On-Demand GPU Instances: Per-Minute Billing, No Contracts.
Pick from 10+ NVIDIA and AMD GPU models and pay only for what you use, with NVIDIA instances live in under 2 minutes. Per-minute billing means you stop paying the moment you're done. On-demand covers both dedicated instances (99.99% SLA, non-interruptible) and spot instances (interruptible, up to 50% off).
Deploy in Three Steps
From zero to a running GPU instance in under a minute. No provisioning queues, no approval workflows.
Pick Your GPU
Browse available GPU models with live pricing. Filter by VRAM, architecture, or price to find the right fit for your workload.
Configure and Launch
Choose your region, storage, and OS image. Hit deploy. NVIDIA instances are provisioned and accessible within 2 minutes.
Build, Train, Iterate
SSH in and start working. Scale up to more powerful GPUs or spin down when you're finished. You're billed by the minute.
Built for Speed and Flexibility
Per-Minute Billing, Zero Waste
Most cloud providers bill by the hour. We bill by the minute. Run a 23-minute training job? Pay for 23 minutes. No rounding up. Every instance carries a 20-minute minimum runtime, NVIDIA included, and a 2x AMD MI300X carries 60 minutes.
10+ NVIDIA and AMD GPU Models
From RTX 4090s for prototyping to H100s for production training, every major NVIDIA architecture is available, plus the AMD MI300X with 192GB of HBM3. Switch between models without penalties.
Full Root Access, Your Way
Every instance comes with NVMe SSD storage and full root access, plus a dedicated IP on NVIDIA instances. Install your frameworks, mount your datasets, run your stack your way.
Global Data Center Network
Deploy in regions closest to your data. Our vetted data center partners span multiple continents with Tier 3/4 certification and enterprise-grade networking.
Available GPU Models
Every major NVIDIA architecture, from budget-friendly RTX 4090s to flagship B300s, plus the AMD MI300X. All available on-demand with per-minute billing.
Compare Instance Types
Dedicated
CurrentOn-demand instances with a 99.99% SLA. Non-interruptible. Deploy instantly, pay by the minute.
Production workloads, long-running training, critical inference
Spot
On-demand instances at a discount using idle capacity. Same hardware, up to 50% off, interruptible.
Batch jobs, experiments, hyperparameter tuning, flexible workloads
Reserved
Locked capacity with volume pricing. Custom clusters and dedicated support.
Large deployments, multi-month projects, enterprise teams
On-Demand GPU FAQ
On-demand is the umbrella for both tiers. Dedicated instances carry a 99.99% SLA and cannot be reclaimed by the provider, so they are the right fit for production workloads and long-running training. Spot instances use idle capacity at up to 50% off the dedicated rate, but can be terminated when demand is reclaimed. Both bill per minute and share the same GPU hardware.
Yes, a short one. Every instance carries a 20-minute minimum runtime, and a 2x AMD MI300X carries 60 minutes. Past the minimum Spheron bills by the minute, so a 45-minute job costs 45 minutes. There is no contract and no long-term commitment. The AMD MI300X can't be stopped, only destroyed.
We offer 10+ NVIDIA models including RTX 4090, RTX 5090, RTX PRO 6000, L40S, A100, H100, H200, GH200, B200, and B300, plus the AMD MI300X (192GB HBM3, ROCm). New models are added as they become available from our data center partners.
Most instances are ready in under 2 minutes. Select your GPU, region, and configuration, then hit deploy. No approval process, no provisioning queue.
Yes. Spin down one instance and launch another model at any time. There are no lock-in periods or penalties for switching. Use an RTX 4090 for development and scale up to an H100 for production training.
Everything you need: a fully provisioned instance with your selected GPU, NVMe SSD storage, network bandwidth, full root access, and a dedicated IP address on NVIDIA instances. No hidden fees.
Deploy in 60 Seconds
No contracts. No commitments. Deploy your first GPU instance now and pay only for the minutes you use.