Partnership Conversation

Completing the
AI Factory


The missing commercial layer
on top of the NVIDIA AI factory



NVIDIA + STAKATER

The landscape

NVIDIA is building
the AI Factory of the world


DGX systems, H100 and Blackwell clusters, AI Enterprise software, CUDA — the full hardware and driver stack for enterprise AI. The infrastructure race is decided.


DGX SuperPOD H100 / H200 / Blackwell NVIDIA AI Enterprise CUDA & cuDNN GPU Operator Base Command

The question now is not which GPU.
It is how do you turn it into a commercial cloud.

The problem

They bought the cluster.
Now nobody can use it.


An organization buys a DGX cluster. The infrastructure team guards it. Teams file tickets. External customers cannot be onboarded. Nobody has chargeback. Nobody has self-service. The GPU investment sits idle while the opportunity passes.



🔒
Access bottleneck
One team controls the GPU cluster. Everyone else waits for an admin to open a ticket.
📊
No chargeback
Finance cannot see which team consumes which GPU-hours. Cost accountability is impossible.
🚧
No isolation
One team's runaway training job starves every other team. No quotas, no policies, no guardrails.

The stack today

The commercial layer is missing

Who needs to consume it
Internal teams and external customers
Data scientists ML engineers Business units Paying customers
Missing commercial layer
Self-service, multi-tenancy, chargeback, catalog, policy
GPU quota per team Self-service provisioning Chargeback per business unit Tenant isolation Service catalog Policy enforcement
NVIDIA delivers
Hardware + AI software stack
DGX / H100 / Blackwell NVIDIA AI Enterprise GPU Operator CUDA Base Command

The solution

Cloud Orchestrator is that commercial layer

Built on OpenShift. Sits directly above the NVIDIA stack. Turns the GPU cluster into a cloud — billable, multi-tenant, self-service — for internal teams with chargeback or external customers paying for GPUaaS.

Teams and customers get
Self-service GPU cluster Notebook environments Inference endpoints Model registry
Cloud Orchestrator
Multi-tenant GPU quotas Service catalog Chargeback Policy enforcement Air-gap ready ITSM integration
NVIDIA infrastructure
DGX clusters GPU Operator AI Enterprise CUDA

What we add

From GPU cluster to AI factory for every team

⚡
GPUaaS
Any team or paying customer provisions a GPU cluster from a portal in minutes. Quota-bound, policy-enforced, billed automatically.
🎯
Inference-as-a-Service
Managed inference endpoints deployed on NVIDIA hardware. Teams consume models without touching infrastructure.
💰
GPU chargeback
GPU-hour metering per team, per project. Finance gets accurate cost data. No more shared clusters with invisible cost.
🔐
Tenant isolation
Hard namespace boundaries between teams. One team's job cannot starve another. Policy enforced by default.
🏛️
Air-gap ready
Full deployment with no internet dependency. Built for banks, government, and defense — environments where public cloud is not an option.
📋
Service catalog
Notebooks, training environments, inference endpoints — all available as catalog items. Teams order; the platform delivers.

The market

Europe is building AI on-premise

GDPR, the EU AI Act, and data sovereignty rules are pushing regulated AI workloads back in-house. These organizations cannot run sensitive AI on public cloud.

🏦
Banks and financial institutions
Fraud detection, credit scoring, risk models — all regulated, all on-prem. Buying DGX clusters now. Missing the platform layer to operate them as a service across business units or sell to fintech clients.
📡
Telcos and national operators
Own massive infrastructure. Running network AI, predictive maintenance, customer analytics. Need internal cloud operating model on top of existing GPU investment.
🏛️
Government and defense
Air-gapped by requirement. Building sovereign AI capability. Cannot use hyperscaler clouds. Need the complete stack on-premise — hardware and operating model.

The partnership

Hardware + commercial layer = a complete AI cloud

NVIDIA brings
DGX hardware and GPU clusters
AI Enterprise software stack
CUDA and GPU Operator
Global brand and enterprise channel
AI factory narrative and positioning
+
Stakater brings
Multi-tenant platform layer
Self-service provisioning and catalog
GPU metering and chargeback
Air-gap and sovereignty compliance
European enterprise deployments
=
An AI factory teams can actually use
NVIDIA hardware utilization goes up
Every DGX sale becomes a platform sale
Banks and telcos can run regulated AI on-prem
Operationalization is solved, not an afterthought

What we are asking for

Three simple steps

1
Evaluate the ISV partnership fit
A technical conversation with your ecosystem team to validate the integration and the architecture story.
2
Identify one joint customer opportunity
A European bank, telco, or government organization buying DGX where the missing platform layer is the blocker. We co-sell.
3
Build a joint reference architecture
A published NVIDIA + Stakater AI factory blueprint for European regulated industries. Shortens every future deal.

About Stakater

Built in Europe. Deployed in regulated industries.

Stakater is a cloud platform company headquartered in Sweden. We build Cloud Orchestrator — the operating layer for private and sovereign cloud.

10+
Years building
OpenShift platforms
EU
Headquartered
in Sweden
RH
Red Hat Advanced
Business Partner
🏦
Deployments in
financial services


Cloud Orchestrator
The self-service platform layer for private and sovereign cloud. Built on OpenShift. NVIDIA GPU Operator compatible. Air-gap deployable. Multi-tenant by design.