1. What Happens When a Data Center Loses Electricity?
A data center is a building full of servers, the computers that store files and answer requests from the internet. Servers use electricity, and they also give off a lot of heat, so air conditioning has to keep running too. The Japan Data Center Council, an industry group, explains that a sudden power cut can damage equipment and data, so centers use backup systems [1].
The first backup is a battery system called a UPS (uninterruptible power supply). The council says a UPS is meant to cover only about 5 to 30 minutes [1]. For longer cuts, a generator takes over, and a generator needs fuel.
That is where the 8-hour example comes from. The council wrote its guide during planned blackouts in Japan, when electricity was switched off on a schedule in different areas (rolling blackouts, as they are sometimes called elsewhere). With fuel for 8 hours and no refill, three 3-hour blackouts would total 9 hours, and the backup would be lost. The guide adds that in a large-scale blackout, deliveries of fuel might not be possible, so fuel has to be arranged in advance [1]. It is an example chosen to show the risk, not a record of something that happened.
The same guide says that moving important systems or backup copies to a data center outside the blackout area is also an option [1]. That leads to the first reason for spreading data centers out.
2. Why Keep Copies in Places That Do Not Share the Same Trouble?
If every copy of your data sits in one neighborhood, one flood or one failed power station can reach all of them. So large cloud companies group their data centers into zones that are built not to fail together. Amazon Web Services (AWS) says each of its "Availability Zones" has its own power, networking and connections. Its zones are designed not to be hit at the same time by problems such as a utility power cut, a water supply failure, cut fiber lines, earthquakes, fires, tornadoes or floods. Generators and cooling equipment are not shared between zones, and they are meant to be fed from different electrical substations [2].
The zones in one AWS region are placed "meaningfully distant" from each other, up to about 60 miles (100 km), which is far enough to avoid shared failures but close enough for fast, single-digit-millisecond communication [2]. Each region has at least three zones [3]. A simple way to picture it: keeping copies of a school project at home, at school and at a grandparent's house, so one spilled drink cannot ruin everything.
This explains why a company wants more than one site. It does not by itself explain why some sites end up far from big cities, which is where power comes in.
3. Why Do Power and Land Push Data Centers Away From the Biggest Cities?
Data centers use a great deal of electricity, and the amount is growing fast. A report from Lawrence Berkeley National Laboratory, published by the U.S. Department of Energy in December 2024, estimates that data center electricity use in the United States tripled over the previous decade and could double or triple again by 2028. The Department's announcement links this to the growth of artificial intelligence (AI) and other new uses of electricity [5].
In Japan, a research group at the think tank Mitsubishi Research Institute describes two mismatches. Data centers crowd into large metropolitan areas such as Tokyo and Osaka, where suitable sites are running short, while low-carbon power sources are spread across rural regions. Also, solar power peaks in the daytime but data centers must run 24 hours a day, 365 days a year [4]. Building in the countryside is one way to bring computing to where the clean electricity is, rather than carrying all the electricity to the city. Japanese policy discussions call linking power supply and communication networks this way "Watt-Bit collaboration." This source is a magazine report on a researcher's presentation, so treat it as a view from one expert, not a measured result.
Not every kind of computing can move, though. The same researcher separates work that can be done far away, such as training an AI model, which can tolerate some delay, from work that needs an answer almost instantly, such as controlling a self-driving car, which is better done close by [4]. So distance matters differently for different jobs.
4. Can a Far-Away Data Center Still Respond Quickly?
Distance adds delay, because signals take time to travel. The answer companies give is to build faster, more direct network lines. AWS says the zones in one region are connected by dedicated fiber with high bandwidth and low delay [2].
A Japanese example is in Ishikari, a city on the northern island of Hokkaido. The city announced that a new data center there, powered entirely by Ishikari's renewable energy, was completed on March 27, 2026, with some data halls planned to start operating in August 2026. When it opens, the city says, the building will be linked to Otemachi, a central business district in Tokyo, by an optical network service from the telecom company NTT East. The city states this will solve the earlier problems of distance and delay [6].
That last statement is the city's announcement, not a test result. The page gives no figure for how many milliseconds of delay are acceptable, or for which uses it will work. Because the data center was only beginning operation at the time of the announcement, how well the link performs is something the sources do not show yet.
5. Does Building Far From Cities Solve Everything?
No. Moving a data center does not shrink the amount of electricity its computers need. A report from the Japan Science and Technology Agency (a public research agency), summarizing a workshop held in March 2026, says that the huge amount of processing expected in the age of AI could cause power use so large that running it continuously becomes difficult. It calls for research into computing methods that are both fast and power saving [7]. Where a building stands and how efficient its computers are are different layers of the problem.
There are also trade-offs the sources leave open. Gathering computers in one place lets a company avoid owning extra machines, which saves energy; spreading them out can reduce that benefit. More sites also means more places to protect. The sources we read do not say which approach is better overall, and the answer likely depends on what is stored and how important it is.
Putting it together: spreading data centers out is supported both by disaster preparation and by electricity supply, and fast fiber lines help make it workable. It still comes with conditions, including daytime-versus-nighttime solar power, rising power demand, and the loss of some efficiency from gathering machines together.
6. How Can You Explore Where Your Own Data Lives?
With a parent or guardian, open the settings of a phone or tablet and look at whether photos are stored only on the device or also online. If they are online, some data center holds them, and you usually cannot see where it is. Then imagine a 9-hour blackout at home and write down what would stop working. That makes the 8-hour fuel example easier to feel.
To read more, the Japan Data Center Council guide on blackouts [1], the AWS fault-isolation paper [2] and the Ishikari announcement [6] are good places to start. When you read them, ask: how many hours of blackout does the writer assume, and how far apart do they assume the sites are?