That sentence still sounds strange, even in 2026.
Lonestar Data Holdings, a Florida-based company, has been one of the main names behind the idea. In 2025, it sent a small lunar data center payload through Intuitive Machines’ Athena lander mission. The payload weighed about 1 kilogram and carried 8 terabytes of SSD storage. It was not a full data center in the way we use that term on Earth. It was a test. But it was a real test, with real hardware, sent toward the lunar surface.
The idea behind it is easy enough to understand.
Some data is important enough that people want it far away from normal risks. Natural disasters, wars, power failures, damaged undersea cables, cyberattacks, local political disputes. A copy of data stored in another city is useful. A copy stored in another country is better. A copy stored 384,000 kilometers away feels like another category entirely.
For archives, government records, legal documents, scientific datasets, and long-term backup, the Moon has a certain appeal. It is remote. It has no floods, no earthquakes, no local residents complaining about noise, and no national government can claim the Moon as its territory under current space law.
So, in theory, lunar storage solves a narrow but real problem: how to keep a copy of important data away from Earth-based disruption.
In practice, it solves far less than the headlines suggest.
The main issue is not rockets, storages, even lunar dust, though that particular problem deserves its own angry memo.
The main issue is the speed of light.
NASA puts the Moon at about 1.3 light-seconds from Earth. That means a signal takes roughly 1.3 seconds to go from Earth to the Moon, and another 1.3 seconds to come back. So before we talk about server processing time, application logic, encryption, routing, retransmission, or anything else, we are already looking at about 2.6 seconds for a basic round trip.
That is acceptable for an archive.
It is painful for almost everything interactive.
A normal web page is not one neat request. It is a conversation between the browser and several servers. The browser asks for the page. Then it asks for scripts, images, style files, fonts, tracking scripts, product data, account status, cart status, location, pricing, stock, payment options, and recommendation data.
The 2025 Web Almanac from HTTP Archive found that the median desktop page made 77 requests, while the median mobile page made 72. It also notes that request volume matters because each request adds latency overhead, even when the total number of bytes is not large.
Now, to be fair, modern browsers do not load every request one after another. Many files are fetched in parallel. Caching helps. HTTP/2 and HTTP/3 help. Engineers are not fools, most days.
But interactive applications still have dependencies. One response often tells the browser what to ask for next.
A shopping page may need to know who you are before it can show your address. It may need your address before it can show shipping. It may need shipping before it can show the final price. It may need your product history before it can show recommendations.
Put that chain on the Moon and the numbers become annoying very quickly.
Five dependent server calls means at least 13 seconds of network waiting.
Ten dependent calls means 26 seconds.
Twenty means 52 seconds.
That is before download time, server processing, JavaScript execution, database lookups, third-party services, and all the other small delays that make web developers quietly lose their hair.
Now imagine autocomplete.
You type “laptop”.
A normal search box may wait until you stop typing, then send a request for suggestions. Some systems send requests after two or three characters. Others update more often.
If each suggestion request has to go Earth to Moon and back, every response carries that 2.6 second minimum delay. Type six letters and the system may still be answering what you typed several seconds ago. By the time it suggests “laptop stand”, you have already finished typing, clicked somewhere else, or closed the tab because you are a busy person with a life.
Recommendations are worse.
A shopping website might need your profile, browsing history, local inventory, current promotions, vendor rules, delivery options, and product availability before it can show a useful list. Even if only six of those steps are sequential, that is 15.6 seconds of minimum waiting. If the site makes ten dependent calls, that becomes 26 seconds.
For a backup archive, 26 seconds is nothing.
For a customer trying to buy a phone charger during lunch, 26 seconds is an insult.
This is why lunar data centers are not really a replacement for data centers on Earth. They are closer to remote vaults. Useful for records that must survive. Poor for anything that needs to respond while a human is still paying attention.
That brings us back to Earth.
Earth-based data centers have their own problems. They use land, electricity, cooling systems, batteries, servers, storage arrays, routers, cables, and a long list of supporting equipment. The International Energy Agency estimated data centers used about 415 terawatt-hours of electricity in 2024, around 1.5% of global electricity consumption. Its base case projects that figure could reach around 945 terawatt-hours by 2030.
Power gets most of the attention. But hardware has an afterlife too.
Every data center, office server room, hospital IT department, bank branch, school lab, and government office eventually removes equipment from use. Servers age, drives fail, and laptops are replaced. Furthermore, network equipment is upgraded, backup tapes are retired, SSDs leave production, racks are cleared.
At that point, the question is not only “Can this equipment be recycled?”
The question is also “What data is still inside it?”
That is where the conversation becomes much less futuristic, and much more practical.
The world already generated a record 62 million tonnes of e-waste in 2022, according to the Global E-waste Monitor 2024. Less than a quarter of that mass, 22.3%, was documented as properly collected and recycled. The same report projects global e-waste could reach 82 million tonnes by 2030.
Inside that waste are metals, plastics, glass, batteries, circuit boards, and hazardous substances. UNITAR also estimated that 31 million tonnes of metals were embedded in 2022 e-waste, including copper, gold, and iron, with a total embedded metal value of about US$91 billion.
So when companies talk about building data centers on the Moon, it is worth asking a simpler question first.
How are we handling the data centers, server rooms, and old IT assets already here?
Because most companies do not have a Moon problem.
- They have a storage room problem.
- They have old laptops from former employees.
- They have hard drives removed from servers and placed in boxes.
- They have backup media nobody has checked in years.
- They have network equipment waiting for resale.
- They have a disposal process that depends too much on trust and not enough on proof.
For most organizations, the final stage of data security is not a rocket launch. It is inventory, chain of custody, verified data erasure, physical destruction when needed, material separation, recycling, resale where safe, and proper documentation.
- That work is not glamorous. It is also where many real risks sit.
- A drive that leaves the building without verified erasure is still a data asset.
- A laptop sold without proper wiping is still a data asset.
- A server waiting in storage is still a data asset.
- Deleting files is not the same as removing risk. Retiring equipment is not the same as closing the data lifecycle.
The Moon data center story is interesting because it shows how far organizations are willing to go to protect information. But it also exposes a more ordinary gap.
We are discussing how to preserve data 384,000 kilometers away, while many companies still cannot say with confidence what happened to the drives that left their office last quarter.
That is the part worth fixing first.
Because before data goes to the Moon, it usually spends years on Earth, inside devices that someone eventually has to collect, verify, destroy, refurbish, or recycle.
And that last step deserves more attention than it gets.