Case study
How a Swiss cloud service provider cut Azure waste in one workshop
With samcloud, Abraxas's engineers gained a repeatable, metric-driven approach to Azure cost optimisation, which they now apply across the CSP subscriptions they manage.
- Industry
- Cloud Service Provider / Managed Services (Switzerland)
- Date
- Workshop held 5 November 2025; follow-up steps agreed
Executive summary
Abraxas, a Swiss Cloud Service Provider moving towards a fully managed-services model, engaged samcloud to improve how its engineers optimise customer Azure subscriptions. In a single session, samcloud’s Stefan Denk worked with the Abraxas team on:
- Evidence-based right-sizing: VM sizing decisions grounded in real 30-day CPU and memory metrics at 1-minute granularity.
- Uncovering high-impact waste hidden from the standard Azure portal view: unattached storage and licence-heavy SKUs.
- Transferring a repeatable method: a checklist and a metric-driven process the team can apply across its customer base.
The outcome was a capability the team can reuse, and the customer reported reducing unnecessary Azure spend by “a significant percentage.”
Client profile
Abraxas is a Cloud Service Provider and managed-services provider operating in Switzerland. It manages a mix of customer Azure subscriptions that have grown over time through projects and lift-and-shift migrations, and is moving towards a “100% outsourcing” managed-services model. As Andi Schmid put it, the session “gave us an idea how we will approach FinOps on Azure as a Managed Services Provider.”
The scope
Before the engagement, the estate looked, in the customer’s own words, “normal”: mixed VM generations, oversized lift-and-shift disks and Windows workloads running on license-included SKUs. Azure's inbuilt CSP Cost Management hard to see the overall picture above subscription level, so optimisation was mostly reactive: “reacting when someone complained about cost.” The goal was to replace firefighting with a defensible, repeatable method the engineers could own.
The samcloud way
SAMcloud worked through the customer’s real subscriptions with the team.
- Phase 1
Measure before moving
Pulled 30-day CPU and memory metrics at 1-minute granularity. Confirmed real utilisation (a sample D4as v5 VM ran at roughly 5% average CPU with more than 60% memory consistently free).
Right-sizing decisions were based on evidence, so the team could move from D-series to E-series or B-series only where it was safe.
- Phase 2
Cut high-impact waste
Identified and fixed unattached 4 TB Standard SSDs. Reviewed Windows licensing against VM SKUs.
Recurring waste was eliminated, including a single licence change that cut cost by a factor of 10 on one workload.
- Phase 3
Hand over the method
Walked through the trade-offs of burstable VMs and Premium disk bursting for spiky workloads. Left the team with a reusable checklist and a clear process.
The engineers now have a repeatable, metric-driven method they can apply across their own customer subscriptions.
Conclusion
Stefan doesn’t push generic ‘best practices’. He walks through real subscriptions with you, explains the trade-offs, and leaves you with a checklist you can reuse.
After working with Stefan, we now have a clear process and can quantify the impact of each optimisation. During this workshop, we reduced our unnecessary Azure spend by a significant percentage.
Abraxas’s engineers can now do this work themselves across every subscription they manage.
Results & financial impact
| Workstream | Challenge | Outcome |
|---|---|---|
| Right-sizing | Oversized VMs from lift-and-shift; a sample VM at ~5% CPU with 60%+ memory free | Moves from D-series to E-series/B-series where safe; E2as v5 is roughly 40% cheaper than D4as v5 |
| Storage waste | Unattached disks hidden by CSP tooling | Two unattached 4 TB Standard SSDs found and fixed (~300 CHF/month each, ~600 CHF/month combined) |
| Windows licensing | Workloads on license-included SKUs | One change cut licence cost by a factor of 10 on a single workload (~$120 → ~$10/month) |
| Method transfer | Reactive, one-off fixes | Repeatable checklist and process; unnecessary Azure spend reduced by “a significant percentage” |
Deliverables
- Live, subscription-by-subscription optimisation walkthrough with metrics evidence.
- A reusable right-sizing and cost checklist owned by the Abraxas engineers.
Long-Term Value
Repeatable method
a metric-driven process the team applies across its customer base.
Practical work
the session covered real subscriptions and their trade-offs.
Knowledge transfer
a checklist the engineers reuse independently.
No lock-in
capability that stays with Abraxas as it scales its managed-services model.
Is your Azure optimisation still reactive?
We measure the estate, cut the waste and leave your engineers a method to reuse.
