Voters Just Rebuked AI Data Centres — And Why Some Say the Answer Is in Orbit

The artificial intelligence revolution has a growing problem that cannot be solved simply by writing better algorithms: it needs somewhere to live.

Every chatbot response, generated image, AI-powered search and increasingly sophisticated digital assistant depends on physical infrastructure. Behind the apparently invisible world of AI are enormous buildings filled with servers, electrical equipment, cooling systems and miles of network cables.

These facilities need land. They need electricity. Many need substantial quantities of water. And they increasingly compete with the communities surrounding them for resources that residents already depend on.

On November 3, 2026, that conflict became a political story across the United States. Local voters confronted proposals involving data-centre construction, restrictions, moratoriums and changes to local development rules. The results represented an unusually strong rejection of unchecked expansion.

A November 5 report by Seven Lights described the result as a 44-to-1 rebuke of AI data-centre development. Other election trackers published different totals because they counted different measures and updated results at different times. The broader message, however, was remarkably consistent: communities want a greater say in what gets built around them.

And an unexpected alternative is attracting attention.

What if some of the computers powering artificial intelligence did not need to be built in towns, industrial parks or rural communities at all? What if future computing infrastructure could operate hundreds of kilometres above Earth, using solar energy in orbit?

The idea sounds like science fiction. Yet space-based computing has moved beyond a purely speculative discussion. Google's Project Suncatcher has already entered its first orbital testing phase, and the US Government Accountability Office has examined the technology's opportunities and engineering limitations.

The question is no longer simply whether humanity can build computers in space. It is whether doing so could eventually make economic, environmental and technical sense.

1. The 44-to-1 Message: AI Infrastructure Has Become a Local Political Issue

For years, the expansion of data centres was largely discussed by technology companies, investors, energy planners and government officials. The facilities were presented as necessary infrastructure for a digital economy, much like power stations, telecommunications networks and industrial warehouses.

That conversation has changed.

During the November 3, 2026, US midterm election cycle, residents in communities across the country were asked to make decisions about data-centre development. The questions were not all identical. Some concerned outright construction bans, others addressed temporary moratoriums, electricity consumption limits, zoning decisions or approval agreements with developers.

The 44-to-1 result cited in the November 5 Seven Lights coverage captures the extraordinary imbalance between opposition and support in one reported tally. It is important, however, to understand that the number represents local ballot questions rather than a single nationwide referendum on artificial intelligence.

Election counts also evolved as additional results were reported. Ballotpedia's November 4 accounting identified 47 decided measures, with 46 favouring greater restrictions and only one favouring development. The Data Centers on the Ballot tracker, maintained by Redbud Consulting, also explains how it groups differently worded questions into two broad outcomes: Restrict and Allow.

These differences matter when interpreting the precise headline number. The Ballotpedia election results and contemporary reporting from KCRA and WDSU illustrate how the published totals varied.

Nevertheless, the political direction is difficult to misunderstand.

Communities are increasingly unwilling to accept large industrial computing facilities simply because technology companies describe them as essential to the future.

The opposition is also more complicated than a general dislike of artificial intelligence. Residents may use AI tools every day while opposing a particular facility near their homes. They may support technological progress but question why their electricity bills, water supplies, roads and local environment should absorb the costs of private infrastructure.

This distinction is central to the story.

A person can believe that artificial intelligence has enormous potential and still ask whether a proposed data centre is being built in the right place, whether its developer is paying its fair share and whether residents have been properly consulted.

The ballot results suggest that these questions are becoming politically powerful.

2. What Exactly Are Voters Rejecting?

A data centre is a physical building or collection of buildings containing servers, storage systems, networking equipment and supporting infrastructure.

Think of it as an enormous library, except that instead of storing books on shelves, it stores digital information and performs calculations. Instead of readers walking through its corridors, millions of computers and applications request information from it.

Traditional data centres already supported websites, financial systems, online shopping, video streaming, cloud storage and business software.

Artificial intelligence has changed the scale and intensity of the computing required.

Training a sophisticated AI model can involve thousands of specialised processors working together for extended periods. Once trained, the model must also be operated. Every request to an AI assistant, every generated image and every complex calculation requires computing resources.

This creates demand for facilities designed around high-performance processors and increasingly dense computing equipment.

The result is a new generation of hyperscale facilities.

A hyperscale data centre is not simply an office building with a server room. It may occupy a large industrial site, contain multiple data halls, require dedicated electricity infrastructure and use extensive cooling equipment.

The ballot questions in 2026 addressed different aspects of this development.

For example:

• Some communities voted on whether large data centres should be prohibited.
• Others considered temporary pauses while local governments studied their environmental and infrastructure impacts.
• Some questions focused on facilities exceeding particular electricity-demand thresholds.
• Others addressed land rezoning or development agreements.
• Certain measures sought public approval before particularly large projects could proceed.
• Some advisory questions asked whether local officials should support or oppose data-centre construction.

This variety is important because a vote against one project does not necessarily mean that a community opposes every form of computing infrastructure.

It may mean that residents object to the scale, location, approval process or financial arrangement associated with a specific proposal.

For example, a town might welcome a smaller facility that uses reclaimed water, finances its own electrical upgrades and creates permanent technical jobs. The same town might reject a much larger project that requires substantial new infrastructure without offering enforceable guarantees.

The political argument is therefore increasingly about conditions, not simply technology.

The underlying question is becoming:

Who should carry the physical cost of the AI revolution, and who should receive its economic benefits?

3. The Land Problem: AI Needs More Physical Space Than Its Digital Image Suggests

The first difficulty is land.

Digital services create the impression that computing happens nowhere in particular. A person opens a browser, asks a question and receives an answer. The process appears almost weightless.

But the physical infrastructure behind that experience occupies real territory.

Large data-centre campuses require space for:

• Server buildings and equipment halls.
• Electrical substations and transformers.
• Backup generators.
• Cooling towers, chillers and radiator equipment.
• Fuel storage and maintenance facilities.
• Roads, security areas and utility connections.
• Expansion space for additional computing capacity.

A large facility can transform the character of a rural or suburban area.

Land that was previously agricultural, residential, wooded or undeveloped may become an industrial campus. Even when a developer purchases land legally and obtains the necessary permits, residents can reasonably question whether the transformation is compatible with the community's long-term plans.

There is also a difference between the land occupied by the building and the wider infrastructure needed to support it.

A facility may require new transmission lines, electrical substations, water connections and road improvements. These supporting systems can affect a much larger area than the server buildings themselves.

The issue is particularly sensitive when development occurs near residential neighbourhoods or agricultural land.

Residents may worry about losing green space, changing the visual character of their surroundings, construction traffic and the possibility that additional industrial development will follow the initial project.

There is another consideration: data centres are often built with future expansion in mind.

A company may initially announce one phase, while the overall site plan allows several additional buildings. Communities therefore need to understand not only the first construction phase but also the maximum potential development.

This is why some 2026 ballot questions addressed land area, rezoning and voter approval rather than simply asking whether AI should be supported.

The issue is not whether digital infrastructure needs buildings. It obviously does.

The issue is whether the land selected for those buildings represents the best use of a community's limited resources.

4. The Electricity Problem: AI Is Becoming a Major Industrial Power Customer

The most significant infrastructure challenge may be electricity.

A conventional office building consumes power for lighting, computers, air conditioning and other everyday activities. A large AI data centre can operate at an entirely different scale.

Its servers may run continuously, while cooling equipment, electrical conversion systems and backup infrastructure add further demand.

A large facility can therefore behave more like an industrial power customer than an ordinary commercial building.

The numbers illustrate why energy planners are paying attention.

The International Energy Agency's Energy and AI report estimated that data centres consumed approximately 415 terawatt-hours of electricity worldwide in 2024. Its base-case projection indicated that global data-centre electricity consumption could rise to approximately 945 TWh by 2030.

That is an enormous increase in demand over a relatively short period.

The US picture is also significant.

Lawrence Berkeley National Laboratory's June 2026 United States Data Center Energy Usage Report: 2025 Update estimated that data centres could account for 11.8% of total US electricity consumption by 2030.

The report provided a wider range of scenarios, from 9.5% to 15.3%, depending on assumptions about equipment installations, specialised AI chips, chip lifetimes and server utilisation.

Its reference-case estimate was 649 TWh of electricity consumption in 2030. The broader uncertainty range was 521–843 TWh.

These are projections, not guaranteed outcomes. Improvements in processor efficiency, cooling systems and AI software could change the trajectory.

However, the potential demand is large enough to affect electricity planning today.

Why Does This Create a Problem for Local Communities?

Electricity is not simply a resource that appears when a company requests it.

Power must be generated, transmitted and distributed through a network of infrastructure.

If a utility has sufficient spare capacity, a new large customer may help it use existing infrastructure more efficiently.

If capacity is insufficient, however, the utility may need to invest in:

• New substations.
• Additional transmission lines.
• Transformer replacements.
• Grid reinforcement.
• New generation capacity.
• Additional reserve capacity to maintain reliability.

These investments cost money.

The important question is who pays.

A June 2026 working paper from researchers associated with MIT's Center for Energy and Environmental Policy Research examined US data-centre entry between 2010 and 2024.

The researchers estimated that data-centre entry raised average retail electricity rates by 2.7% and residential retail rates by 2.1%, with larger effects in some utility systems.

This is evidence of a relationship across the period and locations studied, not proof that every new data centre automatically increases every household's electricity bill.

The Congressional Research Service also highlighted the complexity of the issue in its May 2026 report on data-centre energy consumption.

It noted that a separate Lawrence Berkeley National Laboratory analysis of electricity-price changes between 2019 and 2025 identified grid investment, ageing infrastructure, resilience needs, natural-gas prices and other factors as major influences. Data centres were not identified as major price drivers in most areas during that period.

In some states, increased demand helped utilities spread existing costs across more electricity sales.

Both findings can be true.

The effect depends on the local electricity market, existing infrastructure, how much additional capacity is needed and the rules governing cost recovery.

This is why electricity-rate protections have become an important part of the political debate.

Residents do not necessarily oppose a company purchasing large quantities of electricity. They want assurances that ordinary households will not be left paying for infrastructure primarily built to serve a private technology business.

The Renewable-Energy Question

Technology companies increasingly announce renewable-energy agreements and investments in clean electricity.

These commitments can be valuable, but the relationship between a data centre and a renewable-energy project needs careful examination.

Buying renewable-energy certificates is not always the same as physically supplying a facility with clean electricity every hour.

Solar panels produce electricity when sunlight is available. Wind turbines depend on wind conditions. A data centre, meanwhile, may need electricity continuously.

The IEA's analysis indicates that renewables are expected to meet nearly half of additional global data-centre electricity demand through 2030. It also projects that natural gas and coal together will supply more than 40% of the additional demand in that period.

The actual energy mix varies between countries and electricity systems.

The real challenge is not merely purchasing green electricity. It is building an electricity system that can reliably deliver sufficient power without transferring unacceptable costs or environmental consequences to the surrounding population.

5. The Water Problem: Keeping Powerful Computers Cool

Computers produce heat.

A processor performing millions or billions of operations generates thermal energy. When thousands of powerful processors operate together in a confined building, heat management becomes an essential part of the infrastructure.

Imagine putting thousands of powerful electric heaters inside one enormous room and asking them to operate continuously without overheating.

That is an imperfect but useful way to understand the cooling challenge.

Data centres use different combinations of air cooling, liquid cooling, chillers, heat exchangers and other technologies.

Some facilities use water-based cooling systems that transfer heat away from computing equipment. Evaporative cooling can be particularly effective because water absorbs heat as it evaporates.

But evaporated water must be replaced.

The amount consumed depends on several factors:

• The size and computing density of the facility.
• The local climate.
• The type of cooling equipment.
• The quality and source of the water.
• Whether the system is open-loop, closed-loop or hybrid.
• How frequently water can be reused before it must be discharged.

The Congressional Research Service's July 2026 report, Data Centers and Water: Frequently Asked Questions, provides a useful illustration.

Citing International Energy Agency estimates, it explains that a 100-megawatt US data centre may directly consume approximately as much water each day as 2,600 households.

The IEA's broader estimate, including indirect water consumption associated with electricity generation, is equivalent to approximately 6,500 households. These figures are illustrative rather than universal measurements for every facility.

The report also notes that municipal water systems supplied approximately 97% of US data-centre onsite water needs in 2025, based on the research it reviewed.

The distinction between direct and indirect water consumption matters.

A data centre might reduce the amount of water used inside its own cooling system while still relying on electricity generated by power stations that consume water elsewhere.

A complete assessment should therefore examine both the facility and the electricity supply.

Why Water Becomes a Community Issue

Water availability varies enormously between locations.

A data centre built in a region with abundant water and modern recycling infrastructure may have a manageable impact.

The same facility in a drought-prone area could create a much more difficult situation.

The Congressional Research Service estimated that municipal water systems supplied approximately 97% of US data-centre onsite water needs in 2025, based on the research it reviewed.

It also explained that direct water use represents a relatively small proportion of overall US water consumption, while acknowledging that local impacts can be significant because facilities concentrate demand geographically.

Consider two communities.

The first has a reliable water supply, a modern wastewater treatment system and access to reclaimed industrial water.

The second experiences periodic drought, has limited groundwater resources and already faces competition between households, agriculture and industry.

It would be misleading to assume that a data centre creates the same water problem in both locations.

But it would be equally misleading to claim that national water statistics make local concerns irrelevant.

For a household, the important issue is not the national percentage. It is whether sufficient clean water remains available when the community needs it.

Can Better Cooling Solve the Problem?

There are promising options.

Closed-loop cooling systems can circulate liquid through equipment without continuously consuming the same volume of fresh water as an evaporative system.

Reclaimed wastewater can replace some potable-water demand where suitable treatment and distribution infrastructure exist.

More efficient chips can also reduce the amount of heat produced for a given computing task.

However, every cooling technology involves trade-offs.

A system that saves water may require more electricity. Another may need expensive equipment or specialised maintenance.

The best approach depends on the local environment and the design of the facility.

The public should be able to see those calculations before a major project is approved.

6. Noise, Generators and the Feeling of Losing Control

Electricity and water are measurable. Community experience is more difficult to reduce to a spreadsheet.

A large data centre can create several visible and audible changes in its surroundings.

Cooling equipment may generate a continuous mechanical hum. Large fans, pumps, transformers and other machinery can operate around the clock.

Backup generators may be tested periodically. During an electricity interruption, they may operate for longer periods.

Construction can bring heavy vehicles, temporary road closures, dust and noise.

Depending on the location, residents may also worry about lighting, visual impacts and the presence of industrial infrastructure close to their homes.

These effects are not identical at every site. Modern acoustic engineering, appropriate setbacks, landscaping and carefully designed generator systems can reduce some problems.

Nevertheless, they help explain why a facility that looks harmless on a company presentation may appear very different to someone living next door.

There is also a question of trust.

Residents sometimes discover that important development discussions have already taken place before they receive meaningful information about the proposal.

They may learn about the size of the project, expected electricity demand, water arrangements or tax incentives only after key decisions have been made.

A September 2026 Urban Institute analysis described these difficulties as part of the growing national backlash.

It cited concerns about electricity costs, water use, noise pollution, tax exemptions and the broader implications of AI. Its research discussion also highlighted the importance of meaningful public involvement, enforceable community benefits and transparent negotiations.

The article reported that 75% of Americans did not want a data centre built in their area, according to the survey it cited.

That figure should not be interpreted as a rejection of all data centres or all AI applications.

It does, however, suggest that the industry has a serious public-acceptance challenge.

A company can purchase land, secure permits and negotiate an electricity contract. It cannot automatically purchase the confidence of the people who will live beside the facility.

7. The Economic Argument: Data Centres Also Bring Benefits

A balanced discussion must recognise why local governments have welcomed these projects.

Data centres can bring substantial private investment.

Construction requires contractors, engineers, electricians, equipment suppliers, transport services and other workers.

Operating facilities need technicians, security personnel, network specialists, electrical engineers and maintenance teams.

They may also contribute property-tax revenue and support local infrastructure development.

For some communities, attracting a major technology project appears to offer a route towards economic diversification.

The problem is that the benefits and costs are not always distributed evenly.

Construction employment can be significant but temporary. Permanent staffing levels depend on the size and operational model of the facility.

Tax incentives can also reduce the public revenue that residents expect to receive.

The Urban Institute notes that more than two-thirds of US states have offered significant tax incentives to encourage data-centre development over the past 25 years.

Such incentives may reduce, forgive or defer corporate, property or sales taxes.

That creates an obvious question: if a community provides land, infrastructure and public incentives, what does it actually receive in return?

A good agreement should not rely on attractive promises alone.

It should specify measurable commitments, including:

• Who pays for electricity infrastructure.
• Who finances water-system improvements.
• How much potable water can be used.
• What noise limits apply.
• How many permanent jobs are expected.
• What tax payments and incentives are involved.
• What happens if the developer fails to meet its commitments.
• How residents can monitor compliance.

In other words, communities are not necessarily demanding that the AI industry stop building.

They are demanding a better deal.

8. Why Some People Believe the Answer Could Be in Orbit

This is where the story takes an unexpected turn.

If data centres are creating pressure on land, water and electricity infrastructure, perhaps some computing should take place somewhere other than Earth.

Space-based computing proposes placing computer processors, storage equipment and networking systems aboard satellites.

Instead of building a huge server facility on agricultural land, companies could eventually deploy networks of orbital computing platforms.

The attraction is straightforward.

In carefully selected orbits, satellites can receive sunlight for extended periods, potentially generating electricity through solar panels without depending on a terrestrial power grid.

They would not need to occupy valuable residential or agricultural land.

They would not draw drinking water from municipal supplies for conventional cooling towers.

And they could process certain types of information before sending results back to Earth.

This does not mean that putting computers into space automatically solves every environmental problem. It means that the physical constraints become different.

The US Government Accountability Office examined this possibility in its April 28, 2026, report, Science & Tech Spotlight: Data Centers in Space.

The report identified possible reductions in terrestrial land, electricity and water requirements while emphasising the substantial engineering and economic barriers that remain.

It also noted that smaller facilities processing information generated by satellites may be closer to practical deployment than enormous orbital systems designed to train advanced AI models.

The Important Development After Project Suncatcher

The Jornal has already explored Google's Project Suncatcher and the broader possibility of moving AI computing into orbit.

The next question is what happens after the initial idea.

On October 1, 2026, Google announced that its Project Suncatcher prototype satellite had successfully launched aboard SpaceX's Transporter-18 rideshare mission.

Google confirmed contact with the satellite and said it was operating as expected.

On October 30, Google reported another important milestone: the satellite had successfully returned its first AI-generated response from an onboard machine-learning model.

The company described the achievement as an early research milestone. The next steps involve testing how its Tensor Processing Units perform under thermal stress and radiation exposure and progressing towards tests involving two interconnected satellites.

That distinction is important.

A successful prototype launch and an initial AI response prove that an experimental computing system can operate in orbit. They do not prove that thousands of satellites can economically perform the computing tasks of a modern hyperscale facility.

The next phase involves testing the technology, understanding its limitations and determining which workloads might justify the additional complexity.

The GAO report indicates that some proposed deployments of larger data-centre satellites are being considered for the mid-2030s.

The technology is moving forward, but it remains developmental.

9. What Would Orbital Computing Actually Solve?

The strongest argument for space-based computing is not necessarily that every AI server should leave Earth.

It is that certain types of computing may be better suited to orbit.

Consider the amount of information produced by satellites.

Earth-observation spacecraft collect enormous quantities of imagery and measurements. Weather satellites monitor atmospheric conditions. Scientific missions observe distant objects and record complex datasets.

Sending every piece of raw information to Earth before analysing it can create delays and require considerable communication capacity.

An orbital computer could examine information closer to where it is collected.

For example, an Earth-observation satellite might capture thousands of images of a region. An onboard computing system could identify cloud-free images, detect changes in vegetation or flag possible flooding.

Instead of transmitting every original image immediately, the system could prioritise the most relevant information.

That could reduce communication requirements and accelerate certain decisions.

This is one of the more practical early applications identified by the GAO.

There are other possible uses.

Scientific processing: Space telescopes and research missions could analyse their own observations before transmitting selected results.

Autonomous spacecraft: Satellites could process information and make certain operational decisions without waiting for instructions from a distant ground station.

Specialised AI workloads: Some computational tasks may be worth performing in orbit if they benefit from uninterrupted solar access or proximity to space-generated data.

Distributed computing: A network of satellites could potentially divide large tasks among multiple platforms and exchange information through optical communication links.

The important word is potentially.

A system designed to process satellite imagery does not automatically become suitable for hosting the world's largest AI models.

Different workloads have different requirements for computing capacity, memory, connectivity, reliability and response time.

The future of orbital computing will probably depend on matching the right task to the right environment.

10. The Biggest Engineering Obstacle: Space Is Not a Giant Refrigerator

There is an appealing assumption behind the idea of space-based computing: space is extremely cold, so cooling powerful computers should be easy.

Unfortunately, this is a misunderstanding of how heat behaves in a vacuum.

On Earth, cooling systems can transfer heat into surrounding air or water. Fans move air across hot components, liquid loops carry thermal energy away, and cooling towers can release heat into the atmosphere.

Space does not provide that surrounding atmosphere.

In a near-vacuum, convection is essentially unavailable. A spacecraft cannot simply blow hot air into its surroundings and expect the heat to disappear.

Instead, it must transfer heat through engineered systems and ultimately release it through radiation.

This requires radiators: specialised surfaces that emit thermal energy into space.

The larger the computing system, the more heat it may need to remove. Consequently, the radiator system can become enormous.

Boston Consulting Group's August 2026 analysis of orbital data centres illustrates the scale of the problem.

Its modelling suggests that a satellite supporting 100 kilowatts of computing capacity could require a radiator area of approximately 400 square metres under the assumptions used in its analysis.

That is a substantial amount of equipment to manufacture, transport and deploy in orbit.

The analysis discusses lightweight radiator designs and advanced two-phase cooling loops as possible ways to reduce the mass and surface area required.

These are engineering opportunities, not proof that the cooling problem has already been solved.

The Sunlight Paradox

Space offers access to abundant solar energy, but solar panels alone do not guarantee uninterrupted electricity.

A satellite in low Earth orbit may pass through Earth's shadow, depending on its orbital geometry.

During those periods, solar generation is interrupted and the spacecraft must rely on stored electricity or another power-management strategy.

Some carefully selected sun-synchronous orbits can provide near-continuous sunlight, which is one reason they attract interest for orbital computing.

But these orbits involve specific orbital constraints, and designing a large, reliable computing constellation around them is not straightforward.

The energy system must support more than the processors themselves.

It must also power communications equipment, thermal-control systems, onboard computers and other spacecraft components.

A large orbital data centre could require solar arrays considerably larger than those previously deployed and assembled in space.

That is one of the challenges highlighted by the GAO.

Radiation Can Damage the Computers

Earth's atmosphere and magnetic environment provide important protection against some forms of space radiation.

Satellites do not enjoy the same protection.

High-energy particles can affect electronic components, cause unexpected changes in stored information and gradually degrade hardware.

A computer operating in orbit therefore needs more than ordinary data-centre reliability engineering.

Its designers may require:

• Radiation-tolerant processors.
• Additional shielding.
• Error detection and correction.
• Redundant computing systems.
• Automatic recovery from hardware faults.
• Software designed to detect corrupted calculations.

This is particularly important for AI workloads.

A corrupted calculation could produce an incorrect result without necessarily causing the entire system to shut down.

An orbital AI facility would need to distinguish between a normal computational error, a software problem and a radiation-induced hardware fault.

The GAO identifies radiation damage as one of the central technical concerns. It also notes that protective measures can add cost or affect computing performance.

Communications: The Computer Still Needs to Talk to Earth

A data centre on Earth can connect to a vast network of fibre-optic cables.

These cables provide extremely high-capacity connections between buildings, cities and countries.

An orbital data centre would need a different communications architecture.

It might use radio-frequency systems for certain connections and optical laser links to communicate between satellites.

Laser communication is attractive because it can potentially provide high data rates without depending on conventional radio-frequency bandwidth in the same way.

However, the system must maintain accurate pointing between moving spacecraft.

Communication with Earth introduces another challenge.

An optical link must reach a ground station, and atmospheric conditions can interfere with optical transmission. Cloud cover, weather and geographical coverage all matter.

A satellite network could use multiple ground stations and intersatellite connections to improve availability, but this adds infrastructure and complexity.

There is also latency.

Low Earth orbit is relatively close to the surface compared with geostationary orbit, but communication is not instantaneous. Signals still have to travel between satellites, ground stations and the users requesting information.

For some workloads, that is acceptable.

For others, particularly applications requiring very rapid responses from a nearby user, terrestrial infrastructure remains more convenient.

Maintenance Is Another Expensive Problem

When a server fails inside an Earth-based data centre, technicians can often replace it.

They can inspect a cooling system, install new equipment, change a power supply or repair a network connection.

A satellite cannot be serviced so easily.

A component failure could mean losing an entire computing platform, depending on the design and available servicing capabilities.

Robotic spacecraft servicing is developing, but routine maintenance of large orbital computing systems has not yet become an established commercial service.

This changes the economics.

A terrestrial facility can replace equipment through normal supply chains. An orbital facility may require launching replacement hardware, moving it into the correct orbit and integrating it with the existing network.

The cost of reliability therefore becomes part of the business model.

11. The Financial Reality: Orbital Data Centres Are Not Cheaper Yet

The economic argument is just as important as the engineering argument.

Launching computers into orbit requires rockets, spacecraft structures, solar panels, cooling equipment, radiation protection, communications hardware and extensive testing.

All of that equipment must be designed to survive launch and operate in an environment where repair is difficult.

The result is a substantial upfront investment.

Boston Consulting Group's 2026 study compared the projected costs of terrestrial and space-based AI data centres.

Its model estimated that orbital facilities could carry a cost premium of approximately 2.5 to 3 times over terrestrial infrastructure under current assumptions.

Even under a realistic improvement scenario, BCG estimated that the premium could remain around 1.5 times over the following decade.

These are modelled economic scenarios, not guaranteed market prices or independently established forecasts. They nevertheless demonstrate why orbital computing should not currently be presented as a cheaper replacement for ordinary data centres.

The cost comparison also reveals something interesting.

A terrestrial facility spends substantial money on electricity, cooling, maintenance and physical infrastructure throughout its lifetime.

An orbital facility could potentially reduce some of those operating expenses, particularly if solar electricity becomes sufficiently practical.

However, it carries much higher initial deployment costs.

The question is whether lower operating costs over many years can compensate for the expense of manufacturing and launching the equipment.

That calculation depends on launch prices, hardware lifetime, satellite reliability, cooling efficiency and the amount of computing capacity each spacecraft can provide.

It also depends on the price of electricity on Earth.

A terrestrial data centre located near affordable, reliable and low-carbon electricity may remain far more economical than an orbital competitor.

In other words, space computing does not automatically win simply because sunlight is available.

It must deliver enough useful computing capacity to justify the complete cost of getting there.

12. Space Has Environmental Costs Too

Moving infrastructure away from communities does not make its environmental footprint disappear.

It changes where that footprint occurs.

An orbital computing network would require launches, spacecraft manufacturing and eventual disposal or replacement of satellites.

Rocket launches emit gases and particles into the atmosphere. Researchers are studying how increasing launch activity and satellite reentries may affect atmospheric chemistry and climate.

A 2024 study published in Scientific Data developed a global inventory of rocket-launch and reentry emissions associated with the growing satellite constellation industry.

It documented emissions including carbon dioxide, black carbon and other pollutants. Its findings underline why the environmental consequences of large satellite networks need to be considered alongside their potential benefits.

There is also the increasingly serious issue of orbital debris.

Satellites travel at very high speeds. Even relatively small fragments can damage operational spacecraft.

If a satellite fails and remains in a busy orbit, it can become another object that other spacecraft must avoid.

A collision could create additional fragments, potentially increasing the risk of further collisions.

This is commonly associated with the concept of the Kessler syndrome: a situation in which collisions generate debris that makes additional collisions more likely.

The European Space Agency's September 2026 Space Environment Report warned that the orbital environment is becoming increasingly congested.

It reported that more than 4,000 new payloads were placed into orbit during 2025 and highlighted growing collision risks and the need for better end-of-life disposal and active debris removal.

Adding thousands of AI computing satellites to an already crowded orbital environment would therefore require serious planning.

A responsible system would need:

• Reliable collision-avoidance capabilities.
• Clear orbital coordination.
• End-of-life disposal procedures.
• Effective satellite tracking.
• Long operational lifetimes where possible.
• Repair, refuelling or replacement strategies.
• International cooperation on space traffic management.

There is an important environmental principle here.

Humanity should not solve one infrastructure problem by creating another that future generations will struggle to manage.

If orbital computing becomes commercially viable, it should be designed around long-term space sustainability from the beginning.

13. The Most Realistic Future May Be a Combination of Earth and Space

The debate is often presented as a choice between two extremes.

One side imagines increasingly enormous terrestrial data centres. The other imagines a future in which most computing happens in orbit.

Neither extreme is particularly convincing.

Earth has major advantages.

It already has electrical networks, fibre-optic connections, technicians, industrial supply chains and buildings designed to support complex equipment.

Companies can expand existing facilities, upgrade processors and introduce better cooling systems without launching everything into space.

Terrestrial data centres can also be located near renewable-energy projects, industrial facilities and suitable water infrastructure.

Space has a different set of advantages.

It may offer access to extended periods of solar energy in suitable orbits, the ability to process information close to satellites and a way to reduce certain demands on terrestrial infrastructure.

It could become attractive for specialised applications that do not require the same economics as a general-purpose cloud service.

Consider three examples.

Example one: a business using an AI assistant.

A customer asks an online assistant to explain a document or generate a report.

The service needs fast connectivity, reliable computing and convenient access to the user's information.

A terrestrial data centre is likely to remain the practical choice.

Example two: an Earth-observation satellite.

A satellite captures a large volume of imagery to monitor forests, agriculture or coastal conditions.

Processing selected information in orbit could reduce the need to transmit every raw image to Earth.

An orbital computing platform could provide a meaningful advantage.

Example three: a large AI research organisation.

The organisation wants to train a model using enormous datasets and specialised computing clusters.

It needs extensive memory, fast connections between processors, reliable electricity and access to engineers.

An Earth-based facility may remain more cost-effective, although some specialised tasks could eventually be distributed across orbital systems.

These examples point towards a hybrid infrastructure model.

Computing will be distributed according to the needs of each workload rather than moved into space simply because space is available.

Boston Consulting Group reaches a similar broad conclusion in its 2026 analysis: orbital facilities are more likely to complement terrestrial data centres than replace them.

The report explores the possibility that orbit-advantaged workloads could represent a meaningful portion of the future AI infrastructure market, while acknowledging the continuing cost premium.

14. What Should Communities Demand from the AI Industry Now?

Orbital computing may eventually become an important part of the technology landscape.

But communities cannot wait ten years for a satellite network to solve infrastructure problems that are happening today.

The immediate responsibility belongs to developers, utilities, regulators and local governments.

There are practical measures that can make terrestrial data-centre development more responsible.

Publish Meaningful Resource Information

Developers should disclose projected electricity demand, water consumption, cooling methods and backup-power arrangements.

Information should cover both initial construction and planned expansion.

Protect Ordinary Electricity Customers

Large computing customers should face transparent arrangements that identify who pays for the grid upgrades needed to serve them.

The cost of new infrastructure should not automatically be distributed among households and small businesses.

Make Water Commitments Measurable

Communities should know whether a facility will use potable water, groundwater, surface water or reclaimed wastewater.

Contracts should specify expected consumption and how performance will be monitored.

Require Enforceable Community Agreements

Promises about employment, tax contributions and environmental protection should be written into agreements with meaningful consequences for non-compliance.

Improve Public Consultation

Residents should have access to relevant project information before critical zoning and financial decisions are made.

Compare Alternative Locations

A facility that is unsuitable near a residential area may be more appropriate in an existing industrial zone with sufficient electrical and water infrastructure.

These measures do not require governments to oppose AI.

They recognise that technological development is more sustainable when it respects the people and resources that make it possible.

15. What We Know — And What Remains Uncertain

It is useful to separate established developments from predictions.

What Is Already Established

• AI infrastructure depends on physical data centres with substantial electricity and cooling requirements.
• US voters overwhelmingly favoured greater restrictions or oversight in the 2026 local data-centre ballot measures.
• Electricity and water demand are important considerations in data-centre planning.
• Google launched a Project Suncatcher prototype satellite in October 2026.
• Google reported its first AI-generated response from that satellite on October 30, 2026.
• The GAO has identified potential advantages and significant engineering challenges associated with space-based data centres.
• Orbital debris and satellite congestion are recognised environmental and operational concerns.

What Remains Uncertain

• Whether orbital AI computing can compete economically with terrestrial facilities.
• How quickly large-scale cooling systems can be developed.
• Whether radiation protection and satellite reliability will become sufficiently affordable.
• How much computing capacity can be deployed per kilogram of launch mass.
• Whether optical communications can provide the capacity and reliability required by commercial AI workloads.
• Whether orbital networks can operate sustainably without adding unacceptable debris and atmospheric impacts.

The difference between these two lists is important.

The technology is real. Its commercial future is not yet guaranteed.

The same principle applies to the political backlash. The election results show strong opposition to many development proposals, but they do not establish that AI infrastructure construction will stop.

Developers may change locations, negotiate new agreements, redesign cooling systems or seek regulatory approval under stricter conditions.

Some communities may eventually accept projects that offer a more convincing balance of costs and benefits.

Others may continue to reject them.

The future will depend on how effectively the industry responds.

Key Takeaways

Frequently Asked Questions (FAQ)

1. Why Are Voters Rejecting AI Data Centres?

The main concerns involve electricity demand, water consumption, industrial land use, noise, backup generators, environmental effects and uncertainty about who pays for infrastructure improvements.

Some communities also object to tax incentives and development decisions made without sufficient public consultation.

The opposition does not necessarily mean residents reject artificial intelligence itself. Many are demanding better conditions for infrastructure development.

2. What Does the 44-to-1 Data-Centre Vote Mean?

It describes a reported tally of local ballot outcomes during the November 2026 US midterm election cycle.

These were different local questions involving construction restrictions, moratoriums, zoning and development approvals.

It was not a single nationwide referendum on AI.

Other trackers subsequently reported different totals because they included different measures and updated results at different times.

3. Can AI Data Centres Really Operate in Space?

In principle, yes.

Computers can operate aboard spacecraft, and Google's Project Suncatcher prototype has already entered orbit.

However, operating a small experimental computing platform is very different from running a large commercial AI data centre.

Power generation, thermal management, radiation protection, communication and maintenance must all be solved at a much larger scale.

4. Would Space-Based Data Centres Eliminate Water Consumption?

They could avoid the conventional onsite water requirements associated with terrestrial cooling towers.

That would be a potential advantage for communities facing water shortages.

However, orbital computing would still have an environmental footprint through manufacturing, electricity-storage systems, launch activity and other infrastructure.

It would be inaccurate to describe the technology as completely impact-free.

5. Why Is Cooling Computers in Space So Difficult?

Space is a vacuum, so conventional air-based cooling does not work in the same way it does on Earth.

Heat must ultimately be released through radiation using specialised thermal-management equipment.

Large computing systems can require extensive radiators, increasing the mass, complexity and cost of satellites.

6. Is Google Already Operating an AI Data Centre in Orbit?

Google announced on October 1, 2026, that its Project Suncatcher prototype satellite had launched and established contact.

On October 30, the company reported its first AI-generated response from the satellite's onboard computing hardware.

The mission is an experimental research project testing how AI accelerator hardware behaves under spaceflight, radiation and thermal conditions.

It is not a fully operational commercial AI data centre.

7. Will Orbital Computing Replace Terrestrial Data Centres?

Probably not completely.

Earth-based facilities have established electricity supplies, fibre networks, maintenance teams and manufacturing infrastructure.

Orbital computing may become useful for specialised tasks, especially processing information generated by satellites or workloads that benefit from particular orbital conditions.

A combination of terrestrial and orbital infrastructure is more plausible than the complete replacement of one by the other.

8. When Could Space-Based Data Centres Become Commercially Practical?

There is no confirmed date for widespread commercial deployment.

The GAO identifies some larger orbital data-centre deployments as possibilities for the mid-2030s, while industry analysis suggests that technical feasibility could improve during the next five to ten years.

Those are development expectations, not guarantees that the technology will become economically competitive.

Sources and Further Reading

1. Ballotpedia — Election Results, 2026: Ballot Measures. Election results and local ballot descriptions.
2. KCRA — Voters Overwhelmingly Reject Data Centres in Their Communities. November 2026 election reporting.
3. Data Centers on the Ballot — Redbud Consulting. Local ballot measure tracker and methodology.
4. Lawrence Berkeley National Laboratory — United States Data Center Energy Usage Report: 2025 Update. June 2026.
5. International Energy Agency — Energy and AI: Executive Summary. Global electricity demand projections.
6. Congressional Research Service — Data Centers and Water: Frequently Asked Questions. July 31, 2026.
7. Congressional Research Service — Data Centers and Their Energy Consumption. May 12, 2026.
8. MIT Center for Energy and Environmental Policy Research — Who Pays for Growth? Evidence from Data Centers and the Grid. June 2026, revised October 2026.
9. Urban Institute — Most Americans Do Not Want a Data Center in Their Area. September 11, 2026.
10. US Government Accountability Office — Science & Tech Spotlight: Data Centers in Space. April 28, 2026.
11. Google Research — Our Project Suncatcher Prototype Satellite Is in Orbit. October 1, 2026.
12. Google Research — We've Received Our First Response from the Project Suncatcher Prototype Satellite. October 30, 2026.
13. Boston Consulting Group — Space-Based Data Centers: More Than Hype, but Not a Revolution. August 27, 2026.
14. European Space Agency — Space Environment Report 2026. September 14, 2026.
15. Scientific Data — Global 3D Rocket Launch and Re-entry Air Pollutant and CO2 Emissions. October 2024.

Editorial note: The election figures are reported snapshots of local measures, not a national referendum. Energy, water and orbital-cost estimates are attributed to their respective research organisations. Predictions about commercial orbital computing remain uncertain.