Private LLM hosting for logistics and supply chain
A dedicated DGX Spark for reading manifests, customs declarations, and supplier contracts with a language model that never phones home. The instance is yours alone, in EU-Central, at $0.79/hour for a single node or $1.79/hour for a linked 256 GB cluster.
What logistics teams actually put in front of an LLM
A freight forwarder or 3PL that wants an LLM in its workflow is usually trying to solve one of three problems: turning a stack of shipment manifests and bills of lading into structured data, checking customs documentation for errors before a shipment gets held at a border, or summarizing supplier contract terms across dozens of lanes and carriers. Each of those documents is commercially sensitive on its own terms.
A shipment manifest lists what moved, for whom, on which vessel or truck, and often the declared value. A customs declaration adds tariff codes, country of origin, and the importer of record. Neither is classified information, but in aggregate a competitor or a freight broker who saw a year of your manifests would know your supplier base, your volumes by lane, and your margins. Route optimization data, the model your dispatch team uses to sequence stops or choose carriers, is closer to a trade secret than a shipping label: it encodes years of tuning against your actual network, and it is exactly the kind of thing a shared AI API's logs could leak into a training run or a support ticket you never see.
Supplier contract terms are the third category, and they carry their own confidentiality clauses. Feeding a rate sheet or a service-level agreement into a public chatbot to get a quick summary can put you in breach of the contract you're trying to summarize, independent of whether the AI vendor actually misuses the data.
What dedicated hardware changes
A DGX Spark from GPUwerk is single-tenant: no other customer's workload runs on the machine while it's yours, and the container and workspace are removed from the node before it's offered to anyone else, as described in our Privacy Policy, section 12. You choose the model, whether that's an open-weight model you run yourself or a commercial model you self-host under its own license, and the manifests, customs paperwork, and contract text stay on that instance. We don't access, read, copy, index, or analyse what runs on it.
128 GB of unified memory on a single Spark is enough to run a 70B-class model comfortably for document extraction and summarization work, the kind of task logistics teams use most: pull the HS codes and country of origin off a customs form, flag a manifest line that doesn't match the purchase order, or draft a plain-language summary of what changed in a rate renewal. For higher-throughput batch processing, for example running extraction across a backlog of thousands of historical manifests, the linked 256 GB cluster gives you headroom without changing your workflow.
Where GDPR applies
Shipment manifests and customs documents usually name people: a consignee, a broker, a driver, sometimes an individual importer. If that counts as personal data under the GDPR, our Data Processing Agreement under Article 28 applies to how we handle the infrastructure it runs on; the DPA states plainly that GPUwerk hosts the machine but does not access the content of the instance, and that the controller decides what runs on it. We are not your lawyer and this isn't legal advice: if customs or trade compliance rules affect how you can process this data with AI at all, check with counsel who knows your trade lanes. What we can say concretely is where the data physically sits and who can reach it, which is often the harder question to get a straight answer to from a large AI vendor.
Practical starting points
Most teams don't start with a full rollout. A freight forwarder we'd point to this setup for might start with one narrow task: pulling structured fields, HS code, weight, consignee, value, off a stack of PDF manifests that currently gets keyed in by hand. A 70B-class model on a single Spark handles that kind of extraction well, and you can validate the output against your existing TMS before trusting it with volume.
A second common starting point is contract review: comparing a new carrier rate sheet against your standing terms to flag changes, without sending the rate sheet itself anywhere outside your own infrastructure. A third is customs pre-checks, having the model flag manifest lines where the declared value or HS code looks inconsistent with the rest of the shipment before it reaches a broker. None of these require the model to be perfect; they require it to be fast at a first pass and private enough that you can point it at real supplier and customer data without a second thought about where that data ends up.
Getting started
Deploy a Spark directly from the console, or if you'd rather scope the workload first, an AI Opportunity Session covers your document types, your existing systems (TMS, customs broker software, EDI feeds), and what a first pilot should look like before you commit to a rollout.
Back up before you terminate. Terminating deletes the workspace and GPUwerk's recovery copy of it. Export what you need before you stop paying for it.
Related pages
Run your logistics workload on hardware that's only yours.
Single-node Spark from $0.79/hour, no shared GPUs, no egress fees.
Deploy a Spark Book an engagement