Quelles sont les heures d'ouverture ?

Lundi 08:00 - 18:00 Mardi 08:00 - 18:00 Mercredi 09:00 - 18:00 Jeudi 08:00 - 18:00 Vendredi 09:00 - 18:00

exoscale

The Safe home for your Cloud Applications.

24/06/2026

Choosing between AI performance and strict data compliance? That’s all water under the bridge now. 😉 Exoscale Dedicated Inference is officially out of preview and live for production-grade workloads. 🚀

You can now deploy any Hugging Face model as a production-ready, OpenAI-compatible API endpoint, backed by the security of a fully sovereign European infrastructure AND powered by dedicated NVIDIA GPUs

What this means for your production environment:

𝗔𝗯𝘀𝗼𝗹𝘂𝘁𝗲 𝗜𝘀𝗼𝗹𝗮𝘁𝗶𝗼𝗻:
Dedicated instances mean zero resource sharing. Your proprietary data and prompts never leave your environment.

𝗙𝗿𝗶𝗰𝘁𝗶𝗼𝗻𝗹𝗲𝘀𝘀 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻:
Swap in your preferred models and start querying immediately through standard API frameworks.

𝗘𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗦𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆:
Production workloads are now fully covered by a 99.95% SLA, with operational and support processes fully integrated into the standard Exoscale service lifecycle.

𝗩𝗲𝗿𝘀𝗮𝘁𝗶𝗹𝗲 𝗔𝗜 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲:
Fully optimized not just for LLMs, but also for complex embeddings, RAG applications, AI agents and custom inference APIs.

Our documentation has been completely expanded with deployment scaling guidance, updated CLI references service boundaries and model compatibility requirements.

Learn more: https://changelog.exoscale.com/en/ai-dedicated-inference-is-now-generally-available

23/06/2026

Already using Layer5 for cloud-native learning? You can now find Exoscale there too. ☁️

Engineering teams can access Exoscale learning paths, self-paced workshops and practical resources directly through the Layer5 ecosystem.

That includes material for getting started with us, hands-on workshop content around deploying microservice applications and visual cloud architecture resources for teams working with Infrastructure as Code.

Useful for teams who want to:
• build cloud fundamentals
• onboard new engineers
• work through infrastructure topics hands-on
• keep learning by doing, without putting project work on hold.

Start with the Exoscale Academy on Layer5: https://exoscale.layer5.io/academy

15/06/2026

We are cutting prices by a permanent 30% across all A40 (GPU3) instances in our Frankfurt zone. 🔻

Running active instances? Zero action needed from you. Your environment keeps running uninterrupted, and the lower rate applies to your billing automatically from June on.

Not on board yet? Why you should choose the NVIDIA A40 (GPU3):

• Built with 10,572 CUDA cores and 336 Tensor cores per card to handle intense data parallelization.
• Packing 48 GB GDDR6 memory per card, allowing you to fine-tune and run 13B to 34B parameter LLMs comfortably in production.
• Scale your instances dynamically from 1 to 8 GPU cards, backed by 12 to 96 CPU cores and 56 to 448 GB of RAM per instance.
• Access a maximum of 1.6 TB of NVMe local SSD storage for lightning-fast training data ingestion and checkpointing.
• Works out-of-the-box with Compute, SKS (Managed Kubernetes) and Docker NVIDIA.

𝗧𝗵𝗲 𝗻𝗲𝘄 𝗿𝗮𝘁𝗲𝘀 (𝗽𝗲𝗿 𝗵𝗼𝘂𝗿):
🔺 Small (1 GPU): ~~1.4933~~ ➡️ 1.0453
🔺 Medium (2 GPUs): ~~2.9867~~ ➡️ 2.0907
🔺 Large (4 GPUs): ~~5.9733~~ ➡️ 4.1813
🔺 Huge (8 GPUs): ~~11.9467~~ ➡️ 8.3627

Optimize your AI workloads and view the full changelog here:
🔗 https://changelog.exoscale.com/en/gpu-a40-gpu3-price-reduction-30-off-from-june-2026

09/06/2026

If you believe in a secure, independent European cloud, your voice just got a direct line to the ballot. 🏆
Thanks to the community of builders, partners and clients who trust us with their infrastructure every day, we’ve been nominated for the Cloudcomputing Insider Award 2026 in the Sovereign Cloud category.

Now, we need your help to cross the finish line and bring this trophy home.

Because your time is valuable, we’ve mapped out a Quick Guide in the carousel below to make voting effortless.

1. You don't have to fill out a long survey. You can completely skip the other categories and jump straight to the final section.

2. Just for participating, the publisher automatically enters you into a raffle to win one of three pairs of high-end Teufel Bluetooth headphones! 🎧

Swipe through to see exactly how to lock in your vote. Let’s win this together!
🔗 https://www.cloudcomputing-insider.de/award

Disclaimer: This raffle is organized and administered exclusively by the publisher (Vogel IT-Medien GmbH). Facebook is in no way sponsored by, endorsed by, administered by, or associated with this promotion.

05/06/2026

RFPs, procurement questionnaires and tender documents often contain sensitive company knowledge: security details, pricing logic, customer references, roadmap information, and contract language.

That makes uncontrolled AI workflows difficult to adopt in regulated or enterprise environments.

is now live on the Exoscale Marketplace. 🚀

Its AI agent, Ziva, helps teams:

• find and qualify relevant tender opportunities on regional and global tender portals, such as SIMAP or EU Tender Portal

• draft RFP and questionnaire responses from approved company knowledge

• include source citations for review

• track edits, approvals, and version history

• convert qualified opportunities into response projects

• use a Switzerland-based deployment option on Exoscale infrastructure when data residency matters

Zephior runs as a managed SaaS subscription. For customers with Swiss data-residency requirements, the deployment can run on Exoscale infrastructure in Switzerland.

Marketplace listing: https://www.exoscale.com/marketplace/listing/zephior-ai-procurement/

03/06/2026

If you build your core AI pipeline on a commercial API, you don't own your product lifecycle. You are merely just waiting for a deprecation notice. 🤔

On the blog, we show the exact engineering setup used to migrate a production workload off commercial APIs to a sovereign European deployment.

Swipe through to see the benchmarks and engine configurations. 👋

The backstory is a predictable one: In January, a software company running a production document pipeline got a notification that their core model was being deprecated. No migration path, no extended support. The consequence was a forced shift to an alternative that didn't hit their quality benchmarks.

To fix this dependency, the CIEL project - a collaborative R&D initiative between , -VD, and us - moved the entire workload to dedicated sovereign infrastructure.

The benchmarking data killed a major myth: sovereign cloud is not a performance downgrade. On real production documents, open-weight models beat the commercial baseline. The Qwen3.5 family hit a 0.80 similarity score vs GPT-4.1-mini’s 0.72.

It also fixed the financial trap of per-token pricing. When your structured few-shot prompts average 30,000+ tokens, every prompt edit changes your unit economics. Shifting to dedicated GPU instances completely decoupled operating costs from prompt length.

The full step-by-step migration checklist and vLLM configuration files are live on the blog. Find the link in the comments. 👇

01/06/2026

Attending Cloud Native Zürich on June 10?
After the first day of the conference, we are bringing the community together for an exclusive side event. 🤝

Together with SUSE, we want to bring the cloud-native community to the table. High above the roofs of Zürich! Rooftop BBQ, cold drinks and conversations on practical solutions are all part of the evening. 🌆

𝗪𝗵𝗮𝘁’𝘀 𝗱𝗿𝗶𝘃𝗶𝗻𝗴 𝘁𝗵𝗲 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻:
Data control, vendor lock-in and long-term flexibility are top of mind for many European organizations. We want to turn these challenges into a practical exchange focused on real solutions.

𝗧𝗼𝗽𝗶𝗰𝘀 𝘄𝗲'𝗹𝗹 𝗲𝘅𝗽𝗹𝗼𝗿𝗲:

🛡️ 𝗦𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻 𝗦𝘁𝗮𝗰𝗸 – building a strong, independent foundation with SLES and Exoscale
⚙️ 𝗥𝗲𝗮𝗱𝘆 𝗳𝗼𝗿 𝗥𝗮𝗻𝗰𝗵𝗲𝗿 – managing Kubernetes consistently across environments with SUSE Rancher & Exoscale SKS

Whether you're deep in implementation, scaling platforms, or shaping strategy: This is a space to meet like-minded people and spark ideas you can take back with you after the event.

🎟️ Spots are limited to keep the setting personal: send a quick DM!

👉 Details & agenda:
https://www.a1.digital/events/suse-exoscale-a1-digital-cloud-native-zurich-sideevent/

29/05/2026

A sovereign European cloud is not a technical downgrade.Thomas Perelle and the team at Gravitek just proved it. 🔺

They built a fully functional, production-grade Internal Developer Platform (IDP) running entirely on Exoscale SKS Pro for under €100 a month.

They put our new managed Karpenter integration to the test to see if a 100% European stack could handle real-world platform engineering pressure, without legacy provisioning hacks or giving up automation.

The architecture they deployed is clean:

𝗧𝗿𝘂𝗲 𝘀𝗰𝗮𝗹𝗲-𝘁𝗼-𝘇𝗲𝗿𝗼 𝗮𝘂𝘁𝗼𝘀𝗰𝗮𝗹𝗶𝗻𝗴:
Workload pools drop to zero nodes when idle.

𝗭𝗲𝗿𝗼 𝗽𝘂𝗯𝗹𝗶𝗰 𝗮𝘁𝘁𝗮𝗰𝗸 𝘀𝘂𝗿𝗳𝗮𝗰𝗲:
They used the Tailscale Operator instead of traditional ingress, completely eliminating public internet endpoints and extra load balancer costs.

The full architectural breakdown and configuration details are live on the blog. Take a look at what they built and let us know what your current cluster setup looks like: https://www.exoscale.com/blog/building-an-idp/

27/05/2026

Managed Thanos is LIVE on Exoscale 🚀

We designed it for critical workloads where long-term stability is an operational requirement.

By using low-cost Object Storage for the backend, you get years of history instead of days. It provides a single global view across all your clusters, while we handle the managed control plane and scaling.

It is a rock-solid alternative built for production-first environments.

Technical information here: https://www.exoscale.com/dbaas/thanos/

20/05/2026

We are attending Code Craft, DevOps Days, and SaaStanak next! Our goal: reeeeaaaal good conversations regarding critical infrastructure. Will you join Marcel Judex, Sebastien Pittet or Iva Bosotina? 🤛

If you are managing workloads where production comes first, find us at our booths or in the hallways.

What to ask us there? Take a look at the pictures to find out :)

Adresse

Boulevard De Grancy 19A
Lausanne
1006

Heures d'ouverture

Lundi	08:00 - 18:00
Mardi	08:00 - 18:00
Mercredi	09:00 - 18:00
Jeudi	08:00 - 18:00
Vendredi	09:00 - 18:00

Site Web

https://www.exoscale.com/

Notifications

Soyez le premier à savoir et laissez-nous vous envoyer un courriel lorsque exoscale publie des nouvelles et des promotions. Votre adresse e-mail ne sera pas utilisée à d'autres fins, et vous pouvez vous désabonner à tout moment.