All systems operational 12 elite locations · Europe & Asia Pay with crypto · 0 KYC
Home / Guides / Choosing a jurisdiction: twelve sites
Guide

Choosing a jurisdiction: twelve locations, Europe and Asia

Dark server hall lit by status-LED rows

Rank twelve hosting locations by latency, retention law and takedown posture. One SKU price, Iceland +20% and Switzerland +35%, VPS Core from $4.80/mo annual.

Short answer

Rank twelve hosting locations by latency, retention law and takedown posture. One SKU price, Iceland +20% and Switzerland +35%, VPS Core from $4.80/mo annual.

The flag is not the feature

Brochures sell flags; incidents are decided by four other things: who must reach the box in milliseconds, which legal process you actually expect, how much retention posture matters to the workload, and whether the service draws complaints. Rank those before the country, because the country is the one input you cannot patch later without moving data. Twelve sites share one SKU here, which keeps the ranking honest — nothing is discounted to steer you toward a flag the marketing team prefers this quarter.

What does not change with the flag is the commerce layer: signup is an email plus a 12-character password or a bare token, checkout runs the same 19 coins from XMR to USDT on seven networks, and the account model is no-KYC at every site. The flag changes latency, peering and law — nothing else in the relationship, which means every other factor you might weigh is identical before you start.

Where flag-first thinking breaks: picking a jurisdiction for its reputation and discovering the workload's real constraint was latency to users — or the reverse, optimizing milliseconds for a workload whose actual exposure is legal. Both mistakes cost a migration, and migrations are the one operation this catalog makes deliberately cheap but never free, because data location is not retroactive.

Latency claims are testable before any invoice exists: from the cities where your users sit, time the route to each candidate region and let the numbers, not the map, sort the shortlist. A region that looks close on a globe can route through three extra exchanges, and one that looks remote can ride a straight subsea path. Twelve candidates is a small enough set to measure by hand in under an hour, and measured data beats every brochure ever printed about peering density.

The method in this guide is to read each site as a package: a latency profile, a retention law, a takedown posture, and a price modifier where one exists. Start with the layer most comparisons never examine — the difference between the company you sign with and the country that hosts your disks — because confusing the two is where most forum-thread advice goes wrong.

VPS from $4.80* — Core on the annual term (list $6/mo), one invoice for 12 months, no auto-charges. Dedicated with IPMI from $39.20/mo. Crypto invoice, no KYC.

Launch now

Two layers: the signature and the rack

The contract you sign is with a Saint Kitts and Nevis company; the RAM you rent sits in one of 12 cities. Those are separate layers and both matter — domicile governs billing and corporate process, while the facility country governs the disks and the law enforcement that can reach them. Conflating the two is the standard analytical error in every 'offshore hosting' thread you have ever read, and the error survives because both words get shortened to 'jurisdiction'.

Neither layer is an exemption. The AUP travels with every SKU to every site — CSAM, phishing, spam and attacks are ban-level events everywhere — and local criminal process works normally at every facility. What the two layers actually decide is which civil claims have to travel far, under what retention rules your data sits, and how a takedown claim is read when it arrives. Abuse reports are handled under the law of the facility location, which tells you which layer owns the incident.

The rack layer is also where the operational facts live: Tier III or IV, 2N or N+1 power, uplinks from 20 Gbps to 100 Gbps, and DDoS filtering included at every site. None of those numbers moves with your threat model, and all of them matter to uptime — which is why they are constant across the catalog. When two candidate sites differ only in law and latency, the facility sheet reads as a tie, and that tie is information too.

Where it breaks: assuming the SKN domicile makes the facility unreachable, or that an EU facility overrides the contract entity. Each layer answers different questions — tax and corporate process follow the signature, data access follows the rack. A serious threat model writes down which question it is asking before choosing which layer to optimize, because the two answers rarely point at the same city.

With the two layers separated, the site comparison becomes a table you can actually fill in: latency profile, governing law, peering density, premium. The sections that follow fill it for all twelve sites, ordered roughly by how often each turns out to be the right answer rather than by geography or price.

Amsterdam: millisecond answers, takedown culture

Amsterdam is the latency pick for western Europe: the facility rides 40 Gbps into AMS-IX and DE-CIX, two of the densest peering fabrics in the world, which is why round trips into Germany, France and the UK feel local. For SaaS, VPN egress and anything users refresh constantly, this is the default rack — the one you pick when the argument is milliseconds and the legal posture merely needs to be competent.

The trade is legal texture: EU GDPR applies (EUR-Lex), and Dutch hosting case law reads takedown claims seriously rather than ignoring them. Crypto billing still keeps card processors and their KYC out of the relationship — the checkout is identical at every site — but content that draws documented complaints will draw process here faster than it would in Reykjavik, and that is a property of the jurisdiction, not of the provider.

The combination worth engineering: Amsterdam for the service users refresh and Reykjavik for the archive nobody should reach easily. Two racks on one no-KYC account, each doing the job its jurisdiction is best at, cost little more than one oversized middle-ground box. Deciding per workload instead of per company is the whole move — the flag is a property of the rack, and renting a second rack is a dropdown, not a project.

Where Amsterdam breaks: latency-sensitive workloads assume the whole catalog is slower and settle for it, or the opposite — privacy-sensitive users assume 'EU' means uniform and are surprised that the Netherlands processes claims more administratively than, say, Iceland does. The facility side is Tier III+ with 2N power on the same SKU as everywhere else; the law is the variable, so read it as one.

For the rest of the EU-adjacent map — Frankfurt, Stockholm, Helsinki, Luxembourg — the same GDPR baseline holds with different peering and process flavor. Before those, though, the two sites with the strongest legal posture in the catalog: Iceland and Switzerland, each carrying a published premium because enough people vote with their threat models to make the pricing public.

Iceland: retention silence at +20%

Reykjavik's argument is legal, not electrical: Iceland has no mandatory data retention for hosting, a speech-protective political climate — IMMI's home turf — and a datacenter industry built around exactly that posture. The hall is Tier IV with N+1 power on a 10 Gbps uplink peering FICIX, and the location premium is a published +20% rather than a surprise at invoice time. For publishing platforms, Tor workloads and anyone who wants the disk outside the EU, this is the default answer.

The mechanism worth understanding is transit: Iceland is an island, so packets to continental users take an extra hop through subsea cables, and that hop is the price of the posture. For workloads where latency is secondary — background jobs, publishing, archival, privacy-critical services whose users expect an extra hop — it is a fair trade, and the people paying it are buying the statute, not the scenery.

Operationally, the island posture shows up in maintenance windows rather than in speed tests: an isolated grid on geothermal and hydro power rides out continental events that reorder schedules elsewhere, and the Tier IV / N+1 rating is the mechanical expression of that independence. Publishers who have lived under takedown regimes in other countries tend to price the premium against the cost of one forced migration — a comparison the premium usually wins within a single incident. Distance is a billing line; silence is a feature.

Where Reykjavik breaks: EU-latency-sensitive applications that would have been fine in Amsterdam, and users who read 'no mandatory retention' as 'no law applies'. Icelandic law still operates — it is process-friendly toward speech, not absent — and the AUP is identical to every other site in the catalog. Abuse complaints still arrive and still get handled, just under Icelandic rules rather than Dutch or German ones.

If the goal is a non-EU legal system with strong institutions and a real judicial layer between a requester and the disk, the next site up the price ladder is the deliberate one: Zurich at +35%, selling judicial process rather than distance from Brussels — and it attracts a different kind of buyer for exactly that reason.

Zurich: FADP and a judicial overlay at +35%

Zurich buys you a specific thing: Swiss process. The revised Federal Act on Data Protection (Fedlex) puts a judicial overlay on data requests, Switzerland sits outside the EU's legal instruments, and the facility is Tier IV with 2N+1 power on a 100 Gbps uplink peering SwissIX and DE-CIX. The premium is +35% — the most expensive flag in the catalog, priced that way because a meaningful share of buyers consider judicial process worth exactly that.

Pick it when the buyer of the hosting cares: a client with Swiss data-transfer clauses, a compliance officer who reads statutes, or a threat model where a judge between the requester and the disk is a requirement rather than a preference. Mail and Matrix homeservers on Swiss boxes are the canonical pairing for exactly that reason — long-lived services whose operators want every future request to arrive as a legal question rather than an administrative one.

The buyer profile matters more here than anywhere else in the catalog. If a contract, an auditor or a client's counsel names Swiss process as a requirement, the premium is simply the price of the requirement; if nobody names it, the money buys calm rather than capability. A useful test: imagine the single unwelcome document your workload could receive, and check whether a judge standing between it and the disk changes the outcome. If the answer is no, spend the difference on a bigger port.

Where Zurich breaks: assuming non-EU equals non-cooperative. Swiss criminal law functions normally, and the FADP is a framework for legitimate process, not a wall against it. The other failure is paying +35% for a posture the workload never exercises — a latency-bound game server gains nothing from judicial overlay, and the premium is then pure cost with no mechanism behind it.

Between Amsterdam's speed and Zurich's process sits a fourth posture entirely: institutional. Luxembourg sells EU law with a data-trust regime tuned to finance at a smaller premium, and it reads foreign complaint templates very differently from its neighbors — which makes it the right answer for a workload most guides never consider.

Luxembourg: the institutional middle at +15%

Luxembourg is EU law with a data-trust regime built for finance: a Tier IV / 2N+1 hall on 40 Gbps with LU-CIX and AMS-IX reach, a +15% premium, and a regulatory culture accustomed to holding sensitive institutional data under supervision. Chain nodes, legal seedboxes and price-sensitive EU workloads land here for exactly that combination — GDPR baseline, institutional pacing, and a premium that costs less than Zurich while buying a different flavor of deliberateness.

The characteristic behavior is in how claims are read. Copyright templates arriving from abroad get treated as mail, not as orders, while criminal law proceeds normally — two channels, deliberately kept separate. For workloads that attract automated complaint mills but no criminal exposure, that distinction is the whole feature: the paperwork exists, gets read, and does not turn into a server shutdown before anyone with authority has actually reviewed it.

Two buyer groups dominate this rack. The first runs shared infrastructure where a steady stream of automated complaints arrives weekly — trackers, indexers, mirror lists — and wants those letters to enter a review queue instead of a shutdown script. The second prices EU presence for institutional reasons: funds, registries and the professional services around them want union law under their workloads without Amsterdam's exposure. Both groups accept the premium easily, because their alternative was never the cheapest flag — it was a compliance budget an order of magnitude larger.

Where it breaks: treating 'mail, not orders' as immunity. Documented, legally grounded claims still process under GDPR and Luxembourger law; the premium buys a deliberate administrative pace and institutional context, not a dead letter box. Workloads with real criminal exposure belong to a different conversation entirely, in every jurisdiction on this map without exception.

The Nordics extend the EU baseline with their own accents — a press-freedom tradition in Stockholm, dense peering in Helsinki, EEA membership in Oslo — and each prices differently, which makes them three separate entries in the comparison rather than one region to lump together and forget.

The Nordics: Stockholm, Helsinki, Oslo

Stockholm rides 40 Gbps into Netnod and DE-CIX North under EU GDPR and a press-freedom tradition that predates the modern web; Helsinki peers FICIX on a 20 Gbps uplink under plain EU GDPR; Oslo runs 20 Gbps into NIX under EEA-transposed GDPR with a +20% premium. All three are Tier III with 2N power, all three bill at base list except Oslo, and all three are one SKU with the same checkout as every other flag.

The distinction that matters is legal texture on top of the shared GDPR floor. Sweden's constitutional protections for editorial material have real history behind them, Finland offers some of the shortest paths into the Baltic region, and Norway's EEA position mirrors EU law without EU institutions sitting over it. The premiums differ accordingly — Stockholm and Helsinki at base, Oslo at +20% — and the latency profiles differ more than the brochures suggest, because peering density is not uniform across the region.

A practical note on routing: Helsinki and Stockholm give the shortest paths into the Baltic states and behave well toward northern Poland and Finland itself, while Oslo sits comfortably for Norwegian users but routes the rest of the continent through its neighbors. Measure from where the users actually are; the Nordics reward measurement more than any other region in the catalog because their peering densities differ sharply despite their geographic closeness.

Where the Nordics break: assuming 'Scandinavian' is one jurisdiction. Three parliaments, three data protection authorities and three peering markets behave differently in practice, and a failover pair of Stockholm plus Helsinki is redundancy in geography but not in law. If your threat model involves legal process at all, read the statute you are actually relying on, not the region label on the map.

East and south of that map, the legal baselines change character entirely: four Asia-Pacific sites run under APPI, PDPA and SAR law, with peering and latency profiles to match — and notably, none of them carries a location premium at all.

Asia-Pacific: Tokyo, Singapore, Hong Kong, Taipei

Tokyo rides 100 Gbps into JPIX and JPNAP under Japan's APPI; Singapore peers SGIX on 100 Gbps under the PDPA; Hong Kong runs 40 Gbps into HKIX as a SAR with its own regime; Taipei connects TWIX on 40 Gbps under Taiwan's PDPA. All four are Tier III+ with 2N power, all four bill at base list — there is no location premium anywhere in Asia on this map, and the SKU is identical to the European one.

The selection logic inverts Europe: latency to users dominates first, because regional users notice trans-Pacific round trips far more than they notice statutory nuance. A Singapore or Tokyo endpoint serving Southeast Asian users is the correct pick on that alone; the legal layer then differentiates among the four for workloads with specific data-residency clauses — Japan's APPI, Singapore's PDPA, the SAR's own framework, Taiwan's PDPA — each of which is a functioning statute rather than a marketing adjective.

Interconnection is the quiet differentiator in this region: JPIX and JPNAP in Tokyo, SGIX in Singapore and HKIX in Hong Kong each concentrate the regional carriers differently, and the same physical distance can produce noticeably different round trips depending on which fabric carries the handoff. Buyers who skip measurement because 'the map looks close enough' are the population most often surprised here — the Pacific is wide, and the cables cross it in specific, mappable ways.

Where Asia-Pacific breaks: assuming the legal posture matches a European site's. These are functioning rule-of-law regimes with their own retention and access rules — none of them is an 'offshore' void in the marketing sense, and none is GDPR either. Match clauses, not vibes: if a contract names a statute, the site choice follows the statute, and if it names nothing, latency decides.

Twelve sites cover the map; the remaining decisions are architectural rather than geographic — what happens when one site is not enough, and whether the second box should live in a second jurisdiction. The answer to both is the same argument, and it is cheaper to run than most people expect.

Failover is a jurisdiction decision too

Deploying two servers in two jurisdictions is supported and sensible: nothing in the account forces a single region, both boxes sit on one no-KYC account, and DNS or a load balancer decides who serves. The subtlety is that the pair is only as resilient as the difference between its two laws — Reykjavik plus Amsterdam survives different failures than Stockholm plus Helsinki, at identical prices, so the choice of pair is a threat-model statement whether you make it consciously or not.

Read the pair as covering two axes: geographic failure (power, subsea cables, national incidents) and legal failure (process that reaches one rack). A pair separated on both axes — say, Iceland and Singapore — costs the same per box as a pair separated on neither. Order Core at $6/mo list — $4.80/mo on the annual term — in each of two cities and the whole redundancy question costs under $10/mo effective, which is less than most people spend monitoring the thing they forgot to make redundant.

Test the pair before you trust it: take the primary down on purpose, once, in a quiet hour, and watch how long the second site needs to answer. A failover that has never been rehearsed is a theory, and theories about DNS or balancer behavior fail in creative ways — stale records, health checks aimed at the wrong port, certificates bound to a single name. The rehearsal costs minutes and converts the architecture from plausible to known.

Where failover breaks: replication that crosses the very boundary you chose. Database replication, object storage and DNS hosting each reintroduce a third jurisdiction silently; if the secondary's copies land in a region you never evaluated, the legal redundancy is fictional. Map where the bytes actually rest, then decide whether the answer still matches the threat model you wrote in the first section of this guide.

Jurisdiction ranks the sites, but the same map prices the metal: the dedicated tiers ship identical capabilities to every flag, which means the box choice and the flag choice can — and should — be made independently. That independence is the last mechanism to understand before the recap puts the whole sequence in order.

Metal inherits the map

Every dedicated tier ships the same capabilities at every site: IPMI on the board, ECC memory, NVMe in RAID 0/1 (0/1/5/10 on the top tier), DDoS filtering included, and unmetered ports from 1 Gbps up to 25 Gbps. The metal does not tilt the jurisdiction decision — Forge is the same 16c / 32t 9950X machine in Reykjavik as it is in Frankfurt, and the only thing that differs between the two racks is the statute hanging over them.

Budget the two decisions together to prevent the classic mismatch: a jurisdiction premium paid on a plan the workload never saturates, or bare metal dropped into a flag chosen for its price alone. Write the two numbers side by side — tier and modifier — and check each against the threat model separately. The catalog's flat structure keeps the arithmetic easy: 3 dedicated tiers, 3 VPS tiers, one modifier list, and no bundle pricing to untangle.

So decide the flag first and the hardware second: pick the law and the latency, then size from Core-class shared capacity through Forge to Summit and the dual-EPYC Monolith. Rent Forge at $49/mo list — $39.20/mo on the annual term — when the workload needs the rack's isolation, and spend the location premium only if the law justifies it. The dedicated pages carry the full spec per tier, and racking takes 2–4 hours at any flag, so the hardware decision is reversible in an afternoon while the location decision is not.

Where metal-first thinking breaks: buying top-tier capacity into a site chosen for its price, then discovering the workload's real constraint was legal exposure — or the reverse, paying a premium for law a workload never exercises. Hardware is replaceable in hours; data location is not retroactive. The facility law in force on day one governs everything written to the disk from that day forward, which makes it the least reversible choice on this entire page.

With both layers — jurisdiction and hardware — priced independently, the whole decision compresses into an ordering you can run top to bottom in one sitting. That ordering, in prose rather than in a table, is the recap below.

The jurisdiction choice, in the order to run it

Write the threat model first: who must reach the box fast, which legal process you expect, whether retention posture matters, and whether the workload draws complaints. Those four answers eliminate most of the 12 sites before any price is read, and they are the only inputs in this decision that do not change when prices do — which is why they go first and stay written down.

Verify the paperwork half early, too: the first bill should show exactly one list price plus the published modifier for the flag you picked, with the annual discount applied to the subtotal before any conversion. Any surprise on that page is cheaper to correct before provisioning than after, and the arithmetic is simple enough to check by hand — list, modifier, term discount, total. Ten seconds of reading replaces a billing ticket.

Rank candidates second: latency profile against the users' map, then governing law against the threat model — GDPR sites for EU baselines, Iceland for retention silence, Switzerland for judicial process, Luxembourg for institutional pace, the Asia-Pacific four for regional latency with no premium. Check the premium third: +20% for Iceland and Oslo, +35% for Zurich, +15% for Luxembourg, base everywhere else. Only when the flag and the tier both have reasons behind them, order on the deploy page — the coin and the term are dropdowns that never constrain this choice, as the XRP guide demonstrates.

Verify fourth: the first invoice confirms the premium math, the first week of logs confirms the latency assumptions, and the first incident — there is always a first — shows you which law the rack actually answers to. Revisit annually; statutes move faster than datacenters, and a posture that was right in one year can be merely expensive the next.

The whole sequence is a morning's work: threat model, shortlist, premium check, order, verify, calendar reminder. Twelve sites and one SKU keep the comparison honest; the four questions above keep it fast. Everything after that is operations — the hardening checklist and the WireGuard guide already own that part, so this page can end here.

Does the price change by location?

One list price covers every flag except three published premiums: Iceland and Oslo add 20%, Zurich 35%, Luxembourg 15%. Everything else from Amsterdam to Taipei bills at base list, and the 20% annual discount applies the same way at every site, so the location decision never fights the pricing decision.

Can I run failover across countries?

Yes. Deploy two servers in two jurisdictions, let DNS or a load balancer decide who serves, and manage both from one no-KYC account. The pair is most valuable when the two sites differ in law as well as geography — two identical legal postures is redundancy in hardware only, and hardware is rarely what fails worst.

Does the coin list change by country?

No. All nineteen checkout options — XMR, BTC, ETH, SOL, LTC, TRX, XRP, DOGE, DASH, ADA, GRAM, BNB and USDT on seven networks — clear invoices at every site on the map, and validation timing is identical everywhere. The coin you pick and the flag you pick never constrain each other.

Which flag for a dedicated server?

Pick the facility law first and the latency second. IPMI, ECC and NVMe RAID ship on all three dedicated tiers at every site, so the hardware does not tilt the decision — Forge in Reykjavik is the same machine as Forge in Frankfurt, under different law. The rack choice and the flag choice stay independent, so make them in that order.

Does the Saint Kitts and Nevis entity matter?

It governs the contract: billing, corporate process and the commercial relationship. The facility country governs the disks and which law enforcement can reach them. Neither layer exempts content from the AUP or from local criminal law at the site, and both layers matter to different questions — so write down which question you are asking before choosing which layer to optimize.

Is Iceland or Switzerland automatically more private?

Neither is automatic. Reykjavik offers no mandatory retention and a speech-protective climate at +20%; Zurich offers a judicial overlay on data requests under FADP at +35%. Which is 'more private' depends on the process you expect — write the threat model first, and let it pick the flag rather than the other way around.

Primary sources

More ways to pay

Pay with Bitcoin · Pay with Monero · Pay with USDT (TRC-20/ERC-20/SPL) · Pay with Litecoin · Pay with Ethereum · Pay with Solana · Pay with x402 · Pay with Dogecoin · Pay with XRP · Pay with TRX · Monero on bare metal · Dedicated servers with crypto · Pick the jurisdiction first.

Ready to launch?

Build the box — VPS or bare metal — create the password, pay the invoice that follows.