Giant Swarm

Giant Swarm started inside an ad-exchange company: its founders ran Adcloud, moved to microservices in 2011, and found that operating their own infrastructure consumed a third of engineering time. They turned that experience into Giant Swarm in 2014, and the company is now a GmbH registered in Cologne.

The product is not a bare cluster. The homepage describes a curated stack in which Kubernetes, observability, security, networking and application management come configured together, with Cilium, Kyverno, Flux and Karpenter pre-integrated. For teams comparing against OpenShift, this is the closest European analogue to a bundled platform rather than a cluster-only service.

Two ways to buy it

Giant Swarm offers a fully managed model, in which its engineers monitor the platform around the clock and apply upgrades and security patches, and a self-managed model, in which your own team runs the same stack with training and expert support. The page states that the second model gives no third party access to production systems.

That choice matters in regulated settings: a hospital or bank can keep operators in-house and still use configurations that Giant Swarm says come from more than 150 production clusters.

Where the clusters run

According to the homepage, clusters run on AWS, Azure or on-premises and can be managed from a single control plane. Giant Swarm itself does not sell the hosting, so the legal home of your data is the provider you pick.

A buyer who needs European-only infrastructure has to choose an on-premises or European-hosted target and confirm that Giant Swarm supports it; the pages I read named only AWS, Azure and on-premises.

Open source and exit

The vendor describes the stack as open source with no lock-in, and states that every component remains yours to keep. If the contract ends, the platform keeps running because it consists of community projects rather than a proprietary runtime.

The vendor also lists a CNCF AI-Conformance certification and GPU scheduling features for AI workloads, which matter if clusters will host model serving next to ordinary services.

Limitations to Consider

Pricing is not published; the website shows only a calculator that asks for a vCPU count. The infrastructure targets named on the pages I read are AWS, Azure and on-premises, so sovereignty depends on your own hosting choice. Figures on customer savings and cluster counts are the vendor's own statements and not independently checked. For a plain managed cluster without extra layers, Nine or Scaleway are simpler.

Who Should Use Giant Swarm

Platform engineering teams of mid-sized and large companies that would otherwise spend months wiring policy, observability and GitOps around a bare cluster, and that want an open source result they can operate in-house later.