Call records and network data on hardware only you administer.
Private LLM hosting on a dedicated machine in EU-Central, for carriers and network operators working with customer usage records, call data, and network operations telemetry.
Not compliance or legal advice. This page describes GPUwerk's infrastructure and contractual commitments. Whether your specific use of subscriber, call, or network data meets your national regulator's requirements is for your own compliance function or counsel to assess.
Why call and network data need a closer look
Call detail records reveal who talked to whom and when. Network topology and traffic data expose how the network actually runs. Both are the kind of data operators keep off general-purpose cloud tools by default.
"Support teams already summarize tickets in AI tools"
A support engineer pasting a customer's call history or account details into a consumer AI tool sends that data to a third party outside your contract. On a dedicated instance, the same summarizing and triage workflow runs on hardware assigned only to your team. See the private ChatGPT setup →
"We can't expose network topology"
GPUwerk runs infrastructure, not models: diagnostic logs and configuration data processed on your instance are handled by software you installed, behind SSH keys only you hold. Under the data processing agreement, GPUwerk does not access your workspace's content except at your request for support or where a legal obligation requires it.
"Usage data is regulated, and the rules vary by country"
Retention and handling rules for call and usage data differ across EU member states and beyond. We can tell you where the hardware sits and who operates it; matching that against your specific regulatory obligations is a job for your compliance team.
Infrastructure facts, plainly stated
| Question | Answer |
|---|---|
| Where is data physically processed? | EU-Central, on a dedicated single-tenant machine assigned to you |
| Does this satisfy our national telecom regulator's rules? | Not claimed. That's an assessment for your compliance and legal team, based on your specific obligations around retention, lawful interception, and customer data handling. |
| Who can access instance content? | Through the instance itself, only holders of your SSH keys; password login is disabled fleet-wide. GPUwerk keeps infrastructure administrator access, as on any hosted service, and under the DPA does not use it on your content except at your request or where a legal obligation requires it. |
| Sub-processors for the workload? | None, listed at /legal/sub-processors |
| DPA (GDPR Art. 28)? | Published at /legal/dpa, no charge |
| Certifications held? | None claimed on this page or elsewhere on the site. If your procurement requires a specific certification, verify directly with us before relying on this page. |
Practical starting points
A support organization might start with drafting: a model produces a first-pass reply to a subscriber ticket from the account history and prior correspondence, which a human agent reviews before sending, keeping the account data on hardware the operator controls throughout. A network operations team might point a model at years of incident runbooks and postmortems to search for a prior fix to a recurring fault, work that today usually means someone searching a wiki by keyword and hoping the terminology matches.
A third use is summarizing patterns in aggregated, already-anonymized usage statistics for a capacity planning report, a task where the sensitivity comes less from any one record and more from the fact that the underlying dataset, before aggregation, touched subscriber accounts. Keeping that entire pipeline, from raw metadata to finished report, on a single private instance avoids the question of which processing step counts as sharing subscriber-linked data with a third party, a question that's harder to answer cleanly once the data has crossed onto someone else's infrastructure, even briefly.
Pricing
On-demand spark-1x is $0.79/hour, spark-2x is $1.79/hour. Held (reserved) capacity is 75% of the on-demand rate: $0.59/hour and $1.34/hour. Full detail on the pricing page.
Questions we get from carriers and network operators
Can we run call-record analysis without sending customer data to a model vendor's API?
Yes. Your instance is a dedicated single-tenant machine reachable only over SSH keys you hold. GPUwerk does not access instance content except at your request for support or where a legal obligation requires it, under the published data processing agreement.
Does this meet our national telecom regulator's data handling rules?
GPUwerk can describe the physical location, operator, and access controls of the hardware. Whether that satisfies your national regulator's rules on call data retention, lawful interception, or customer data handling is an assessment for your compliance and legal team.
Is GPUwerk a GDPR processor for subscriber and usage data?
Yes, in the sense that a standard Article 28 DPA is published at /legal/dpa at no charge, with no sub-processors for instance workloads. Determining the correct legal basis for processing subscriber or usage data is your responsibility as controller.
Can network operations data stay off shared infrastructure?
Yes. A dedicated machine keeps network topology data, traffic logs, and diagnostic output on hardware assigned only to your team, not a shared multi-tenant pool. Email hello@gpuwerk.com with questions.
Related pages
See how other teams handling regulated data use the same infrastructure: insurance, public sector, or back to private LLM hosting overview.
Related pages
Bring your network and customer data workload to a machine you control.
Talk to the people who run the racks, or start with a pilot and a practical rollout plan.
Talk to us Deploy an instance