How Market Displacement Could Help Solve the Data Center Crisis
The AI economy is creating an enormous demand for computing power. The question is no longer simply whether America will build data centers, but where they will be built, who will pay for the infrastructure, and whether a better economic model can displace the most disruptive projects.
Updated September 13, 2026
Key Takeaways
- Data centers are becoming a major U.S. electricity-load problem. Lawrence Berkeley National Laboratory's 2025 update estimates that U.S. data centers could consume about 11.8% of national electricity by 2030, with a scenario range of 9.5% to 15.3%.
- Community opposition has a legitimate infrastructure component. Large facilities can create demands for electricity transmission, substations, backup generation, water, roads and industrial-scale cooling in places that were not previously designed for such loads.
- Noise is a genuine environmental-health issue. Chronic environmental noise is associated with sleep disturbance, annoyance and cardiovascular and other health effects. That does not mean every symptom attributed to a data center is caused by electromagnetic fields or by the facility itself.
- The evidence for “dirty electricity” near data centers is interesting but not definitive. A 2024 Bloomberg analysis of approximately 770,000 residential sensors found a strong geographic association between power-quality distortion and proximity to significant data-center activity. However, the analysis was observational and utilities challenged aspects of the methodology. It should therefore be treated as a signal requiring independent engineering investigation, not proof that data centers are causing illness.
- Putting data centers in space is an intriguing long-term technology experiment, not yet a demonstrated replacement for terrestrial computing. SpaceX has sought FCC authorization for a potential constellation of up to one million orbital data-center satellites, while Google is testing its own space-computing concept through Project Suncatcher.
- Advanced nuclear energy may become part of the displacement strategy. The United States is already moving beyond conventional reactor development, but commercial thorium molten-salt reactors remain developmental. Advanced uranium-fueled reactors are substantially closer to deployment.
- The most practical near-term strategy may be better siting rather than eliminating data centers. Remote locations with abundant power, cold climates, available land and dedicated energy infrastructure could reduce conflict with residential communities.
1. The “Law of Displacement”: A Useful Hypothesis, Not a Scientific Law
Strictly speaking, this is not a recognized law of economics, engineering or public policy. It is better understood as a recurring market hypothesis (1):
When a harmful, expensive or politically unpopular technology has become deeply entrenched, opposition alone may not be enough to remove it. A sufficiently competitive alternative can be far more powerful because it allows capital to move without requiring the entire economic system to give up the function that the old technology provided.
This distinction matters.
It is tempting to interpret every historical technology transition as proof of a universal law. History is more complicated than that. Technologies disappear for many reasons: regulation, changing consumer preferences, resource constraints, technical failure, competition, patents, economics, liability and political pressure.
Nevertheless, the underlying idea is important. If society needs a service, simply demanding that the service disappear is often less effective than developing a superior way to provide it.
That may be particularly relevant to artificial intelligence.
2. America's AI Boom Has Become an Electricity Problem
AI is not just a software story.
It is increasingly an infrastructure story.
Training and operating large AI systems requires enormous quantities of computing hardware. Those chips require electricity. The electricity requires generation, transmission, substations and cooling infrastructure.
The resulting growth in electricity demand is now large enough to influence national energy planning.
According to Lawrence Berkeley National Laboratory's 2025 U.S. Data Center Energy Usage Report, U.S. data centers could account for approximately 11.8% of total U.S. electricity consumption by 2030, with modeled scenarios ranging from approximately 9.5% to 15.3%.
The U.S. Energy Information Administration has also projected record national electricity consumption in 2026 and 2027, with AI-intensive data centers identified as a major contributor to load growth.
The important point is not any single forecast.
It is that the direction of travel is unmistakable: AI is converting computing demand into physical energy demand.
3. Why Data Centers Create Local Political Conflict
A data center can be economically valuable while simultaneously creating concentrated local costs.
The same facility may bring investment, construction activity, tax revenue and jobs while also requiring:
- large amounts of electricity;
- new transmission lines and substations;
- backup generation;
- large amounts of industrial cooling infrastructure;
- water or alternative cooling systems;
- roads and heavy construction traffic;
- large parcels of land; and
- continuous mechanical equipment that can generate substantial noise.
This creates a classic infrastructure problem.
The benefits are often distributed while some of the costs are concentrated.
A national technology company may receive the economic benefit from additional computing capacity, while nearby residents may experience construction, traffic, visual disruption, noise or concerns about utility costs and infrastructure strain.
That does not automatically mean a particular data center is harmful or that opposition to it is justified. It does mean that the economic accounting should include the full infrastructure cost.
4. Electricity Prices: The Question Is More Complicated Than “Data Centers Raise Your Bill”
One of the most politically powerful claims in the data-center debate is that large computing facilities inevitably increase electricity prices for surrounding households.
The answer is more complicated.
A data center can create additional demand that requires new generation and transmission investment. Whether that translates into higher residential bills depends on utility regulation, rate design, who pays for infrastructure, the timing of electricity consumption, the availability of new generation and whether the data center directly funds some of the required capacity.
That distinction should be made explicit.
The strongest policy question is not simply:
“Do data centers use too much electricity?”
It is:
“Who pays for the marginal infrastructure required to serve them?”
That is a much more useful question for regulators and communities.
5. The “Dirty Electricity” Question Deserves Investigation—but Not Overstatement
One of the most interesting claims in the original article concerns power-quality distortion.
In December 2024, Bloomberg analyzed data from approximately 770,000 residential power-quality sensors and compared those measurements with the locations of nearly 1,500 U.S. data centers.
The analysis reported a strong geographic association between higher total harmonic distortion and proximity to significant data-center activity. Bloomberg also noted that multiple other factors—including industrial loads, solar generation and electric-vehicle infrastructure—can affect power quality.
Importantly, the analysis does not establish that data centers are the sole cause of the measured distortion, nor does it establish that such distortion is causing widespread human illness. One major utility cited in the investigation disputed aspects of the sensor methodology.
That means the scientifically responsible conclusion is:
The observed relationship is sufficiently interesting to justify independent engineering studies, standardized measurements and transparent utility data—but it should not be presented as proof of causation.
This is an important editorial correction.
Correlation is a reason to investigate, not permission to declare a mechanism proven.
6. Data Centers, Noise and Health: What We Actually Know
There is a stronger evidence base around environmental noise.
Large data centers contain extensive mechanical cooling systems, fans, pumps, transformers and backup-generation equipment. Depending on design and location, these can become persistent sources of environmental noise.
The World Health Organization identifies excessive environmental noise as a contributor to annoyance, sleep disturbance and increased risk of cardiovascular and other health effects.
Therefore, it is entirely reasonable to take persistent data-center noise seriously.
But a revised article should avoid another common mistake: assuming that every symptom reported by residents must have the same cause.
Reported experiences such as headaches, poor sleep, stress, fatigue, tinnitus or breathing irritation can have many causes. Noise, air pollution from backup generators, construction activity, anxiety, sleep disruption and other environmental factors can overlap.
That is why a community assessment should measure actual exposures:
- daytime and nighttime sound levels;
- low-frequency and tonal noise;
- vibration;
- air pollutants associated with backup generation;
- water use and wastewater impacts;
- grid voltage and harmonic distortion; and
- other site-specific environmental variables.
This approach is much stronger than attributing all symptoms to an unproven single mechanism.
7. Be Careful With Electromagnetic-Field Claims
The original article moved from data-center noise and power-quality concerns into broader claims regarding electromagnetic hypersensitivity.
That section needs substantial qualification.
There is ongoing scientific research into biological effects of electromagnetic fields, but the existence of electromagnetic fields around electrical infrastructure does not by itself establish that the fields are causing particular symptoms in nearby residents.
Likewise, evidence that a person experiences symptoms does not automatically establish the environmental mechanism responsible for those symptoms.
A credible public-health discussion should therefore separate:
1. Measured exposure — what electromagnetic fields, noise or pollutants are actually present?
2. Epidemiological association — are exposed populations demonstrably experiencing more disease?
3. Mechanistic evidence — is there a plausible biological mechanism supported by reproducible experiments?
4. Causality — does the total evidence support the conclusion that the exposure causes the health effect?
These are different questions.
That distinction strengthens the article rather than weakening it.
8. The AI Industry's First Response: Build More Generation
The industry's most obvious response is to provide more electricity.
That can take several forms:
- natural-gas generation;
- renewable energy;
- battery storage;
- grid expansion;
- existing nuclear generation;
- restarted nuclear plants;
- small modular reactors; and
- other advanced nuclear technologies.
This is already happening.
The U.S. nuclear industry has entered a very different phase from the prolonged stagnation of previous decades. In March 2026, the Nuclear Regulatory Commission authorized issuance of a construction permit for TerraPower's Kemmerer Power Station Unit 1 in Wyoming—the first commercial reactor construction approval from the NRC in roughly a decade and the first approval for a non-light-water reactor in more than 40 years.
The NRC has also introduced a new technology-inclusive licensing framework intended to provide more flexibility for advanced reactor deployment while retaining safety requirements.
This is significant.
The future of AI power may not be solved by one technology. The emerging model is likely to be a portfolio:
existing nuclear + new nuclear + renewables + storage + grid modernization + better data-center siting.
9. Thorium: Promising, Interesting—and Still Experimental
One of the most attractive ideas in the original article is thorium-based nuclear power.
There is genuine science behind the concept.
Oak Ridge National Laboratory operated the Molten Salt Reactor Experiment from 1965 to 1969. The reactor demonstrated molten-salt fuel technology and became the first reactor to operate on uranium-233.
Thorium can be converted through neutron capture into uranium-233, which can then serve as fissile fuel.
Molten-salt reactors are also fundamentally different from conventional light-water reactors because some designs use liquid fuel dissolved in circulating salt rather than solid fuel assemblies.
These characteristics have attracted decades of research.
But a crucial distinction must be made:
A successful 1960s experimental reactor is not the same thing as a commercially deployable modern power plant.
The Nuclear Regulatory Commission's review of the thorium fuel cycle makes clear that thorium introduces its own safety, fuel-cycle, materials and regulatory issues.
Modern molten-salt systems also face important engineering challenges involving corrosion, materials compatibility, fuel processing, tritium management, component lifetime, licensing and supply chains.
Therefore, the strongest argument for thorium is not:
“Thorium is the nuclear technology that will solve everything.”
It is:
“Thorium and molten-salt technologies deserve serious development because their characteristics could eventually provide another pathway to high-density, reliable electricity.”
10. Advanced Nuclear Is More Important Than Thorium Alone
This distinction is particularly important in 2026.
America does not need to wait for a hypothetical commercial thorium reactor before expanding nuclear capacity.
Advanced reactors using uranium fuel cycles are already progressing through licensing and construction.
TerraPower's Natrium project has received an NRC construction permit. Other advanced-reactor and small-modular-reactor projects are progressing through U.S. licensing pathways.
That leads to a better strategic framework:
| Technology | Current position | Potential role |
|---|---|---|
| Existing nuclear | Commercial today | Near-term reliable power |
| Advanced uranium reactors | Moving through licensing/deployment | 2020s–2030s expansion |
| Small modular reactors | Developing | Distributed industrial power |
| Molten-salt reactors | Developmental | Potential long-term high-temperature power |
| Thorium fuel cycle | Research/development | Potential long-term fuel option |
That is a much more credible nuclear strategy than presenting thorium alone as the answer.
11. The Second Possible Displacement: Move Data Centers Away From Communities
There is another solution that deserves considerably more attention.
Move the computing load.
Data centers do not necessarily need to be built next to large population centers.
The optimal location for a hyperscale computing campus may instead be a place with:
- abundant electricity;
- low population density;
- adequate transmission or dedicated generation;
- low cooling requirements;
- available industrial land;
- high-capacity fiber connectivity;
- minimal conflict with residential neighborhoods; and
- state and local governments willing to host the infrastructure.
This changes the debate from:
“Should America build AI infrastructure?”
to:
“Where should America build AI infrastructure?”
12. Alaska Is an Interesting Case Study
Alaska illustrates why geography may eventually become a competitive advantage in the data-center industry.
The state is cold, sparsely populated and rich in land and energy resources.
Cold temperatures can reduce cooling requirements, while low population density reduces the number of households directly exposed to the industrial footprint.
The state is already exploring the possibility of attracting major data-center projects, including proposals for large campuses on the North Slope. At the same time, federal facilities and military sites in Alaska have also been considered as potential locations for AI infrastructure.
Alaska is not a magic solution.
Its major weakness is connectivity.
Data centers need enormous quantities of high-speed fiber connectivity, and the Arctic environment creates logistical challenges for construction, maintenance and transmission.
That means Alaska demonstrates an important principle:
The best location for an AI data center may not be the place with the most customers nearby. It may be the place where energy, climate, land, water, infrastructure and political acceptance combine to create the lowest total cost.
13. The Orbital Data Center Idea
There is an even more radical idea: move computing into space.
This is no longer purely science fiction.
In January 2026, SpaceX submitted an FCC application for a proposed non-geostationary-orbit system of up to one million satellites, described by the company as an Orbital Data Center system. The FCC accepted the filing for consideration and comment.
Google is pursuing a related concept.
Its Project Suncatcher proposes interconnected solar-powered satellites carrying Google's AI accelerators. Google announced plans for an early-2027 learning mission involving prototype satellites.
The attraction is obvious.
In space:
- solar energy is continuously available in suitable orbits;
- large terrestrial land requirements disappear;
- terrestrial cooling-water consumption can be avoided; and
- computing infrastructure can potentially scale beyond the limitations of local grids.
But major challenges remain.
14. Why Space Probably Will Not Replace Terrestrial Data Centers Soon
The economics are still highly uncertain.
The hardest problems include:
- launch costs;
- satellite manufacturing at enormous scale;
- radiation effects on electronics;
- thermal management;
- communication bandwidth;
- orbital congestion and debris;
- maintenance and replacement;
- space-qualified hardware costs; and
- the enormous number of satellites required for meaningful compute capacity.
Google itself describes Suncatcher as a research moonshot rather than an immediately deployable commercial data-center replacement.
The appropriate conclusion is therefore not that orbital computing is impossible.
It is that the economic comparison has not yet been won.
For the foreseeable future, terrestrial computing is likely to remain dominant for most general-purpose applications.
Space could nevertheless become attractive for specialized workloads, space-native applications and possibly certain types of AI computation in the longer term.
15. A Better Strategy: Displace the Most Disruptive Data Centers
This brings us back to the displacement hypothesis.
The realistic goal is probably not:
“Stop data centers.”
The realistic goal is:
“Make the current way of building data centers economically inferior to a better alternative.”
Imagine a future in which the economics favor data centers that:
- generate or contract for dedicated power;
- pay the full marginal cost of new grid infrastructure;
- are built away from residential communities;
- use advanced cooling technology;
- minimize water consumption;
- operate with transparent environmental monitoring;
- use low-carbon, reliable electricity; and
- select locations based on total system cost rather than proximity to major cities alone.
At that point, communities would no longer need to defeat AI infrastructure politically.
The economics would naturally favor a less disruptive architecture.
16. What Governments Should Require
If governments want to minimize conflict between AI expansion and local communities, several principles make sense regardless of political ideology.
A. Full-cost infrastructure accounting
Developers should not be allowed to socialize the infrastructure costs created primarily by their projects while privatizing the resulting economic benefits.
B. Transparent electricity contracts
Utilities and regulators should disclose how major data-center loads affect generation, transmission investment and rate structures.
C. Independent power-quality monitoring
Where claims of unusual harmonic distortion exist, measurements should be conducted by independent engineers using standardized instrumentation rather than relying solely on either industry or advocacy-group measurements.
D. Independent noise monitoring
Noise should be measured at property boundaries and nearby residences under realistic operating conditions, including overnight periods.
E. Better zoning
Hyperscale data centers should generally be treated as major industrial infrastructure, not simply as ordinary commercial development.
F. Dedicated generation where appropriate
Large computing projects should increasingly be expected to demonstrate how they will provide or contract for sufficient generation and grid capacity without creating unacceptable reliability or affordability impacts on existing users.
17. The Bigger Opportunity: AI Could Accelerate America's Energy Transition
There is an irony in the data-center crisis.
AI is increasing electricity demand at exactly the moment when America needs to modernize an aging energy system.
That creates an enormous economic incentive.
If technology companies are willing to pay premium prices for reliable electricity, they may unintentionally provide something the nuclear and energy industries have lacked for decades:
a large customer willing to finance new power infrastructure.
That could accelerate development of:
- advanced nuclear reactors;
- small modular reactors;
- grid-scale storage;
- long-distance transmission;
- geothermal energy;
- renewables combined with firm generation;
- advanced cooling;
- microgrids; and
- more efficient computing hardware.
The key is ensuring that the economic incentive is directed toward infrastructure that creates broad public value rather than merely enabling faster construction at the lowest private cost.
18. The Data Center Crisis May Be an Infrastructure Design Problem
It is tempting to view the current conflict as a battle between environmentalists and the technology industry, or between communities and economic development.
That framing is too simplistic.
The deeper problem is that America's AI infrastructure is being designed in real time.
There is no reason to assume that today's architecture will remain the optimal architecture in 2030 or 2040.
Hardware changes.
Cooling changes.
Power generation changes.
Chip efficiency changes.
Networking changes.
AI models themselves change.
The location of computing can change too.
That means today's controversial data centers should not automatically become tomorrow's permanent infrastructure model.
19. The Real “Law of Displacement”
Perhaps the most useful interpretation of displacement is therefore broader than the historical examples usually offered.
Society does not necessarily need to choose between:
technology and the environment.
It can sometimes choose between two different technological architectures.
The first may be:
high-density urban data centers + residential-grid dependence + transmission expansion + extensive cooling + community conflict.
The second might eventually become:
remote computing campuses + dedicated energy + advanced nuclear + renewables + storage + efficient cooling + low-density sites + transparent environmental monitoring.
If the second architecture becomes cheaper, more reliable and easier to finance, the first does not need to be defeated politically.
It simply becomes obsolete.
20. What the Next Decade Could Look Like
| Period | Likely development |
|---|---|
| 2026–2028 | Rapid terrestrial data-center construction, grid upgrades, gas generation, nuclear restarts and accelerated power procurement. |
| 2028–2032 | Greater deployment of advanced reactors, data-center microgrids, dedicated generation and remote industrial campuses. |
| 2030s | Potential commercial deployment of additional advanced reactor technologies, alongside increasingly efficient AI hardware and cooling systems. |
| 2030s+ | Orbital computing could move from experimental demonstrations toward specialized commercial applications if launch, thermal and manufacturing economics improve. |
These are scenarios rather than predictions. Technology deployment is highly sensitive to financing, regulation, manufacturing capacity, electricity prices and unexpected engineering breakthroughs.
Conclusion: Don't Just Fight the Data Center Boom—Change Its Economics
The explosive growth of artificial intelligence is creating a genuine infrastructure challenge.
America needs more computing capacity, but it also needs reliable electricity, affordable power, resilient grids and communities that retain a meaningful voice over major industrial development.
Simply declaring that data centers are good or bad does not solve that problem.
Nor does assuming that protest alone will stop an industry backed by enormous amounts of capital.
A more productive strategy is to ask a different question:
What technology, location or energy architecture could provide the same AI computing capacity with lower total economic and social costs?
That question opens several pathways.
Advanced nuclear could provide reliable high-density electricity.
Remote locations could reduce conflicts with residential communities.
Improved cooling could reduce water and electricity requirements.
Dedicated generation could reduce pressure on local grids.
Better power-quality monitoring could distinguish legitimate engineering problems from unsupported claims.
And eventually, space-based computing could provide another architecture for selected applications if its economics mature.
The most important idea is therefore not that thorium will solve the data-center crisis, or that Alaska will become America's AI capital, or that orbital satellites will replace terrestrial data centers.
Those remain hypotheses.
The larger proposition is simpler:
The best way to reduce the negative externalities of a rapidly expanding technology may be to make a better alternative more profitable.
That is the useful part of the “law of displacement.”
Not a law of physics.
Not an inevitable historical rule.
But potentially a powerful way to think about how technological systems evolve.
Frequently Asked Questions
What is the law of displacement?
In this article, “law of displacement” is a proposed market hypothesis, not a recognized scientific or economic law. It describes the idea that entrenched technologies may be displaced more effectively by commercially competitive alternatives than by opposition alone.
Are data centers causing higher electricity prices?
Large data centers can create substantial new electricity and infrastructure requirements, but the effect on household electricity prices depends on utility regulation, rate design, generation availability, transmission investment and how much of the incremental infrastructure is paid for by the data-center customer.
Is the “dirty electricity” problem proven?
No. Bloomberg's analysis of approximately 770,000 residential sensors found a strong association between power-quality distortion and proximity to major data centers, but the analysis was observational, other sources can affect harmonics, and utilities challenged aspects of the methodology. Independent engineering measurements are needed before drawing causal conclusions.
Can data-center noise affect health?
Persistent environmental noise can affect sleep, annoyance and cardiovascular health. The World Health Organization recognizes environmental noise as an important public-health issue. However, individual symptoms near a data center should not automatically be attributed to one environmental cause.
Are data centers causing electromagnetic hypersensitivity?
The available evidence does not justify claiming that nearby data centers cause electromagnetic hypersensitivity. Measurements of electromagnetic fields, power quality, noise and other exposures should be separated from claims about causation.
Will thorium reactors solve America's data-center power problem?
Not in the near term. Thorium and molten-salt reactors are promising areas of nuclear research, and Oak Ridge demonstrated molten-salt reactor technology in the 1960s. However, modern commercial thorium systems still require substantial development, engineering validation, fuel-cycle infrastructure and regulatory approval.
What nuclear technology is closest to helping with future data-center demand?
Existing nuclear plants, nuclear restarts, advanced uranium-fueled reactors and some small modular reactor programs are closer to deployment than commercial thorium reactors. The U.S. NRC's 2026 approval of TerraPower's Kemmerer construction permit illustrates how advanced nuclear deployment has begun to move from research toward commercial construction.
Could data centers be built in Alaska?
Alaska has several characteristics that are attractive for large computing campuses, including cold temperatures, abundant land and low population density. Its main challenges include connectivity, logistics, infrastructure and the cost of building and maintaining facilities in remote environments.
Could AI data centers eventually be built in space?
Possibly for selected applications, but this remains experimental. SpaceX has filed an FCC application involving up to one million proposed orbital data-center satellites, while Google's Project Suncatcher is testing solar-powered satellite computing. The economics, thermal management, communications, radiation environment and manufacturing requirements remain major challenges.
What is the most realistic solution to the data-center crisis?
There probably will not be one solution. A combination of better siting, dedicated generation, grid investment, advanced nuclear power, renewable energy, storage, more efficient AI hardware, improved cooling and transparent cost allocation is more plausible than any single technology replacing the entire terrestrial data-center industry.
Evidence & Source Notes
Primary/authoritative sources used in this update:
- A Midwestern Doctor - How the Law of Displacement Can Solve the Data Center Crisis (2026)
- Lawrence Berkeley National Laboratory — United States Data Center Energy Usage Report: 2025 Update
- U.S. Energy Information Administration
- World Health Organization — Guidance on Environmental Noise
- U.S. Nuclear Regulatory Commission — Advanced Reactor Highlights, 2026
- NRC — TerraPower Kemmerer Power Station
- Oak Ridge National Laboratory — Molten Salt Reactor History
- NRC — Safety and Regulatory Issues of the Thorium Fuel Cycle
- Federal Communications Commission — SpaceX Orbital Data Center application notice
- Google — Project Suncatcher
- Bloomberg — AI data centers and power-quality distortion
.png)
Comments