Guide
14-day trial: run a first workload
Form, no credit card. VMs, Kubernetes, GPUs included.
Published 2026-08-21
The Hikube trial lasts 14 days, with no credit card. It is not a fifteen-minute self-serve signup: work email and company, then an engineer in Switzerland opens the tenant, and your 14 days run from that opening. NVIDIA GPUs (L4, L40S, A100-80, RTX 6000 Pro, H100, H200) are included. You work on the same platform as paying accounts, operated by Hidora SA: three independent datacenters in Geneva, Gland and Lucerne. Request it on the free-trial page.
What the trial opens
Linux instances (Ubuntu, Debian, CentOS). Windows Server is a licensed image on the same instances; licenses are sized with the engineer by vCPU. Managed Kubernetes (Certified Kubernetes Hosted, Talos, Cilium CNI). Databases: PostgreSQL, MariaDB, Redis, MongoDB, ClickHouse. S3 object storage, backup, vault, VPC. Control: Hikube API and console; Terraform and Cluster API integration in preparation. kubectl, Helm and S3 clients stay.
First test: a VM
Once the tenant is open, a Linux instance is the shortest path to validate network, images and console. Pick flavor, image, VPC. Replication across the three sites is included in the volume price, with no HA surcharge: synchronous (real time) or asynchronous (improved performance). There is no separate multi-AZ SKU. Windows only if the workload needs it, with the license sized before the pilot. Detail: cloud instances.
Kubernetes cluster
The control plane is operated for you (KCSP). Workers stay in your tenant, on Talos: no SSH on the node. You keep kubectl, Helm, Flux or Argo CD. PVCs can replicate across Geneva, Gland and Lucerne, sync or async. Worker autoscaling with load. GPU node groups are in the trial. This is not a “Zurich region” GKE. Guide: Kubernetes across three Swiss datacenters.
GPUs during the trial
The trial includes GPUs. The engineer opens access and sizes L4 24G, L40S 48G, A100-80, RTX 6000 Pro 96G, H100 80G or H200 141G, duration and VPC isolation. Attach the GPU to an instance or a Kubernetes node group (GPU Operator). Datasets and checkpoints on Swiss S3 object storage; secrets in Vault.
Databases, S3, observability
For persistence: PostgreSQL or MariaDB privately via the VPC; MongoDB if you speak document (not Atlas). Backups to Swiss S3 are optional, not a magic switch without a runbook. Object storage: put/get/list with AWS CLI, rclone or an SDK; test your tool. On Kubernetes, platform monitoring uses Grafana, VictoriaMetrics and VictoriaLogs, not a US SaaS by default.
What you actually validate
That kubectl, Helm and an S3 client talk to Hikube. That a VM or cluster fits in your VPC. That an L4, L40S, A100-80, RTX 6000 Pro, H100 or H200 GPU fits the inference or fine-tuning you planned. That compute, storage and backups stay in Geneva, Gland and Lucerne. Hikube provides a DPA and a compliance page; Hikube does not certify your nFADP or GDPR file. If the topic is legal, read the CLOUD Act guide.
After the 14 days
The trial does not bind you. If you continue, billing is usage-based in CHF; GPUs and the Windows licence are at published catalogue rates. Endpoints stay the same. Support is an engineer in French and English, from Geneva, on business days 08:00–18:00 CET; infrastructure monitoring and on-call are 24×7. Priority detail on the support and operations page. Regulated volume or VMware exit: talk to an engineer. Operator: Hidora SA (Lancy), about page.
Frequently asked questions
No. Work email and company are enough.
Yes, subject to availability. All six cards are in the 14-day trial: L4, L40S, A100-80, RTX 6000 Pro, H100 and H200. There is no published cap in hours or quantity; the engineer scopes the allocation against free capacity, and sizes the model and VPC isolation.
Nothing is deleted at the deadline: your resources and your data stay in place. Billing only starts on a written agreement from you, a signed quote or a purchase order; without that agreement the trial ends and the engineer agrees with you what happens to the data. The 14 days run from when access was opened, not from your request.
No. It is a licensed image on cloud instances. Licenses are sized with the engineer by vCPU.
Yes. Compute, storage, backups and metadata stay in Geneva, Gland and Lucerne, with Hidora SA. This is not legal advice.
kubectl and Helm stay. S3 clients (AWS CLI, rclone, SDK) for put/get/list too. Terraform and Cluster API integration in preparation; today: Hikube API and console. Docs: docs.hikube.cloud.
Get the checklist
This estimate stays on this page: print it or export it as a PDF from your browser. For a costed study on your own scope, talk to an engineer.
Contact usReady to run on 100% Swiss infrastructure?
14-day trial, no credit card. GPUs included.