When we sat down to architect the infrastructure for Octigen, we knew we were walking a tightrope. On one hand, we needed highly scalable, Kubernetes-native infrastructure with access to cutting-edge AI capabilities. On the other hand, handling highly sensitive data for our clients meant we had to take data sovereignty - specifically the implications of the US CLOUD Act - incredibly seriously.

To build out our stack, we evaluated virtually the entire landscape of cloud providers. We looked at the traditional hyperscalers and the European independent alternatives, specifically focusing on cloud-native maturity and AI tokens-as-a-service offerings in the Switzerland/EU regions.

We have thoughts. Here is what we found, and how we ultimately decided to build the foundation for Octigen.

US Hyperscalers (AWS, Azure, and GCP)

There is a reason the "Big Three" dominate the market, but for European startups handling sensitive data, their US ties present a fundamental challenge: the US CLOUD Act. Even if the data is stored in European data centers, US-based companies can technically be compelled to hand over data to US authorities. For companies with strict compliance needs, this is a glaring red flag.

Setting the legalities aside, here is how they stacked up for our specific use cases:

Amazon Web Services (AWS)

• The Good: AWS provides fantastic resources, notably through their internal Solution Architects. They also have an incredible track record for backward compatibility - it's rare that a service is completely deprecated without a clear migration path. If you want a service, AWS has it (with some asterisks).
• The Bad: It is highly complex. You need specialized, dedicated AWS knowledge just to get by without shooting yourself in the foot.
• The AI Angle: Surprisingly, the Zurich region doesn't yet offer a great model choice for their AI tokens-as-a-service offering (Amazon Bedrock).

Microsoft Azure

• The Good: For AI tokens-as-a-service within Swiss borders, Azure is probably the best provider at the moment, largely thanks to their Foundry models.
• The Bad: As a startup, getting direct support from Microsoft was tough; you are heavily pushed toward their (admittedly vast) local partner network. Furthermore, the security model is incredibly complex. Entra ID (formerly Azure AD) is an absolute beast to manage, potentially even more complicated than AWS IAM and security groups.

Google Cloud Platform (GCP)

• The Good: GCP is remarkably simple to get started with and feels very startup-friendly.
• The Bad: While we don't have enough long-term experience with GCP to fairly compare their maintainability or migration paths against AWS, their AI offering fell short for our specific geolocation needs. Vertex AI is available, but model choice localized within Switzerland is almost nonexistent.

The European Contenders: Exoscale and OVH

Knowing the CLOUD Act was a potential dealbreaker for our most sensitive workloads, we turned our attention to the European market.

Exoscale (The Swiss Independent)

Exoscale is a fantastic independent Swiss cloud provider. They offer all the vital basics: Managed DBs, S3-compatible object storage, VMs, and managed Kubernetes.

• Highlights: Everything is available across multiple Swiss locations, making it an absolute home run for local data sovereignty. Furthermore, their support is top-tier. Even for a startup-sized entity like us, that famous "Swiss quality" really shines through.
• Trade-offs: Their managed Kubernetes offering isn't quite as technologically mature as the hyperscalers. For example, there is no managed Docker registry and no built-in autoscaler, which translates to more upfront engineering effort on our end. Additionally, an AI tokens-as-a-service model is currently nonexistent.

OVHcloud (The European Giant)

OVH sits somewhere between Exoscale and AWS.

• Highlights: The technology stack maturity - especially their Managed Kubernetes - gives you basically everything you need for a scalable, cloud-native stack out of the box. Crucially for a growing company, their startup program is phenomenal. It was incredibly simple to sign up and secure generous credits that essentially paid for our infrastructure for a year. They also offer AI tokens-as-a-service based on open-weight models (which we are actively investigating).
• Trade-offs: The support quality leaves a bit to be desired compared to Exoscale, though we haven’t faced any major issues so far. They don't have Swiss data centers, but they are a perfect fit if broader EU sovereign hosting is your goal.

The Octigen Way

So, where did we land? We launched Octigen using a combination of OVH and Exoscale.

By bridging Exoscale’s undeniable Swiss data sovereignty and incredible support with OVH’s mature Kubernetes capabilities and startup-friendly ecosystem, we built a highly robust, sovereign foundation for our services.

But here is the most important lesson we learned: You shouldn't have to lock yourself into a single vendor's ecosystem.

Because we engineered Octigen from day one to be strictly cloud-agnostic and Kubernetes-native, we aren't bound to Exoscale or OVH. We follow a highly flexible deployment model that allows us to meet our clients exactly where their needs dictate.

If you require strict EU/Swiss data sovereignty, we have you covered. If your enterprise is already deeply embedded in AWS, Azure, or GCP, we can run there. And if your security posture mandates that everything stays entirely in-house? We can deploy Octigen directly into your private cloud or internal, on-premise data centers.

We did the heavy lifting of evaluating the cloud landscape so that you don't have to compromise on security, AI capabilities, or performance.

Curious about how Octigen can fit into your specific infrastructure needs? Let’s talk about building a scalable, secure solution that works for you.