatures needed for steel production, cement manufacturing, and clean hydrogen generation — industrial processes that are currently very difficult to decarbonize with solar or wind. This is potentially a bigger deal for climate than powering data centers.
🏔️ Energy Independence for Remote Communities
Microreactors (Oklo Aurora, etc.) could provide reliable, carbon-free electricity to remote Alaskan communities, island nations, and off-grid industrial sites that currently run on diesel generators — one of the most expensive and carbon-intensive forms of electricity generation. This is a genuine humanitarian application.
🔋 Grid Stability
As the grid absorbs more intermittent renewable energy, having dispatchable baseload nuclear power — power you can count on regardless of weather — becomes more valuable, not less. SMRs could serve as the "backbone" that makes a high-renewable grid reliable.
💼 Economic Development
The SMR buildout will create high-skilled, high-wage jobs in manufacturing, engineering, construction, and operations — many of them in rural communities that have seen economic decline for decades. Kemmerer, Wyoming; Oak Ridge, Tennessee; Richland, Washington — these are places where good industrial jobs matter enormously.
Part Ten: The Skynet Footnote — A Brief Meditation on Irony
Let's take a moment to appreciate the full arc of what's happening here.
The same technology companies that:
- Built social media platforms now linked to documented mental health crises in adolescents
- Deployed recommendation algorithms optimized for engagement over wellbeing
- Concentrated unprecedented wealth and political power in the hands of a small number of individuals
- Promised that each new technology would make life better for everyone
...are now telling us that:
- AI will make life better for everyone
- The nuclear plants needed to power AI will be safe and well-managed
- The communities near these facilities will benefit
- Trust them
To be clear: AI does have genuine potential to improve lives — in medicine, in scientific research, in education, in accessibility. The technology is real and the benefits are real.
But the track record of the specific institutions making these promises — on privacy, on mental health, on market concentration, on the distribution of benefits versus harms — is not a track record that invites unconditional trust.
The communities being asked to host data centers and nuclear reactors are, in many cases, the same communities that have historically borne the costs of industrial development while watching the profits flow elsewhere. They are right to ask hard questions. They are right to demand answers before ground is broken, not after.
Skynet, for what it's worth, didn't announce its intentions in a press release. It just quietly became indispensable until it wasn't optional anymore. We're not saying that's what's happening here. We're just noting that the timeline rhymes.
The Bottom Line: A Scorecard
Here's where things actually stand as of mid-2026:
| Category | Current Status | Risk Level |
|---|
| AI power demand growth | Confirmed, accelerating | 🔴 High |
| Grid infrastructure gap | Real and widening | 🔴 High |
| Ratepayer cost impacts | Already occurring | 🟡 Medium-High |
| SMR technology readiness | Promising but unproven at scale | 🟡 Medium |
| HALEU fuel supply | Constrained; Russia-dependent | 🔴 High |
| Nuclear waste storage | No permanent solution exists | 🔴 High |
| Proliferation risk | Increased with HALEU expansion | 🟡 Medium |
| Financial cost overrun risk | Historically severe | 🔴 High |
| Indigenous land rights | Active legal conflicts | 🔴 High |
| Long-term climate benefit | Potentially significant | 🟢 Positive |
| Industrial decarbonization potential | Genuinely promising | 🟢 Positive |
| Remote community energy access | Real near-term application | 🟢 Positive |
What to Watch For
The next five years will be decisive. Here are the specific milestones that will tell us whether the SMR revolution is real or another expensive promise:
2027 — Does Oklo's Aurora microreactor at Eielson AFB actually go live on schedule? First-of-a-kind nuclear projects almost never do.
2028 — Does Three Mile Island Unit 1 (Crane Clean Energy Center) successfully restart and deliver power to Microsoft's grid commitment?
2030 — Do the Kairos, TerraPower, and X-energy projects hit their operational targets? Cost overruns at any of these will reverberate across the entire industry.
Ongoing — Does Congress resolve the Yucca Mountain/permanent repository impasse? Without a permanent waste solution, every new reactor adds to an unresolved liability.
Ongoing — Do FERC and state utility regulators successfully prevent data center grid costs from being socialized onto residential ratepayers?
Ongoing — Do the legal challenges from Indigenous Nations over waste storage sites succeed or fail?
The answers to these questions will determine whether the nuclear-AI industrial complex delivers on its promises — or delivers another generation of cost overruns, stranded assets, and communities left holding the radioactive bag.
The road to AI making life better for everyone is paved with genuinely good intentions, extraordinary engineering ambition, serious unresolved risks, and a financial structure that has a long historical habit of privatizing the gains while socializing the losses. The technology is real. The benefits are possible. The dangers are documented. The communities being asked to host all of this deserve honest answers — not just glossy renderings of reactors that haven't been built yet.
Whether this ends up being the clean energy revolution we need, or the next chapter in a very long story about who pays for progress, depends almost entirely on whether the people asking hard questions get heard before the concrete is poured.
The atoms don't care either way. They'll be radioactive long after the press releases are forgotten. ⚛️
Sources drawn from: U.S. Energy Information Administration (EIA), Nuclear Regulatory Commission (NRC), Stanford University / University of British Columbia SMR waste study, Union of Concerned Scientists, Food & Water Watch, NRDC, Bulletin of the Atomic Scientists, TerraPower, Kairos Power, X-energy, Oklo Inc., Federal Energy Regulatory Commission (FERC), Department of Energy Advanced Reactor Demonstration Program, and reporting from Inside Climate News, CNBC, and Utility Dive.
Full Source List & Reference Links
🔌 AI Power Demand & Data Center Energy Consumption
IEA — The Path to a New Era for Nuclear Energy
International Energy Agency — global nuclear generation forecasts and data center demand projections
🔗 https://www.iea.org/reports/the-path-to-a-new-era-for-nuclear-energy/executive-summary
ASCE — Demand for Data Centers Soars; Could SMRs Meet the Need?
American Society of Civil Engineers — infrastructure analysis of data center power surge
🔗 https://www.asce.org/publications-and-news/civil-engineering-source/article/2025/12/17/demand-for-data-centers-soars-could-small-modular-reactors-meet-the-need
ScienceDirect — The Potential Role of SMRs in Data Centers
Peer-reviewed academic analysis of SMR deployment for distributed data center power
🔗 https://www.sciencedirect.com/science/article/pii/S1738573324005643
U.S. Energy Information Administration — Data Center Energy Use
Official federal statistics on electricity consumption by sector
🔗 https://www.eia.gov
☢️ SMR Technology — What It Is & How It Works
World Nuclear Association — Small Modular Reactors
Comprehensive technical overview of all SMR reactor types, fuel requirements, and deployment models
🔗 https://world-nuclear.org/information-library/nuclear-fuel-cycle/nuclear-power-reactors/small-nuclear-power-reactors
U.S. Department of Energy — Advanced Reactor Demonstration Program
Federal program funding TerraPower, X-energy, Kairos, and other SMR developers
🔗 https://www.energy.gov/ne/advanced-reactor-demonstration-program
Idaho National Laboratory — SMR Research & Development
Primary federal nuclear research facility overseeing advanced reactor testing
🔗 https://inl.gov/nuclear-energy/advanced-nuclear-reactors
Nuclear Energy Institute — SMR Overview
Industry association technical briefings on modular reactor designs
🔗 https://www.nei.org/advantages/clean-air/nuclear-energy-overview
🤝 Big Tech Nuclear Deals
Constellation Energy — Three Mile Island / Crane Clean Energy Center Restart (Microsoft PPA)
Official press release of the 20-year Microsoft power purchase agreement
🔗 https://www.constellationenergy.com/news/2024/Constellation-to-Launch-Crane-Clean-Energy-Center-Restoring-Jobs-and-Carbon-Free-Power-to-The-Grid.html
World Nuclear News — Constellation to Restart Three Mile Island for Microsoft
Independent reporting on the TMI Unit 1 restart deal and timeline
🔗 https://www.world-nuclear-news.org/articles/constellation-to-restart-three-mile-island-unit-powering-microsoft
NucNet — Constellation Secures $1 Billion Federal Loan for TMI Restart
Federal loan guarantee details for the Crane Clean Energy Center
🔗 https://www.nucnet.org/news/constellation-secures-usd1-billion-federal-loann-for-three-mile-island-restart-11-3-2025
Penn Capital Star — Microsoft Describes Three Mile Island as "Once-in-a-Lifetime"
Regional investigative reporting on the deal's local economic and grid impacts
🔗 https://penncapital-star.com/economy/microsoft-describes-three-mile-island-plant-as-a-once-in-a-lifetime-opportunity/
Amazon — Official SMR Investment Announcement (X-energy & Energy Northwest)
Amazon's official press release on its nuclear energy strategy and SMR investments
🔗 https://www.aboutamazon.com/news/sustainability/amazon-nuclear-small-modular-reactor-net-carbon-zero
X-energy — Amazon, Korea Hydro & Doosan Partnership Announcement
X-energy's official announcement of its expanded SMR manufacturing partnership
🔗 https://x-energy.com/news/x-energy-amazon-korea-hydro-amp-nuclear-power-and-doosan-enerbility-announce-partnership-to-scale-advanced-nuclear-energy-for-ai-infrastructure/
Google Blog — First Advanced Nuclear Reactor Project with Kairos Power & TVA
Google's official announcement of the Hermes 2 Oak Ridge deployment
🔗 https://blog.google/company-news/outreach-and-initiatives/sustainability/google-first-advanced-nuclear-reactor-project-with-kairos-power-and-tennessee-valley-authority/
CNBC — Google & Kairos Power Plan Advanced Nuclear Plant for Tennessee
Independent financial reporting on the Google-Kairos-TVA deal structure
🔗 https://www.cnbc.com/2025/08/18/google-kairos-nuclear-smr-tennessee-valley-authority-tva-data-center-ai.html
Kairos Power — Google Partnership Official Page
Kairos Power's official multi-plant agreement details with Google
🔗 https://www.kairospower.com/google
NucNet — Google, Kairos & TVA Sign Landmark Nuclear Power Deal
Independent nuclear industry reporting on the TVA grid integration agreement
🔗 https://www.nucnet.org/news/google-kairos-and-tennessee-valley-authority-sign-landmark-nuclear-power-deal-8-2-2025
Introl — Nuclear Power for AI: Inside the Data Center Energy Deals
Comprehensive analysis of all hyperscaler nuclear strategies: Microsoft, Google, Amazon, Meta
🔗 https://introl.com/blog/nuclear-power-ai-data-centers-microsoft-google-amazon-2025
📍 SMR Deployment Locations
TerraPower — Natrium Plant, Kemmerer Wyoming
Official project page for the flagship sodium-cooled fast reactor demonstration
🔗 https://www.terrapower.com/natrium
Energy Northwest — Cascade Advanced Energy Facility (Richland, WA)
Official utility page for the Amazon/X-energy pebble-bed SMR project in Washington State
🔗 https://www.energy-northwest.com/whoweare/cleanenergy/Pages/Advanced-Nuclear.aspx
Inside Climate News — SMR Deployment Tracking
Independent investigative journalism covering SMR siting, community impacts, and timelines
🔗 https://insideclimatenews.org
Utility Dive — Holtec Palisades / Covert Township Michigan
Trade publication reporting on Holtec's SMR deployment alongside Palisades restart
🔗 https://www.utilitydive.com
⚠️ Dangers, Risks & Criticism
Bulletin of the Atomic Scientists — Data Centers Powered by Next-Gen Nuclear? Don't Fall for Big Tech's PR Hype
Critical analysis of SMR timelines, HALEU risks, and tech company nuclear claims — July 2026
🔗 https://thebulletin.org/2026/07/data-centers-powered-by-next-gen-nuclear-dont-fall-for-big-techs-pr-hype/
Union of Concerned Scientists — Nuclear Power Safety & SMR Risks
Watchdog organization's technical analysis of proliferation, waste, and safety concerns
🔗 https://blog.ucs.org
Food & Water Watch — SMR Opposition & Climate Distraction Argument
Environmental advocacy group's case against federal SMR subsidies
🔗 https://www.foodandwaterwatch.org
Natural Resources Defense Council — Nuclear Energy & Climate Policy
NRDC's policy position on SMRs versus immediately deployable renewables
🔗 https://www.nrdc.org/issues/nuclear-energy
Stanford / UBC Study — SMR Waste Production Analysis
The landmark peer-reviewed study finding some SMRs produce 2–30x more low-level waste
🔗 https://www.pnas.org/doi/10.1073/pnas.2111833119
💸 Financial Risks & NuScale Cancellation
The Energy Mix — NuScale Cancellation & SMR Cost Overruns
Investigative reporting on the Carbon Free Power Project collapse and cost escalation
🔗 https://www.theenergymix.com
Environmental Working Group — Nuclear Ratepayer Cost Analysis
Consumer cost impact analysis of nuclear subsidies and utility rate structures
🔗 https://www.ewg.org
World Nuclear Industry Status Report
Annual independent assessment of global nuclear economics, construction delays, and cancellations
🔗 https://www.worldnuclearreport.org
🏛️ Regulation, Policy & the ADVANCE Act
Nuclear Regulatory Commission — Advanced Reactor Licensing
Official NRC page on SMR design certifications, Part 53 rules, and streamlined review pathways
🔗 https://www.nrc.gov/reactors/new-reactors/advanced.html
Federal Energy Regulatory Commission — Co-location & Grid Policy
FERC rulings on data center direct nuclear co-location and grid cost allocation
🔗 https://www.ferc.gov
Carnegie Endowment for International Peace — Nuclear Proliferation & HALEU
Policy analysis of HALEU supply chain risks and weapons proliferation concerns
🔗 https://carnegieendowment.org
Multistate.us — State-Level SMR Regulatory Legislation Tracker
Tracks state bills fast-tracking SMR permitting across the U.S.
🔗 https://www.multistate.us
🗑️ Nuclear Waste & Storage
Nuclear Regulatory Commission — Spent Fuel Storage
Official NRC guidance on dry cask storage, spent fuel pools, and interim storage
🔗 https://www.nrc.gov/waste/spent-fuel-storage.html
Bulletin of the Atomic Scientists — Nuclear Waste & Indigenous Land Rights
Analysis of waste storage siting disproportionately impacting tribal nations
🔗 https://thebulletin.org
GIS Reports Online — SMR Waste & Community Opposition
Geopolitical and community-level analysis of nuclear waste siting conflicts
🔗 https://www.gisreportsonline.com
🏭 Existing U.S. Nuclear Fleet
U.S. Energy Information Administration — U.S. Nuclear Plant List
Complete official list of all 57 U.S. commercial nuclear plants and 96 reactors
🔗 https://www.eia.gov/nuclear/power-plants
Nuclear Regulatory Commission — Operating Reactors
NRC's official database of all licensed and operating commercial reactors by state
🔗 https://www.nrc.gov/reactors/operating.html
All links were verified as active sources as of August 2026. Government (.gov), academic (.edu), and established institutional sources are the most stable long-term references. Trade publications and news sources reflect reporting current at time of article publication.
