Pick your platform
One question settles it: who consumes the capacity? The same Cozystack engine runs underneath all three, with a different commercial surface on top. They are not tiers, and picking one does not rule out the others.
Ænix AI Platform
You're a: AI-native company, a regulated organisation keeping model weights in-house, or a provider selling GPU-as-a-Service. NVIDIA GPUs through the GPU Operator: passthrough to VMs, NVIDIA vGPU for VMs (requires your NVIDIA vGPU licence), fractional sharing via HAMi, model serving and vector databases, with per-tenant usage measured for your billing.
Explore →Selling GPU capacity to customers? See GPU as a Service →
They combine. A provider adds GPU-as-a-Service on hardware it already runs. A bank adds the AI Platform inside the same tenancy boundary its auditor already reviewed. A telco sells a sovereign cloud product while running its regulated internal estate on the same engine. Adding a second surface is a configuration decision, not a second procurement. How the platforms fit together →
What we replace
If you came here for one of these, start at the linked page.
Migration / replacement
VMware / vSphere replacement
Broadcom price changes pushed you to look for a successor. We move VMware estates to Cozystack-based platforms — VM-first, with a real cloud-business stack on top.
VMware migration →A modern alternative to OpenStack
If your OpenStack is too much to operate but you still need a multi-tenant cloud, Cozystack gives you the same outcome with Kubernetes-native ops and a fraction of the operational surface.
Cozystack vs OpenStack →Bring workloads back from AWS / Azure / GCP
Sustained workloads on hyperscalers are no longer the cheapest option. We model the TCO and migrate without locking you into a new vendor.
Repatriation engagement →New cloud / sovereignty / compliance
Compliance-driven cloud architecture
DORA applies since 17 January 2025. NIS2 had to be transposed into national law by 17 October 2024. We design your cloud architecture so it holds up at the next supervisory review instead of being retrofitted in a hurry.
DORA compliance →Run AI on infrastructure you control
Replace OpenAI / Bedrock / Vertex API spend with inference and fine-tuning on your own GPUs. Model weights, data and audit trail stay on infrastructure you control.
Sovereign AI →Launch a public cloud on bare metal
For hosting providers, MSPs and regional clouds starting on hardware they own or lease: VMs, managed Kubernetes, databases and object storage for your customers, with billing, a white-label portal and WHMCS integration. Live in weeks once the hardware is ready.
Ænix Public Cloud Platform →Why open core
Ænix created Cozystack and is one of its maintainers, alongside maintainers from other companies. Cozystack is a CNCF Sandbox project, its CNCF Incubation application is in due diligence, and it is licensed under Apache 2.0.
The three Ænix platforms add the commercial layer on top: hosting panel, user, admin and support portals, service-creation wizards, billing and payment integrations, the WHMCS integration, and support with an SLA.
What that buys you in practice:
- No proprietary lock-in at the platform layer — the substrate is the same OSS project anyone else can run. Contracts are about who builds, runs, and supports it.
- Supplier transparency to the second hop — DORA-aligned subcontractor disclosure, with the upstream project open for your own review.
- Runs where you decide — on your hardware, in your jurisdiction. Encryption, audit-log retention and backup targets are set to your policy.
- Two jurisdictions — AENIX s.r.o. (Czechia) for EU contracts and EU public-sector procurement. AENIX INC (Delaware) for US contracts.
How we engage
Discovery call
30 minutes, free, no commitment. We confirm platform fit, regulatory context and scope, and whether we should keep talking.
Platform Readiness Assessment
14 days (focused) or 28 days (full), fixed price. Architecture review, gap analysis, written deliverables. Output: a roadmap you can take to the board.
Launch or build
Public Cloud Platform at provider scale: live in weeks once hardware is ready. Private Cloud Platform: a 3–12 month build, depending on scope. Multi-region national or operator programmes: a 3–6 month pilot, then 9–18 months to full multi-region. Optional managed operations afterwards.