We have a new infrastructure partner. Verda, a European AI cloud provider, runs NVIDIA GPUs in Europe, powered entirely by renewable energy. Since July, Verda has been part of our pool of compute providers, expanding our sovereign AI infrastructure with sustainably powered compute capacity.
Why Verda
We had a short list of requirements. Verda met all of them.
Compute had to stay in the EU. Not "EU data residency available on request." Physically in the EU, under EU law, with zero US jurisdiction exposure.
Power had to be clean. We were not going to run AI on fossil fuels. Our customers ask about sustainability, and so do we.
Hardware had to be current. Real NVIDIA GPUs with InfiniBand interconnects, proper storage. Not last-generation chips repackaged as "AI-ready."
Verda checks every box. SOC 2 Type II compliant, NVIDIA Preferred Partner, running workloads for Nokia, Magnific, and 1X. They raised $117 million in April to expand their business. We talked to a few providers before landing on Verda. What sold us was the InfiniBand clusters. Most providers offer standalone GPU instances. Verda offers real clusters with proper interconnects, which matters when you run production workloads.
What changes for you
What changes is the layer below. Some of the GPUs that process your prompts, your fine-tuning jobs, your embeddings, and your coding sessions now run on Verda hardware in Europe. We are hosting GLM 5.2 and DeepSeek V4 Flash 0731 on Verda, giving you more sovereign capacity for two of the most capable open-weight models available. Powered by renewable energy, under EU jurisdiction. No US CLOUD Act, no data leaving Europe.
You still get Zero Data Retention. You still get open-weight models. Now running on European green energy, on hardware that meets our standards.
Our upcoming coding subscription runs on the same infrastructure. Verda backs it with sovereign GPU capacity for GLM 5.2, so developers get unlimited access to the latest open-weight coding model without sending their code to US providers.
Why we did this
European companies have been stuck between two bad options. US hyperscalers that are subject to the CLOUD Act, meaning your data is not really sovereign even when stored in Frankfurt. Or buying your own GPUs, which is expensive, slow to procure, and hard to scale.
We built Melious to be the third option. Sovereign AI in Europe, Zero Data Retention, no US exposure. Verda backs us with the hardware to deliver that at scale, alongside our existing providers, on clean energy and current hardware.
Rollout
We are routing some workloads to Verda over the coming weeks. The coding subscription opens for early access soon. If you already use Melious, you do not need to do anything. Your traffic moves gradually, not overnight. If something breaks, you hear from us first.
Build on sovereign European AI
Open-weight models. Zero Data Retention. European GPUs on renewable energy.