Public or Private Cloud: Choosing the Right Home for Every Workload

Articles Cloud
10 min read
Share to


Cloud First, but Which Cloud?

Cloud First has been the direction for many organisations for years. And it remains a relevant starting point. But Cloud First does not automatically mean Public Cloud with Microsoft, AWS or Google Cloud. 

Today, organisations can consume IT infrastructure as a cloud service in several ways. Public Cloud from the large global hyperscalers is one. Private Cloud delivered as a service from Norwegian data centres is another. For many, the right solution will be a combination: Hybrid Cloud. 

The key question is which workloads and data should run where, and why. Here, workloads means the applications and tasks the infrastructure needs to support. 

New Requirements for Control and Data Location

Geopolitics and global uncertainty have changed the way organisations think about their data and infrastructure. Organisations are increasingly considering who controls the data and infrastructure they depend on, which countries’ laws apply, and how vulnerable they are if the conditions around them change. Digital sovereignty has become part of the strategic IT discussion, and for many organisations these questions are now being discussed both by executive management and at board level. 

At the same time, regulations such as NIS2 and DORA have made risk management, digital resilience and dependencies on suppliers even more important. 

But this is about far more than security and regulation. 

A customer service platform with major traffic peaks, an AI solution requiring significant GPU capacity, and an OT environment where downtime can bring production to a standstill place fundamentally different demands on the infrastructure. 

In our conversations with customers, we also see these requirements moving further through the value chain. Our customers are increasingly challenged by their own customers, including requirements for data to be stored and processed in Norway or the Nordic region. For some, meeting such requirements can be critical to winning or retaining customers. Expectations are increasing, making infrastructure choices increasingly business critical. 

For some workloads, flexibility and rapid scaling are essential. For others, the priorities are very high availability, low latency, control and predictable performance. When downtime can have serious consequences for the business, deciding where a workload should run becomes far more than a technology decision. 

The choice depends on what each workload requires, and how the organisation balances breadth of services and functionality against control, security, performance, data location and cost. 

Such an assessment does not necessarily mean moving out of Public Cloud. It means making deliberate choices about what to place where. A good cloud decision therefore does not start with the platform. It starts with the data and how critical it is to the business. 

Start with the Need

A financial system, a digital customer service platform, a production-critical OT system and an AI solution place very different demands on the infrastructure. Before choosing a platform, the organisation should answer some important questions: 

  • Data and requirements: What data is being processed, who should have access, where should the data be processed, and what requirements apply?
  • Performance: How quickly must the solution respond, and which other systems does it depend on?
  • Availability: What happens if the service stops? How quickly must it be restored, and how much data can the organisation afford to lose?
  • Risk: What are the consequences for the business if data becomes unavailable, compromised or lost? What dependencies does the organisation have on suppliers, technology and infrastructure?
  • Capacity: Is demand stable, seasonal or difficult to predict?
  • Cost: What will the complete solution cost over time, including operations, licences, data transfer and the necessary expertise?
  • Responsibility: Who is responsible for security, maintenance, capacity and incident management?

These questions help separate what is non-negotiable from what can be balanced against cost, flexibility and complexity. They also give IT, security, finance and the wider business a common foundation for decision-making. 

Once the requirements are clear, the organisation can assess which platform is best suited to each workload. 

Assess the Platform to the Requirements

Private Cloud gives an organisation dedicated cloud resources in an environment that can be adapted to requirements for control, security, data location and operations, among other things. 

Private Cloud can be delivered with different levels of operational responsibility. In NetNordic’s model, customers can choose between IaaS, PaaS and SaaS, with a clearly defined division of responsibilities. Monitoring, patching, backup and preparedness form part of the foundation around the infrastructure. 

The choice between Public and Private Cloud is therefore about more than flexibility alone. Requirements for control, security, data location, performance and the services the organisation actually needs must also be taken into account. 

The Threat Landscape Makes Control More Important Than Before

A more challenging threat landscape places greater demands on control. It is not only about where data is physically located, but also about who has access to it, how the infrastructure is monitored, and how quickly an incident is detected and handled. 

At NetNordic, security is an umbrella over everything we do. We therefore see it as a given that our Private Cloud solutions are protected by a comprehensive SOC/MDR service at the infrastructure level. When critical workloads and data are entrusted to an external provider, we believe organisations should set clear requirements for continuous security monitoring and threat management. 

The questions organisations should ask are the same regardless of platform: Where is our data located? Who operates the infrastructure? Who has access? How are redundancy and resilience addressed? Which suppliers are involved in the delivery? Which jurisdiction is the organisation exposed to? And who is actually responsible when something goes wrong? 

This is particularly relevant for business-critical systems, sensitive data and OT environments where downtime can have significant operational or financial consequences. The point is not to argue for one cloud model. It is to understand what the workload requires and make the choice accordingly. 

"Our customers entrust us with highly sensitive employee data, so strong control over where data is stored and processed, who can access it and how it is protected is fundamental to us. We are therefore exploring NetNordic’s private container platform as a complementary Nordic infrastructure option alongside our existing public-cloud capabilities. For workloads tested on the platform, it offers Nordic data residency, predictable costs and a stable environment, while allowing our developers to retain the same container-based way of working. We are still early in the evaluation, but exploring different infrastructure options supports our continuing work to strengthen the security, resilience and flexibility of our services”.

Øyvind Ødegård

VP DevOps, CatalystOne 
Øyvind Ødegård

The Provider Is Part of the Decision

The platform and provider should be assessed together. The technical solution and price are important factors, but so are expertise, understanding of the organisation’s needs and the ability to provide ongoing support. 

“The decision to choose NetNordic was based on the technical solution, price and their reference customers. We felt that you understood our needs and adapted the solution to what we required. Your follow-up on questions around possible solutions also weighed positively. At the same time, we believe that your expertise will help us in our future customer relationships".

Espen Vangstein

Head of IT Norway, Bravida Norge AS.
Espen Vangstein

Hybrid Cloud Requires Everything to Work Together

A hybrid model connects different cloud environments, for example Public and Private Cloud. This allows organisations to place applications and services where they fit best. 

But placing each workload in the right environment is not enough. The environments also need to work together. 

Consider a customer service platform running in Public Cloud while the information it needs is stored in a system in Private Cloud. The connection between the environments then affects response time, availability and cost. Even if the platforms work well individually, the connection between them can become a weak point. 

Organisations therefore need to plan for more than where each individual application should run. Networking, identity, access management, monitoring and incident management need to work together. Responsibility must also be clear when an issue affects multiple environments. 

Hybrid Cloud is not about having a little bit of everything. It is about using different platforms where they create the most value, while making sure the whole environment works together. 

Language Models, Jurisdiction and Cost

When using external language model services, organisations need to establish where data is processed, who may access it and which countries’ laws apply. The provider’s terms, subcontractors and technical setup are part of that assessment. Data location alone does not provide the full answer. 

At the same time, the cost picture becomes more difficult to predict as AI moves from pilot to production. Token-based pricing is easy to get started with but can be harder to budget for when usage grows from a pilot to hundreds of users. 

Organisations that want greater control over where AI data is processed can choose to run open language models in Norwegian data centres. This can provide a different balance between data control and cost predictability compared with token-based services from global providers. 

At NetNordic, this is delivered through Model-as-a-Service, with dedicated capacity rather than payment per token. Customers can choose where inference takes place, and Norwegian language models may be relevant where the quality of Norwegian language output is critical to the results. 

Just as important as the model itself is the data users upload to it. When the model runs in an environment controlled by the organisation, documents, questions and responses can be managed within the same framework as the rest of its critical data. This requires logging, storage, access and any external integrations to be covered by those requirements as well. 

We also see that many customers need support in choosing the right model for the right task. One large model is rarely the best answer to every need. For organisations where Norwegian data residency is essential, it is important that the data remains on Norwegian soil. 

GPU as a Service

When the need for GPU capacity arises, the organisation also needs to decide how that capacity should be sourced. 

GPU as a Service provides access to GPU capacity without requiring investment in and operation of dedicated hardware. But the same questions still apply: Where does the capacity run? Where is the data processed? What performance is required? And what will the total cost be? 

Organisations should also understand whether GPU capacity is shared or dedicated, what capacity is actually available, how it can be scaled and who is responsible for the infrastructure and software. 

For organisations that want GPU capacity in a Norwegian Private Cloud environment, it can also be delivered with the same division of responsibilities, monitoring and security model as the rest of the infrastructure. 

Agents in Production

The next step for many organisations is AI agents that do more than answer questions. They perform tasks across networks, systems and other critical parts of the environment. 

This means organisations need to decide which identities and permissions agents should have, how their actions are logged and where the boundaries should be set. An agent with overly broad permissions can increase the risk of unintended actions and amplify the consequences of misuse or errors. 

This needs to be addressed before agents are put into production, not afterwards. 

For AI as well, cost must be assessed alongside requirements for data, capacity and operations. 

Compare Total Cost, Not Just the Price of Capacity

A low hourly rate does not necessarily result in the lowest-cost solution. What matters is the actual cost of meeting the workload’s requirements for quality, security, performance and availability. 

The comparison should include: 

  • Capacity and how efficiently it is utilised
  • Storage and data transfer
  • Licences and contract terms
  • Operations and required expertise
  • Security, backup and recovery
  • Migration and adaptations
  • Any parallel operations required during a transition period

For GPU capacity, the comparison should also take into account GPU type, memory, runtime and what the service actually includes. Stable and high utilisation can have very different economics from occasional demand. The total cost over time should therefore be compared, not a single price point. 

Are Your Workloads Still in the Right Place?

A review of your cloud strategy does not have to begin with a new and extensive migration plan. 

Start with a few workloads where there is a clear reason to challenge their current placement. It could be a system with high costs, a service with changing requirements, a business-critical OT environment or an AI project requiring significant capacity. 

A cloud strategy should not start with the conclusion that everything belongs in the same place. 

Some workloads belong in Public Cloud. Others are better suited to Private Cloud. For many organisations, the right architecture will be a combination. 

What matters is knowing why. 

Start with one workload and assess its requirements, risk and total cost together. 

Author

Therese Olafsen Mørck

Head of Growth Sales & Marketing
About Therese Olafsen Mørck
Therese Olafsen Mørck is Head of Growth Sales & Marketing at NetNordic Norway, where she leads initiatives focused on accelerating growth and strengthening NetNordic’s market position. She brings extensive experience in the cloud domain from previous roles, combining strong technology insight with commercial expertise in sales, marketing and business development.

Get in touch

Fill in the form and we will get back to you as soon as possible! Thanks!