A tunnel encrypts every packet and puts it inside another packet. That wrapping travels over the network too — so more data goes out than without it. The only question is how much.
In short. A VPN does not save data, it adds to it. In our test on the night of 16 September 2026, the same 20 MB file cost about 10 % more through the tunnel than without it: +10.7 % on our line in the Netherlands and +10.1 % on the Polish one. While you do nothing at all and the tunnel stays on, it costs about 2 KB a minute — roughly 3 MB a day. We measured on our own servers and on one large download, not on a phone on a mobile network.
We make Mayak. We publish our measurements as they are, including the ones that do not flatter us — the 10 % overhead in this article is one of them. 7 days free after you confirm your email, no card needed. The trial is free and needs no card; without an email it is 3 days.
🍎 On iPhone Mayak works through the third-party Happ app — with the subscription link from your account. Happ is not in the Russian App Store; there incy (publisher LLC ITDEV) works instead, checked on 22 September 2026.
Our test: the same file with and without the tunnel
We took two production lines — our exit in the Netherlands and our exit in Poland — and did the same thing on each: downloaded one and the same 20 MB file first outside the tunnel, then through it, counting the bytes that actually crossed the network. The tunnel was brought up by the same client and with the same settings that go into a phone.
| What we measured | Exit in the Netherlands | Exit in Poland |
|---|---|---|
| Idle, 10 minutes | 2.1 KB/min — about 2.9 MB a day | 1.9 KB/min — about 2.7 MB a day |
| 20 MB file, no tunnel | 20.58 MB crossed the network | 20.62 MB crossed the network |
| The same file through the tunnel | 22.78 MB crossed the network | 22.71 MB crossed the network |
| Tunnel overhead | +10.7 % | +10.1 % |
Why even a plain download costs more than the file. The file weighs exactly 20 MB, and 20.58 crossed the network. The difference is the service headers every connection has, tunnel or no tunnel. So the fair comparison for a tunnel is not the size of the file but what the same download cost without it — which is where 10 % comes from, rather than 14.
The two lines differ by half a percent. That does not mean the country you pick has no effect: two points are far too few to turn into a rule.
What an idle connection costs
A tunnel that is on is never completely silent, even with the phone in your pocket. Every ten seconds it sends a short packet — “still here”. Without that signal your carrier’s equipment would close the connection within a minute or two, and your next action would wait for the tunnel to come back up. On top of that the two sides renew their keys from time to time, which is a few more packets.
Ten minutes of a completely idle connection cost 2.1 KB a minute on the Dutch line and 1.9 on the Polish one. Over a day that is 2.9 and 2.7 MB — less than a single photo.
⚠️ That is not all the background usage. The app itself also talks to our server: it renews the lease of the connection roughly every half hour and checks whether a new version is out. Those bytes are not in the test, and we did not measure them separately.
Where the overhead comes from
None of the reasons is about “extra data”. All of them are about wrapping.
- An envelope on every packet. Your packet goes whole inside another one, and that one has its own header and integrity tag — roughly 60 bytes per packet over IPv4 and about 80 over IPv6. The same 60 bytes seen from the other side are in our article on MTU and packet size.
- Packets are smaller inside a tunnel. Inside the tunnel a packet is not 1500 bytes as usual but 1420 — and 1360 when the path runs through two tunnels in a row. The same data therefore needs more packets, and each one gets its own envelope.
- The replies are wrapped too. While you download, your device keeps sending short acknowledgements back. They are tiny, and their envelope is the same size as a big packet’s. Small packets are where the share of wrapping shows most.
- The protocol itself adds a little. At the start of a connection AmneziaWG sends several packets with random contents, and on some lines a few random bytes are appended to every packet, so that the shape of the connection stops being recognisable. Next to the download itself that is small: everything together stayed inside the same 10 %.
Which leads to something more useful than the number: 10 % is about a large download. The smaller the packets — messages, notifications, short requests — the bigger the share of wrapping in them. How much bigger, we did not measure.
What that is in gigabytes a month
What follows is not a new measurement but arithmetic on our percentage: we take the 10 % overhead and apply it to familiar monthly volumes.
| Without a tunnel | With a tunnel | Extra |
|---|---|---|
| 1 GB | about 1.1 GB | about 100 MB |
| 10 GB | about 11 GB | about 1 GB |
| 30 GB | about 33 GB | about 3 GB |
| 50 GB | about 55 GB | about 5 GB |
Add the idle cost: with the tunnel on around the clock, about 80–90 MB a month goes out even if you never touch the phone.
⚠️ This is our percentage recalculated, not a measurement of your phone. Your percentage will differ: it depends on what you do and how large the packets are.
Why a tunnel never saves data
A VPN is sometimes expected to save data — a memory of browsers and extensions that compressed pages on the way and genuinely shrank them. Our tunnel does not do that and will not: it encrypts what you sent, and encrypted data does not compress — by construction it has none of the repetition compression lives on.
So the honest answer to “will a VPN save me data” is one word: no, it adds. How much is in the table above. That overhead is never below zero.
What matters more than the tunnel
The tunnel adds about a tenth. What makes differences of several times over is something else.
- What you are doing, and at what quality. A video call, video and file downloads cost wildly different amounts, and the gap between them is far larger than ten percent. We did not measure that.
- Packet size. The smaller the packets, the more the wrapping costs — see above.
- Drops and reconnects. Every new connection means another handshake. If the status keeps jumping between “Protected” and “Disconnected”, all of them add up, and that is a separate problem: why the connection drops by itself.
- Switching networks. Moving from Wi-Fi to mobile is a new connection as well. What that costs over a day of ordinary life we did not measure: our test rig sat on a wired link.
How to measure your own usage
Our percentage belongs to our test rig. You can get your own in an evening, and the instrument is already in the phone: the data counter in its settings.
- The overhead. Take a file of a known size and download it twice: once without the tunnel, once with it. Write down the counter before and after each download. The difference between those two numbers is your overhead.
- The idle cost. Turn the tunnel on, write down the counter, leave the phone alone for an hour and look again. Close the apps that go online on their own first.
⚠️ Trust the difference between two of your own measurements, not the absolute numbers. Your phone’s counter and your carrier’s counter count differently, and how exactly a phone splits traffic between apps while a tunnel is up is something we have not checked. Why one and the same instrument can give neighbouring numbers that differ several times over is in our article about speed tests.
At Mayak a number comes with what it does not say. The test above was run on our production lines — the same ones the app connects to. We ask for no payment details for the trial. A confirmed email opens the access — 7 free days. An email is optional: without one the trial is 3 days.
🍎 On iPhone Mayak works through the third-party Happ app — with the subscription link from your account. Happ is not in the Russian App Store; there incy (publisher LLC ITDEV) works instead, checked on 22 September 2026.
What we tested
- Where. Two production lines: our exit in the Netherlands and our exit in Poland. The tunnel was raised with the same AmneziaWG client, in the same version that runs inside the app, using the settings from our client configs — including the short packet every ten seconds.
- What counted. A byte counter on the server: it counts everything that crossed the network for that line, in both directions.
- What we measured. Ten minutes of a fully idle connection; then one and the same 20 MB file, first outside the tunnel and then through it. Each number is a single run, not an average of many.
- When. On the night of 16 September 2026.
What we do not know
We try to make sure nothing more can be read out of our text than we put in it.
- That you will get exactly the same. We measured on servers with a wired link, on two lines and with one file. On a phone on a mobile network the numbers will be different.
- How your carrier will count it. Billing, rounding and special terms for particular services (and the discussed fee for foreign traffic) are not our field, and we did not measure them.
- How much your own use adds. Video, calls, games and file sharing travel in packets of very different sizes. We measured one large download.
- The app’s own background usage. Its requests to our server were not part of the test and were not measured separately.
- The cost of switching networks. During the test the device never moved between Wi-Fi and mobile.
- Whether the country matters. Two lines are two points, not a pattern.
Short answers
How many gigabytes does a VPN use? By itself, almost none: about 3 MB a day while idle, that is under a tenth of a gigabyte a month. Everything else is your ordinary traffic plus roughly a tenth on top.
Does a VPN eat data? In our test it added about 10 % to a large download. On small packets the share of wrapping is higher, but by how much we did not measure.
Does a VPN use data when I am not using it? If it is on — yes, about 2 KB a minute to keep the connection alive. If it is off — no.
Can a VPN save data? No. We do no compression and do not plan to: encrypted data does not compress.
Does my carrier count the data that goes through a VPN? The bytes still travel over the carrier’s network, so it sees them and counts them. Exactly how is a question about your plan, and that we did not measure.
Should I turn it off overnight to save data? For data there is almost nothing to save: a day of idling costs about 3 MB. If the phone gets hot or drains fast, the cause is usually elsewhere — the answer in our help centre.
Does the country I pick change the usage? In our test the two lines differed by half a percent: +10.7 % and +10.1 %. Two points are too few to state a general rule.
Does the app use data while the tunnel is off? Now and then it asks our server about a new version. How many bytes that is, we did not measure.
Why does my phone show different numbers? Your phone’s counter, your carrier’s counter and our counter on the server all count at different points. Compare your own measurements with each other, not with ours.