Webinar · Took place online on Wednesday 19 August 2026 · Recording available

Add Kubernetes, databases and GPU to your price list

One hour with Andrei Kvapil, the creator of Cozystack: how a working provider widens the catalog on the racks it already owns — beside the platform it already runs, keeping its own billing and its own panel.

  • Recording on YouTube
  • 60 minutes, with a live Q&A
  • Free to watch
  • Questions about your own stack — talk to the team
Timur Tukaev — workshop host

Your infrastructure already earns. The question is what else it could sell

You own the racks, you run the platform, and you have the customer relationships. What customers ask for keeps widening — managed Kubernetes, managed databases, object storage, GPU, self-service, predictable billing — and every service you cannot offer is a conversation that ends somewhere else. This session is about closing that gap without rebuilding the business underneath it.

Where this usually starts

A customer asks whether you do managed Postgres, and the answer is no. A renewal quote arrives with a number on it that changes the maths on your VM margin. The platform works, but keeping it working is somebody's entire job. Or the panel your team wrote years ago still runs provisioning, and every new service turns into another month of glue code.

None of these calls for a rebuild. They call for one more platform beside what you run, and a way to sell what it can do.

What lands in your price list

Cozystack is an open-source cloud platform and CNCF Sandbox project that turns your hardware into a catalog: virtual machines, managed Kubernetes, managed databases, S3-compatible object storage and GPU — each one ordered from a self-service panel, provisioned automatically, backed up and monitored. You set the price.

On top of it, Ænix modules turn the platform into a business: billing and usage metering, a hosting panel, WHMCS and custom integrations, and a migration plan that moves customers one at a time.

Open-source foundation
  • Managed Kubernetes
  • Managed databases
  • Virtual machines
  • S3-compatible storage
  • GPU · passthrough, vGPU, HAMi sharing
+ Ænix modules
  • Billing & usage metering
  • Hosting panel
  • WHMCS & custom integrations
20+
managed services in the catalog, ready to price and sell
1 GPU → many tenants
whole-card passthrough, vGPU and fractional sharing on supported cards
€0
per-core hypervisor licensing — Apache 2.0, CNCF Sandbox project

European providers already running this

Not pilots. Production platforms selling to real customers, built on the same foundation the session walks through.

01
3 data centres
Switzerland · synchronous replication · more than a dozen tenants in production

A Swiss cloud provider moved off a hypervisor stack and a Jelastic/Virtuozzo legacy onto a full commercial public cloud — virtual machines, managed Kubernetes, databases and GPU, sold under its own brand. Its own engineers became Cozystack maintainers.

02
5× cheaper GPU
~11,000 active users moved with no downtime

A European academic-computing platform moved off a public hyperscaler onto owned bare metal and kept a single Cluster API spanning bare metal, a hyperscaler and a sovereign Swiss cloud — with fractional GPU sharing across jobs.

Both are anonymised at the customer's request. Andrei walks through what each of them actually did — and what he would do differently on your stack.

What the session covered

Forty-five minutes of practice, then your questions.

  1. 01

    The catalog, and what it earns. Managed Kubernetes, databases, object storage and GPU as sellable SKUs: what each takes to turn on, what to price it against, and which ones sell first.

  2. 02

    A second platform beside your current one. How Cozystack sits next to VMware, OpenStack, Proxmox or Virtuozzo without a cutover: where the two stacks touch on the network, on storage and at the load balancer.

  3. 03

    Moving customers across, one at a time. Migration in practice — what moves cleanly, how long a per-customer move takes, and the workloads Andrei would leave exactly where they are.

  4. 04

    Billing, metering and your panel. Usage metering out of the platform and into your invoicing. WHMCS as a worked example, custom panels over the API, and tenant lifecycle tied to payment status.

  5. 05

    GPU as a product. Renting a whole card or a share of one: passthrough, vGPU and fractional sharing (HAMi), what each isolates, and what you can put in a customer contract.

  6. 06

    The pilot, and the first ninety days. One node, roughly two weeks to a working environment, and the pilot becomes your production — no reinstall, acceptance criteria agreed by both sides.

The session ended with a live Q&A with the speaker.

What you take away

01

A realistic next catalog — the services worth adding first, and what each one is worth to you.

02

A clear-eyed view of billing, usage metering and panel integration — how they work in practice.

03

A migration path that keeps your customers online and your calendar your own.

04

A pilot you can actually start — one node, two weeks, and it becomes your production.

Who it was for

Providers that already sell infrastructure and are deciding what to sell next. If your platform today is VMware, OpenStack, Proxmox, Virtuozzo, CloudStack, OpenNebula or something you wrote yourself, the session was built around your situation.

  • Hosting providers
  • Cloud providers
  • Data centre operators
  • MSPs
  • Telecom companies

Especially the people who own the service catalog and the number under it:

  • Founders
  • CTOs
  • COOs
  • Product leaders
Andrei Kvapil

Your speaker

Andrei Kvapil
Creator of Cozystack · Co-founder and CEO of Ænix

Andrei created Cozystack, the open-source cloud platform and CNCF Sandbox project, after more than fifteen years of building clouds and high-load infrastructure. He contributes to Kubernetes, KubeVirt, Cilium and LINSTOR, and speaks at KubeCon and other industry events. At Ænix, he helps providers across Europe turn their infrastructure into commercial cloud services.

Watch the recording

The webinar took place online on Wednesday 19 August 2026. The recording is on YouTube.

Weighing what to add to your own catalog? Talk to the team, or see the upcoming workshops.

About the webinar

This was a free live webinar, held on 19 August 2026, for hosting providers, cloud providers, data centre operators, MSPs and telecoms that already sell infrastructure and are deciding what to sell next. Andrei Kvapil — the creator of Cozystack, an open-source cloud platform and CNCF Sandbox project — walked through how a working provider widens its catalog with managed Kubernetes, managed databases, S3-compatible object storage and GPU: what runs beside the current stack, what happens to billing and to the customer panel, how customers move across one at a time, and what the first ninety days look like. The recording is free to watch on YouTube.

Quick facts

  • Format A live online webinar, about 60 minutes: a practical walkthrough followed by a live Q&A with the speaker
  • Date Took place on Wednesday 19 August 2026, 16:00 CEST (14:00 UTC), online
  • Price Free; the recording is public on YouTube
  • Language English
  • Who it's for Founders, CTOs, COOs and product leaders at hosting providers, cloud providers, data centres, MSPs and telecoms that already sell infrastructure
  • Host Andrei Kvapil — creator and maintainer of Cozystack (CNCF Sandbox project), co-founder and CEO of Ænix
  • After the webinar The recording, plus a practical view of a catalog and migration plan you can adapt to your own stack

Frequently asked questions

We already run a cloud platform. Is this relevant?

That is who it was built for. Most of the hour was about adding a second platform beside an existing one, widening the catalog and moving customers gradually — not about starting from zero.

Do we have to migrate off VMware, OpenStack or Proxmox?

No. The two run side by side for as long as you want. Several providers keep the old stack for the workloads that are happiest there and sell the new services from the new one.

What happens to our billing?

It stays yours. WHMCS has a ready integration, anything else connects over the platform API, and usage metering exports into whatever you invoice from today. Your customer database stays the source of truth.

We wrote our own control panel. Do we throw it away?

No. The platform exposes a REST API and its own panel is optional. Andrei covered both patterns: keeping your panel as the customer-facing surface, or white-labelling ours.

Can we sell GPU from the same platform?

Yes. Whole-card passthrough, plus NVIDIA vGPU (requires your NVIDIA vGPU licence) and fractional sharing (HAMi) on supported cards, so one card can serve more than one tenant; tenant Kubernetes clusters on MIG-capable cards can also use MIG partitions. Andrei covered what each mode gives you and what you can put in a customer contract.

How many engineers does it take to run?

Fewer than OpenStack. The platform is one coherent Kubernetes-native stack rather than a dozen services you integrate yourself, and upgrades are a release rather than a project.

What does a pilot involve?

One node’s worth of hardware to start, access, and someone to accept the result against criteria agreed by both sides. The pilot environment becomes your production — you add nodes as you grow into it.

Is there a recording?

Yes. The webinar took place on 19 August 2026 and the recording is on YouTube: watch the recording. For questions about your own stack, talk to the team.

Webinar · Took place online on Wednesday 19 August 2026 · Recording available

Bring your stack to a conversation

The webinar took place on 19 August 2026. Watch the recording, then talk to the team about the catalog that fits your own platform.