CloudSigma

What CloudSigma actually is

CloudSigma is a Swiss IaaS provider, founded in Zurich in 2009 by Robert Jenkin and run today out of Zug as CLOUDSIGMA AG. It built its name on unbundled cloud servers — you size CPU, RAM, storage and traffic independently instead of picking an instance shape — and on a per-resource billing model that reads your aggregate consumption every five minutes. Sixteen years on, the company is in the middle of a significant pivot: the front door at cloudsigma.com now sells a white-label "sovereign cloud" platform to telcos, MSPs and data-center operators, who run CloudSigma's stack under their own brand across the partner network. Individual customers can still sign up and run VMs in the direct clouds, but the strategic direction is partner-delivered.

The direct product remains what it was: KVM or VMware virtualisation, any OS including custom images, root access, free-form sizing up to 40 CPU cores and 128 GB RAM per server, and — this is the signature detail — CPU measured in GHz rather than fixed cores. One virtual core equals 2 GHz by default, and you decide how to slice it: a 10 GHz server can be ten 1 GHz threads or four 2.5 GHz threads depending on whether your workload wants parallelism or per-core speed. Storage is NVMe block plus Amazon S3-compatible object storage (HDD, NVMe and caching tiers), and the platform exposes a REST API with Terraform and Ansible support. Snapshots, backups, VLANs, VPN, NAT, firewalls and load balancing are all in the toolbox. On the newer partner-facing side there is GPU compute (passthrough, vGPU and MIG slices) and TaaS, an OpenAI-compatible gateway to 40-odd models with per-token pricing.

Our editorial read: CloudSigma has always been the tinkerer's cloud — more configuration freedom than the big providers, priced per resource, headquartered in a privacy-friendly jurisdiction. The question for new customers is no longer whether the platform is capable; it is how much attention a self-managed individual account gets while the company chases its partner business.

Pricing: what you really pay

CloudSigma runs two purchasing models side by side: subscriptions (bought upfront, cheaper) and burst (pay as you go), with burst rates automatically covering anything above your subscribed levels. The gap between the two is large. A third-party 2026 review by C9 Group priced a small starter workload in Geneva at €34.19 per month on a one-month subscription versus €151.71 pay as you go, and a bigger production workload (compute, disk and IPv4 only) at €342.69 subscription versus €1,621.68 pay as you go. In our view that spread is the single most important pricing fact about CloudSigma: if you run steady workloads on burst rates, you overpay several-fold compared to the same capacity under subscription.

Egress is comparatively generous: the same review documents 5 TiB per month free per account in Geneva and Frankfurt, 1 TiB in Dublin, then €0.0399 per GiB on subscription and up to €0.2193 per GiB on pay as you go. The vendor's own pricing material lists example bundles — one virtual core, 2 GB RAM, 50 GB NVMe and 5 TB of transfer as the small shape, scaling up through Medium and Large tiers to 20 cores with 64 GB RAM and 640 GB NVMe — alongside the promise that free IOPS, firewalls, API calls and 24/7 support are never charged. There is a seven-day free trial, and a free resource tier that automatically applies to accounts spending over USD 10 per month, except in Honolulu, Perth and Melbourne.

What we like is the granularity: per-resource, per-five-minute metering with a near-real-time balance, no minimum commitment on burst, and a live quoting tool at pricing.cloudsigma.com. What we dislike is that the per-unit rates are no longer legibly published the way they once were — the marketing pages have pivoted to partner economics, and you end up building a quote to find the real number. That is a transparency regression for a provider whose identity was transparent pricing.

Performance & infrastructure

CloudSigma's direct clouds are individual locations rather than one global network: each region is effectively its own cloud with its own account and contract. The locations listed in the vendor's pricing tool include Zurich and Geneva in Switzerland, Boden in Sweden, Dublin in Ireland, Frankfurt in Germany, London in the UK, Riyadh in Saudi Arabia, Tokyo in Japan, Manila, Manila-2 and Clark in the Philippines, Honolulu, San Jose and Washington DC in the US, and Melbourne and Perth in Australia; the partner network claims 40+ regions across 35+ countries. Be aware that some regions are operated by partners under their own terms — Düsseldorf is run by enthus, Dublin by ServeCentric — and that Frankfurt and Düsseldorf were reported closed to new sign-ups in 2026.

The SLA is unusual and, on paper, the strongest shape in the market: 100% virtual server availability in any calendar month (hardware and hypervisor layer), 100% network availability, and a 1 ms latency guarantee between servers on the same network. Miss it and the credit is 50 times the fees for any outage lasting more than 15 minutes, capped at 100% of the previous 30 days' fees, claimable by email within 30 days. Read the fine print though: CloudSigma is the sole arbiter of claims, credits are against future service only, and DDoS attacks, upstream failures and law-enforcement activity are exempt. No SLA covers storage, the API or object storage. There are no availability zones anywhere — every European location reports none — so high availability means a second location, which means a second account and contract. The terms also disclaim responsibility for backups and cap liability at one month's fees.

On the operations side, the vendor promises 24/7 support with sub-one-minute response by chat and email. Third-party experience is less flattering: the C9 Group review found support on "reasonable endeavours" with no response targets, no paid support tier, and reported the ticketing system broken since late September 2026. A status page per location exists (Geneva's shows 37 incidents back to April 2019), but post-mortems go out by email rather than being published. We have not benchmarked CloudSigma ourselves, so we keep our own numbers on /benchmarks for the providers we actually measured.

Where it falls short

Start with the pivot. CloudSigma's homepage is now a partner pitch; self-service individual accounts are a legacy path that still works but is no longer where the roadmap points. For a solo developer or a small team that is a strategic risk worth pricing in — the platform will not disappear overnight, but feature velocity, support staffing and sales attention all flow toward the 35+ partner countries now. Then the architectural gaps: no availability zones, no managed PostgreSQL, no managed Kubernetes control plane, no object storage in every region (Geneva, for instance, does not sell it), and a per-location account model that makes cross-region HA a paperwork exercise as much as a technical one.

The pricing model deserves its own paragraph. Unbundled per-resource billing sounds fair and is, in steady state, but the pay-as-you-go versus subscription gap means the naive customer subsidizes the informed one by a factor of four or five. Missing a subscription renewal or running an unattended burst workload is expensive. And while the GHz-based CPU model is genuinely flexible, it also means published "vCPU" comparisons against other providers are meaningless without doing the arithmetic. Finally, the credit formula's "sole arbiter" clause and the one-month liability cap are contractual realities that sit oddly next to the marketing's 100% guarantee.

Who should pick this

Our pick for European teams with steady, self-managed workloads that want Swiss or EU data residency and per-resource control — scientific computing, media processing, agencies running fleets of custom-sized VMs, and anyone who genuinely uses the ability to shape a server as 24 threads of 500 MHz rather than four fat cores. The subscription pricing in Geneva, Frankfurt and Dublin is competitive with the mid-market VPS providers in /category/vps-providers once you commit, and the S3-compatible object storage plus Terraform support make it automatable. Organizations that want sovereignty narratives with local delivery should also look at the partner network in their country.

We would not pick it for workloads that need zone-level redundancy out of the box, for teams that want managed databases or Kubernetes, or for bursty, unpredictable traffic that cannot be subscribed. And we would not start a new project on a self-service account without first asking CloudSigma directly where individual accounts sit in the partner-first roadmap.

Key takeaways

• Swiss IaaS veteran (founded 2009, Zurich; CLOUDSIGMA AG in Zug), now pivoted to white-label sovereign cloud for service providers while still running direct clouds
• Truly unbundled resources with GHz-based CPU sizing, five-minute metering, and subscription vs burst pricing — the same workload can cost 4-5x more on pay as you go
• Strong SLA on paper: 100% VM and network availability with 50x-fee credits, but no zones, no storage/API coverage, and CloudSigma is sole arbiter of claims
• 16 direct locations across Europe, US, Middle East, Asia-Pacific; S3-compatible object storage, snapshots, Terraform/API; GPU and AI gateway (TaaS) on the partner side
• Real caveats for individuals: no managed databases or Kubernetes, per-location accounts, closed sign-ups in some regions, and third-party reports of support tooling problems in 2026

Sources

• https://www.cloudsigma.com/pages/about
• https://cloudsigma.com/
• https://cloudsigma.com/pages/cloud-servers
• https://cloudsigma.com/pages/business-model.html
• https://blog.cloudsigma.com/bg/pricing-dev
• https://pricing.cloudsigma.com
• https://fra.cloudsigma.com/ui/5.0/assets/legals/TBC/sla.pdf
• https://c9group.dev/en/articles/cloudsigma-european-cloud-review
• https://docs.cloudsigma.com/en/latest/virtualrouters.html