Gcore GPU Cloud & Everywhere Inference

Offering overview

Two products make up Gcore's AI line. GPU Cloud supplies bare-metal clusters of eight NVIDIA L40S, A100, H100, H200, or B300 GPUs, plus virtual H100 instances billed only while powered on, with GB300 NVL72 racks available on request. Everywhere Inference is the serverless counterpart: pick a model from the library or upload your own, and Gcore deploys it on dedicated GPUs and uses its CDN network to steer requests to the nearest location. That edge routing is the differentiator for latency-sensitive applications and for compliance cases where requests must be served in a chosen region. Deployment takes a few clicks or an API and Terraform call, autoscaling and multi-tenancy adjust capacity across regions, and observability and real-time cost estimates are built in. Headquartered in Luxembourg, the company reports 50+ cloud locations and more than 210 points of presence on six continents. Pricing is per minute or per second of GPU time rather than per token. Everywhere Inference starts at EUR 1.08 per hour for a single L40S, while bare-metal GPU clusters begin at EUR 1.27 per GPU-hour for L40S and EUR 2.73 per GPU-hour for H100 SXM; all prices exclude VAT.

Technical specifications

Containers

Unknown

Kubernetes

Unknown

Root access

Unknown

API access

Unknown

Autoscaling

Unknown

Persistent processes

Unknown

AI capabilities

Model serving

Not stated

AI frameworks

Not stated

Inference endpoints

Unknown

Fine-tuning

Unknown

Vector database

Unknown

Model catalog

Unknown

Monitoring

Not stated

Evaluation

Not stated

Security & compliance

Geography

Provider headquarters are separate from offering infrastructure and availability.

Available plans

Pricing can change. Promotional prices are shown separately from standard prices.

Questions & answers

What is the difference between Gcore GPU Cloud and Everywhere Inference?

GPU Cloud provides bare-metal GPU clusters and virtual H100 instances. Everywhere Inference is a serverless model-serving layer that deploys models on dedicated GPUs and routes requests through Gcore's CDN network.

Can teams deploy their own models with Everywhere Inference?

Teams can select a model from the library or upload their own model. Deployment is supported through a few clicks, an API or Terraform.

Does Everywhere Inference support regional request routing?

It routes requests to the nearest location and supports serving requests in a chosen region for compliance-related use cases.

Which GPUs are offered by Gcore GPU Cloud?

Bare-metal clusters contain eight NVIDIA L40S, A100, H100, H200 or B300 GPUs. Virtual H100 instances are also offered, and GB300 NVL72 racks are available on request.

How does Gcore bill GPU usage?

Pricing is based on minutes or seconds of GPU time rather than tokens. Virtual H100 instances are billed only while powered on.