Gcore Launches GPU Virtual Machines on NVIDIA Hopper

Why It Matters

The offering lowers entry barriers for AI workloads in Europe, providing sovereign, on‑demand high‑performance compute while eliminating idle‑GPU expenses. This flexibility accelerates experimentation and short‑term projects, strengthening the region’s AI ecosystem.

Key Takeaways

• Gcore offers Hopper GPU VMs in Portugal sovereign region
• Pay‑only‑while‑active model eliminates idle GPU costs
• Scale from 1 to 8 GPUs instantly via portal
• Same InfiniBand networking as bare‑metal GPU cloud
• Targets EU startups, R&D labs, research institutions

Pulse Analysis

European enterprises are increasingly demanding AI infrastructure that can be provisioned on‑demand and remain compliant with data‑sovereignty rules. Gcore’s decision to launch Hopper‑based GPU VMs in its Sines‑3 region directly addresses this need, offering a cloud‑native alternative to traditional bare‑metal clusters. By situating the service within a sovereign zone, Gcore not only satisfies regulatory concerns but also positions itself as a regional competitor to global hyperscalers that often host data outside the EU. This strategic placement could attract firms wary of cross‑border data flows while still requiring cutting‑edge compute.

Technically, the VMs deliver the same Hopper GPU performance and NVIDIA Quantum InfiniBand bandwidth found in Gcore’s bare‑metal GPU Cloud, but with the elasticity of virtualized resources. Users can start with a single GPU for prototype work and seamlessly expand to eight GPUs for large‑scale training, all through a unified portal. The pay‑as‑you‑go model pauses GPU billing when instances are shut down, meaning organizations only incur storage and IP costs during idle periods. This granular cost control is especially valuable for early‑stage startups and research labs that run intermittent, high‑intensity fine‑tuning jobs.

The broader market impact could be significant. As AI development cycles shorten, the ability to spin up powerful GPUs for brief bursts without capital expenditure becomes a competitive advantage. Gcore’s flexible offering may pressure other providers to introduce similar VM‑based GPU services, intensifying price competition and driving innovation in billing models. For European AI players, the combination of sovereign hosting, Hopper performance, and cost‑efficient elasticity could accelerate product timelines and foster a more vibrant AI ecosystem across the continent.