Does the number of VPN servers in a country actually matter?

Every VPN homepage leads with a number: 6,000 servers, 7,000 servers, 9,000 servers, 140 countries. The number is real, and it is also mostly beside the point. Here is what a server count can tell you, what it can’t, and the three things that decide whether a VPN works in the country you care about.

Short version

• Zero servers in a country is the only count that settles anything: no server, no IP there, no local library.
• Above zero, what matters is capacity per server, how many distinct IP ranges the provider has there, and whether the location is physical or virtual. A count hides all three.
• The providers say so themselves. NordVPN’s blog: “Does the VPN server count really matter?” and PIA’s: “Why server count doesn’t matter as much as you think.” Both sell server counts on their homepages anyway.

What a “server” even is

There is no standard. One provider’s server is a virtual machine on a shared box; another’s is a dedicated 10 Gbps machine. Mullvad publishes which of its roughly 570 servers it owns outright and which it rents, and moved its fleet to 10 Gbps links with a hosting partner and to RAM-only operation, where nothing survives a reboot. NordVPN, Surfshark and ExpressVPN have also moved to RAM-only fleets. None of that shows up in a count. Eight well-provisioned machines in Tokyo can carry more traffic than eighty crowded ones. TechRadar’s analysis and Tom’s Guide’s reach the same conclusion from their own testing: providers with modest counts regularly out-perform larger fleets.

The counts also aren’t comparable across providers. In the table below, NordVPN, Mullvad and Proton count servers; Surfshark counts locations; PIA counts WireGuard endpoints; ExpressVPN and CyberGhost publish cities and no count at all. We print each provider’s own unit rather than pretend they line up.

Japan, from the providers’ own lists

Proton lists 315 servers in Japan, Surfshark lists one location. Does that make Proton 300 times better for Japanese Netflix? No. Surfshark’s “location” is a cluster of machines behind one entry; Proton’s 315 are individual servers, some of them the same physical box seen through different IPs. What the table does tell you: everyone has Japan, Mullvad and PIA say theirs are physical, ExpressVPN has four cities, and NordVPN has obfuscated servers there, which matters if you’re on a hotel network that blocks VPN protocols.

Where the count does matter: IP ranges and streaming

Netflix and the BBC don’t block VPN accounts; they block IP addresses they’ve identified as datacenter ranges, and they watch for many users on one address. Netflix keeps a blocklist of datacenter IPs and shows the “you seem to be using a VPN” error. The BBC is reported to block whole hosting ranges rather than single addresses, and one provider’s support team says it is moving toward allowing only residential UK addresses. So in the UK the useful number isn’t servers, it’s how many distinct IP ranges a provider has and how quickly it rotates them when a range is burned. A provider with 400 UK servers in one hosting provider’s range can lose all of them in one afternoon.

No provider publishes its IP ranges per country, so the honest answer is: counts are a weak proxy, test results are the real signal. The 2026 tests we cite on each provider page found ExpressVPN worked in 5 of 5 Netflix libraries, NordVPN and Surfshark in 4, Proton in 3, which does not track server count at all. Our own test box will replace those with dated results per service.

Virtual locations: the number that should be on the homepage

A virtual location gives you an IP registered to a country while the machine sits somewhere else. ExpressVPN labels these on its server list; Surfshark tags them in its API; PIA marks them as “geo”. Proton and NordVPN don’t say per country, which is why our tables show “not stated” for them rather than guessing. Virtual is fine for streaming (the IP is what the service sees) and worse for speed (your traffic goes further than the flag suggests). India is the clearest case: after the 2022 CERT-In logging rule most providers removed physical Indian servers, so an “India” location today is usually a Singapore or UK machine with an Indian IP.

Load and distance: the two numbers that actually move speed

Your speed through a VPN is mostly three things: the distance to the server, how many people share it right now, and the protocol. The first is why cities matter more than counts: a provider with Osaka and Tokyo beats one with 200 servers all in Tokyo if you’re in Kansai. The second is why counts matter a little: more capacity in a city means fewer people per machine at 9 pm. NordVPN exposes per-server load in its API; most apps pick the least-loaded server for you, which makes the raw count invisible in practice. The third is WireGuard (or NordLynx and Lightway, which are WireGuard-class): every provider in our table offers it except ExpressVPN, which uses its own Lightway.

So what should you look at?

1. Is the provider in the country at all, and in which cities. Our country pages pull this from each provider’s own list, with the fetch date.
2. Physical or virtual, where the provider says. Virtual is fine for streaming; for speed and for a provider’s own privacy claims, physical is what you want.
3. Recent test results for the service you care about, dated, because a working server today is a blocked range next month.
4. Obfuscation if the network you’ll be on blocks VPNs (China, hotel and campus networks, the UAE). A count of ordinary servers is worth nothing there.

A big number on a homepage is a marketing choice, not a lie. It just answers a question nobody asked. The question people ask is “will it work in Japan next week?”, and the honest answer has a date on it.

Server figures in this post are read live from our catalogue at build time and carry their fetch date in the tables. Provider claims link to the provider; test results link to the tester. If a number is wrong, tell us and it will be fixed with a new date.