Server locationsEight jurisdictions, chosen for their law.

Where a server sits decides who can compel its contents, and that is a legal question long before it is a technical one. These eight are on the list because of what it takes to reach them through a court — not because a map looked balanced.

edge capacity1,960 Gbit

See all 8

Same price everywhere · Move between them with one API call

The estate

One price, whichever you pick.

A region never costs more than another here. Charging a premium for the jurisdiction that protects you best would be charging for the thing we say we are for, so the price list is flat and the choice is yours to make on the merits.

8

jurisdictions

1,960

Gbit to the edge

13

internet exchanges

1.2 Tbit/s

mitigation, always on

Reykjavík

Iceland · KEF

Flagship Operational
Best for
Anything that expects to be complained about. Press, leaks, archives, mirrors.
Facility
Purpose-built, N+1 power and cooling, seismic-rated
Power
100 % geothermal and hydroelectric — not offset, generated
Uplink
2 × 100 Gbit, unmetered
Exchanges
RIX
Transit
Multiple tier-1 and regional carriers

Deploy here Uptime record Test targets

Bucharest

Romania · OTP

Core Operational
Best for
Bulk capacity inside the EU without the EU-average disclosure posture.
Facility
Carrier-neutral campus, N+1, dense south-east European transit
Power
Grid with N+1 UPS and generator, dual utility feed
Uplink
4 × 100 Gbit, unmetered
Exchanges
InterLAN · RONIX
Transit
Multiple tier-1 and regional carriers

Deploy here Uptime record Test targets

Sofia

Bulgaria · SOF

Core Operational
Best for
The cheapest EU floor space we would put a customer on.
Facility
Tier III-equivalent, N+1, low-cost EU capacity
Power
Grid with N+1 UPS and generator
Uplink
2 × 100 Gbit, unmetered
Exchanges
BIX.BG
Transit
Multiple tier-1 and regional carriers

Deploy here Uptime record Test targets

Chișinău

Moldova · KIV

Offshore Operational
Best for
The furthest outside EU reach while still being a two-hour flight from it.
Facility
Independent facility, N+1, no EU jurisdiction
Power
Grid with N+1 UPS and generator
Uplink
2 × 40 Gbit, unmetered
Exchanges
MD-IX
Transit
Regional carriers with diverse paths west and east

Deploy here Uptime record Test targets

Zurich

Switzerland · ZRH

Core Operational
Best for
When the point is that a court has to be convinced first.
Facility
Tier III+, N+1 throughout, physical security to banking standard
Power
Hydroelectric-dominant grid, N+1 UPS and generator
Uplink
2 × 100 Gbit, unmetered
Exchanges
SwissIX · CIXP
Transit
Multiple tier-1 carriers

Deploy here Uptime record Test targets

Amsterdam

Netherlands · AMS

Core Operational
Best for
Latency to Europe, and to anywhere reached through Europe.
Facility
Carrier-neutral, AMS-IX adjacent, N+1 throughout
Power
Grid with N+1 UPS and generator, dual utility feed
Uplink
6 × 100 Gbit, unmetered
Exchanges
AMS-IX · NL-ix · SpeedIX
Transit
Multiple tier-1 carriers

Deploy here Uptime record Test targets

Panama City

Panama · PTY

Offshore Operational
Best for
The Americas, without American jurisdiction.
Facility
Carrier-neutral, N+1, submarine-cable landing adjacency
Power
Grid with N+1 UPS and generator
Uplink
2 × 40 Gbit, unmetered
Exchanges
Panama-IX
Transit
Regional carriers with diverse subsea paths

Deploy here Uptime record Test targets

Singapore

Singapore · SIN

Core Operational
Best for
Asia-Pacific latency. Choose another region if jurisdiction is the point.
Facility
Carrier-neutral, N+1, major subsea landing adjacency
Power
Grid with N+1 UPS and generator, dual utility feed
Uplink
2 × 100 Gbit, unmetered
Exchanges
SGIX · Equinix IX
Transit
Multiple tier-1 carriers

Deploy here Uptime record Test targets

How to actually choose one

Two questions decide it, and they pull in opposite directions. Most people get this wrong in the same way.

The mistake worth avoiding

Latency is the loud question. Jurisdiction is the one that ends you.

Forty milliseconds is a number you can feel and will forget about in a week. The law where your disks sit is a number you never feel until the morning it decides everything. If the work could ever attract a complaint, pick on the law and put a cache in front of it. If it could not, pick the nearest region and stop thinking about it.

How a takedown is handled here
  • No premium for the safe ones Iceland costs what Amsterdam costs. We do not price the protection separately.
  • No lock to the region you picked Virtual instances move between any two with one API call, at no charge.
  • No published latency figures We have not measured yours. The looking glass gives you the targets to measure it.

On thirteen internet exchanges

Where a region has an exchange, we are on it. Peering shortens the path for your visitors and costs them nothing.

  • AMS-IX
  • BIX.BG
  • CIXP
  • Equinix IX
  • InterLAN
  • MD-IX
  • NL-ix
  • Panama-IX
  • RIX
  • RONIX
  • SGIX
  • SpeedIX
  • SwissIX

Our peering policy is open: any network, at any exchange we share, without a ratio requirement. The transit and peering page has the detail, including RPKI, IRR objects and the communities you can send us.

About the regions.

Does a region cost more than another?

No. Every product on the price list is the same figure in all eight regions. Charging extra for the jurisdiction that shelters you best would be charging for the thing this company exists to provide, so we do not.

Can I move a server between them later?

Virtual instances migrate between any two regions with a single API call and no change in price — the disk is copied, the address changes, and you are told the new one. Bare metal is a physical move, arranged with the network team inside a maintenance window you pick.

Which one should I choose?

If the work could ever attract a legal complaint, choose on the law: Iceland, Moldova, Switzerland or Panama, and put a cache in front of it for speed. If it could not, choose the region nearest your visitors — usually Amsterdam for Europe, Singapore for Asia, Panama for the Americas.

Why do you not publish latency figures?

Because we would be making them up. Latency is a measurement between two points, and one of those points is you. The looking glass publishes the targets for each region so you can measure the number that actually applies to you rather than trust one that does not.

Is my data replicated between regions?

Only where you ask for it. A server lives in the region you chose and nowhere else, which is the point of choosing. Object storage can replicate to a second jurisdiction you nominate, and private clusters can span two regions — both are things you switch on, never defaults.

What happens if a region has to close?

You are told before it is announced anywhere else, with at least ninety days to move and free migration to any other region, including a pro-rata refund of anything prepaid. A region leaving this list has never been something we would let a customer discover from a status page.