Which data center processes Claude? Tracing the connection and limiting processing to Switzerland or the EU

Anyone using Claude through claude.ai, Claude Code, or the API cannot tell which data center processes the request. Only the first section of the connection, up to the nearest edge node, is visible. For companies that provide personal data or confidential content to a language model, that is insufficient: they must be able to demonstrate to customers, data protection advisors, or auditors where processing takes place.

This article shows how far the connection itself can be traced, what is publicly known about Anthropic’s data centers, and which options reliably restrict processing to the EU or the EU-plus-Switzerland area. Prices and regions are current as of October 2, 2026.

Switzerland focus: The revised Federal Act on Data Protection (revFADP) is decisive. Disclosure of personal data abroad is permitted under Art. 16 revFADP if the Federal Council attests that the destination country provides adequate protection (Annex 1 of the Data Protection Ordinance, DPO) or if appropriate safeguards such as standard contractual clauses are in place. EU and EEA states are on this list; the U.S. is included only for companies certified under the Swiss-U.S. Data Privacy Framework. A processing location in the EU therefore makes the justification considerably easier, but does not replace the other obligations (data processing agreement, informing data subjects, data security).

EU note: Companies based in the EU are instead subject to Art. 44 et seq. GDPR. The European Commission has recognized Switzerland as providing an adequate level of data protection (adequacy decision 2000/518/EC, confirmed in January 2024); processing in Zurich is therefore not a problematic third-country transfer for EU companies.

Which stops a request to Claude passes through

A request to claude.ai or api.anthropic.com passes through three stops:

The hostnames claude.ai and api.anthropic.com both point to address 160.79.104.10. It belongs to prefix 160.79.104.0/23, which is registered to Anthropic (AS399358) and, according to RPKI, is validly originated by Anthropic. In global routing, however, the prefix is announced exclusively through Cloudflare (AS13335): Anthropic brings its own IP addresses into Cloudflare’s Anycast network. As a result, the same address is answered from every Cloudflare location worldwide, and the connection reaches the nearest edge node.

Determining the edge node yourself

Cloudflare reveals the serving location in a diagnostic page and in the cf-ray header. The suffix after the Ray ID is an IATA airport code: ZRH stands for Zurich, GVA for Geneva, FRA for Frankfurt, and MRS for Marseille.

curl -s https://api.anthropic.com/cdn-cgi/trace

The relevant output lines are colo= (edge location) and loc= (the country to which Cloudflare assigns your source address). On Windows, PowerShell provides the same information:

$trace = Invoke-WebRequest -Uri "https://api.anthropic.com/cdn-cgi/trace" -UseBasicParsing
$trace.Content

Read the header using a HEAD request:

curl -sI https://api.anthropic.com/ \
  | grep -i -E '^(server|cf-ray):'

Options explained

Option	Effect
-s	Suppresses curl’s progress display and error messages
-I	Sends a HEAD request and outputs only the response headers
grep -i	Searches without regard to case
-E '^(server|cf-ray):'	Extended regular expression: only lines beginning with server: or cf-ray:

From a Swiss network, ZRH or GVA can generally be expected. If the internet egress is elsewhere, for example in a corporate network with centralized egress abroad, the edge node there responds instead, for example with colo=FRA or colo=MRS. The edge location therefore follows your network egress, not the processing location.

Why traceroute ends at the edge

A traceroute shows the path to the edge node and no further:

Test-NetConnection -ComputerName api.anthropic.com -TraceRoute -Hops 20

Options explained

Option	Effect
-ComputerName api.anthropic.com	Target host of the measurement
-TraceRoute	Determines router hops on the path to the destination
-Hops 20	Maximum number of hops to check

On Linux and macOS, the equivalent is traceroute -n api.anthropic.com (-n suppresses DNS resolution of the hops). In the measurement, the final hops before 160.79.104.10 were in address range 162.158.0.0/15, which belongs to Cloudflare. The trace ends after that: the connection is terminated at the edge node, and forwarding to the backend occurs over a new internal connection that is not visible externally.

RIPEstat shows who owns the prefix and through which network it is announced:

curl -s "https://stat.ripe.net/data/prefix-overview/data.json?resource=160.79.104.10"
curl -s "https://stat.ripe.net/data/bgp-state/data.json?resource=160.79.104.0/23"

In the second result, all AS paths end at 13335 399358: Cloudflare is the only upstream for the Anthropic prefix.

Based on my research, there is no public measurement that traces the path all the way to an inference data center; because termination occurs at the edge, this is also not possible with network tools. Measurement services such as llmlatency.dev capture only the response time to the first byte at the edge (as of September 2026: 96 ms median from US Central, 199 ms from Germany). Time to First Token could be compared from several locations, but it depends more heavily on the model, prompt length, and load than on distance. It therefore cannot provide a reliable inference about location.

What is known about Anthropic’s data centers

Anthropic has not yet operated its own publicly identified inference data center with an address. Its computing capacity comes from partnerships compiled by Tim Cadenbach on the TCDEV blog:

Training and inference do not necessarily take place in the same location. Inference can run in any AWS or Google Cloud region, and Anthropic does not disclose which location processes a response for a user in Switzerland. The only certainty is what the privacy policy states: the contracting party and controller for customers from the EEA, the United Kingdom, and Switzerland is Anthropic Ireland, Limited in Dublin; data is transferred to servers in the U.S. or other countries outside the EEA on the basis of standard contractual clauses.

Controls directly with Anthropic: U.S. only or global

Since generation 4.6 models, the Claude API has supported parameter inference_geo. It has exactly two values:

There is no value for the EU or Switzerland. The Workspace storage location (workspace geo) can currently also be set only to us. The response reports the setting applied in field usage.inference_geo, but not a specific region. For Haiku 4.5 and older models, the API returns error 400 for this parameter.

There is no processing-location setting for claude.ai and Claude Code with a subscription (Pro, Max, Team, Enterprise). Through Anthropic’s direct offerings, processing therefore cannot be restricted to either the EU or Switzerland. That is possible only through cloud vendors that operate Claude in their own regions.

AWS Bedrock: EU-plus-Switzerland area from Zurich

AWS Bedrock runs Claude on AWS infrastructure. According to AWS, Anthropic has no access to customers’ prompts, responses, or logs. You select the region through the source region of your API call and through the Inference Profile:

For Switzerland, region eu-central-2 (Zurich) is relevant. Opus 5.5, Sonnet 5.5, and Haiku 4.5 are available there, but only as Global or EU profiles and not In-Region. A profile that computes exclusively in Zurich does not exist. If the EU profile is called from Zurich, AWS distributes requests according to the model card (documented for Haiku 4.5 and Sonnet 4.6) across these regions:

Zurich is a possible target only if the request originates from Zurich. Processing thus remains within the EU-plus-Switzerland area. For Opus 5.5 and Sonnet 5.5, the model card does not specify the target regions; the API returns the actual list:

aws bedrock get-inference-profile \
  --region eu-central-2 \
  --inference-profile-identifier eu.anthropic.claude-sonnet-5-5 \
  --query "models[].modelArn"

Options explained

Option	Effect
bedrock get-inference-profile	Reads the definition of an Inference Profile
--region eu-central-2	Source region Zurich; the target list depends on the source region
--inference-profile-identifier	ID of the EU profile for Sonnet 5.5
--query "models[].modelArn"	Outputs only the model ARNs; the region appears in each ARN

By default, data is stored only in the source region. One exception is content retained for abuse detection: it is stored in the target region. Transport between regions is encrypted over the AWS network.

Demonstrating the processing region for each request

Bedrock is the only one of the described options that logs the actual processing region for each request. CloudTrail writes it in the source region to field additionalEventData.inferenceRegion:

aws cloudtrail lookup-events \
  --region eu-central-2 \
  --lookup-attributes AttributeKey=EventSource,AttributeValue=bedrock.amazonaws.com \
  --max-results 20 \
  --query "Events[].CloudTrailEvent" \
  --output text \
  | jq -r '[.eventTime, .eventName, .additionalEventData.inferenceRegion] | @tsv'

Options explained

Option	Effect
cloudtrail lookup-events	Searches CloudTrail events from the past 90 days
--region eu-central-2	Source region where calls are logged
--lookup-attributes AttributeKey=EventSource,AttributeValue=bedrock.amazonaws.com	Filters for Bedrock events
--max-results 20	Limits output to 20 events
--query "Events[].CloudTrailEvent"	Outputs only the complete event as JSON text
--output text	Output without JSON wrapper, one event per line
jq -r '… | @tsv'	Extracts timestamp, action, and processing region as a tab-separated line

In addition, a Service Control Policy (SCP) in AWS Organizations can block all regions outside the EU and Switzerland. If a profile target region is blocked, the request fails rather than falling back to another region. With this combination, the processing location is technically enforced and demonstrable for every request.

Claude Code through Bedrock in the EU

Claude Code can also be operated through Bedrock. In an eu-* region, Claude Code automatically selects prefix eu.:

export CLAUDE_CODE_USE_BEDROCK=1
export AWS_REGION=eu-central-2
claude

Prefix selection can be overridden with ANTHROPIC_BEDROCK_REGION_PREFIX. AWS authentication uses the standard mechanisms (profile, SSO, environment variables). Billing is through the AWS account, not a Claude subscription.

Google Vertex AI: EU only, explicitly excluding Switzerland

On Google Cloud (Vertex AI, now under the Gemini Enterprise Agent Platform name), the situation is less favorable for Switzerland:

Google explicitly excludes Switzerland from the EU multi-region endpoint: it covers only EU member states; the United Kingdom and Switzerland are not included. Several Swiss guidance pages cite europe-west6 as a path to Claude in Zurich; according to Google’s official location table, no Claude model is available there. Google does not document a log field containing the actual processing region for each request. For the global endpoint, Google states that the processing region can neither be controlled nor determined.

Vertex AI is thus suitable for processing exclusively in the EU, but not for processing in Switzerland.

Microsoft Foundry: currently no EU option

Microsoft offers Claude in Foundry in two variants. “Hosted on Azure” runs on Azure infrastructure, either as Global Standard or as Data Zone Standard; for Claude, the Data Zone exists only for the U.S. “Hosted on Anthropic” runs on Anthropic infrastructure, and Microsoft notes that data may be processed outside Azure and outside the selected region. Anthropic lists Foundry for Europe as “Coming soon.” In Microsoft 365 Copilot and Copilot Studio, Anthropic models are excluded from the EU Data Boundary and are disabled by default in the EU, EFTA (including Switzerland), and the United Kingdom.

Overview: which option guarantees which processing location

No vendor currently offers a path that guarantees Claude processing exclusively in Switzerland. AWS Bedrock with Zurich as the source region and the EU profile comes closest. If it must additionally be ruled out that data leaves the EU, for example because of contractual commitments to EU customers, the EU profile can be called from an EU region such as Frankfurt; Zurich then ceases to be a target.

What regional binding costs

All three cloud vendors and Anthropic itself charge a 10% premium for binding to a region or geography compared with the global endpoint. For Bedrock, pricing depends on the source region; Zurich, Frankfurt, and North Virginia cost the same. Routing between regions is not billed separately.

List prices in USD per 1 million tokens (input / output):

For Vertex AI, the EU-column price applies to Haiku 4.5 and Sonnet 4.6 in region europe-west1. A calculation example for an internal application with 50 million input and 10 million output tokens per month:

Regional binding therefore costs 10% more than the global endpoint. Compared with Anthropic’s direct list price using inference_geo: "us", an EU profile on Bedrock costs the same. More important considerations are likely to be the ancillary costs: an AWS or Google Cloud account with organizational policies, logging, and budget alerts must be set up and operated, and the fixed monthly prices of Claude subscriptions are replaced by strictly usage-based billing. Whether switching pays off depends on volume; a subscription has a fixed price per person but provides no control over the processing location.

Retention and training

In addition to location, how long data is stored matters:

• Anthropic API: Prompts and responses are not used for training unless explicit consent is provided. Zero Data Retention (ZDR) is available upon request per organization and can be combined with inference_geo. Content flagged by abuse detection remains stored for up to two years even with ZDR. For the Fable and Mythos models, 30-day retention is mandatory.
• AWS Bedrock: Content does not go to Anthropic. Retention can be selected per model between none, default and aws_review; Fable requires aws_review with up to 30 days of storage for review by AWS.
• Google Vertex AI: Google uses data for training or fine-tuning only with prior consent.