Transparent pricing for world-class AI infrastructure. Prepaid CPU and GPU compute with unused time returned, and inference, storage and networking invoiced every 30 days.
$600 in launch credit + 20M free inference tokens
Double the typical big-cloud free trial. Applies to inference, storage and networking in every region.
Free
$600 in launch credit, 20M free inference tokens, every region.
Pay as you grow
Per-second metering, priority capacity, 24/7 engineering desk.
Custom
Dedicated fabric, air-gapped regions, named mission team.
Data preparation, retrieval, vector search, orchestration, agents, caches and serving run on CPUs. Pick a family, an instance type, a count and the hours you need, prepaid with unused time returned.
| Instance | vCPU | Memory | vCPU type | Per hour |
|---|---|---|---|---|
| General purpose | ||||
| General 4 vCPU · 8 GB | 4 | 8 GB | Shared | $0.09/hr |
| General 8 vCPU · 32 GB | 8 | 32 GB | Shared | $0.29/hr |
| General 16 vCPU · 64 GB | 16 | 64 GB | Shared | $0.58/hr |
| Compute optimized | ||||
| Compute 8 vCPU · 16 GB | 8 | 16 GB | Dedicated | $0.29/hr |
| Compute 16 vCPU · 32 GB | 16 | 32 GB | Dedicated | $0.58/hr |
| Compute 32 vCPU · 64 GB | 32 | 64 GB | Dedicated | $1.15/hr |
| Memory optimized | ||||
| Memory 8 vCPU · 64 GB | 8 | 64 GB | Dedicated | $0.45/hr |
| Memory 16 vCPU · 128 GB | 16 | 128 GB | Dedicated | $0.89/hr |
| Memory 32 vCPU · 256 GB | 32 | 256 GB | Dedicated | $1.79/hr |
Rates are per instance-hour. Explore CPU compute
GPUs accelerate training and heavy inference. GPU hours are prepaid, paid time only counts while your cluster is running, and unused time returns to your compute balance.
| Accelerator | Memory | Per hour | Per second |
|---|---|---|---|
| NVIDIA B200 192GB | 192 GB HBM3e | $7.79/hr | $0.002164 |
| NVIDIA H200 141GB | 141 GB HBM3e | $7.19/hr | $0.001997 |
| NVIDIA H100 80GB | 80 GB HBM3 | $2.99/hr | $0.000831 |
| NVIDIA A100 80GB | 80 GB HBM2e | $1.29/hr | $0.000358 |
| NVIDIA L40S 48GB | 48 GB GDDR6 | $0.99/hr | $0.000275 |
| RTX 4090 24GB | 24 GB GDDR6X | $0.59/hr | $0.000164 |
Storage
$0.016 / GB-month
Object storage and persistent volumes, encrypted by default.
Data egress
$0.009 / GB
Simple, flat egress pricing from every region.
Inference
20M free tokens, then per token
Per-model token rates are shown in the dashboard before you call them.
CPU and GPU time are prepaid, so there are never surprises on compute. Inference, storage and networking accrue in real time on your dashboard and settle by itemized invoice every 30 days.
Build
Prepay CPU and GPU hours and launch. Inference and storage draw on launch credit first.
Meter
Compute time and per-token inference are tracked in real time on your dashboard.
Settle
An itemized invoice is issued every 30 days for the usage in that cycle.
Grow
Scale up or down at any time. No commitments, and unused compute time returns to your balance.
Everything you need to know about prepaid CPU and GPU compute, credits and monthly invoicing.
CPU and GPU time are both prepaid. Choose a CPU instance type or GPU, how many and how many hours you want, then pay upfront or from your compute balance, and your instances launch. Paid time only counts while they are running — never while queued or self-healing — and any unused time returns to your compute balance when you delete them. Turn on auto-renew to top up hours from your balance automatically.
Inference is metered per token, and storage and networking per unit used, then settled every 30 days by an itemized invoice (revenue-share accounts: every five days). CPU and GPU compute are prepaid, so they never appear as a surprise. Short cycles mean you always know exactly where you stand, with no surprise month-end bills.
If a balance passes its due date, access is limited to the billing and overview pages so you can settle up quickly. Everything else is restored automatically once the full balance is paid.
Usage-based accounts receive an itemized invoice every 30 days, emailed to you and payable online in one click. Revenue-share accounts can pay any amount at any time — partial payments are credited immediately, and the full balance restores access to all features.
Every new account receives $600 in launch credit plus 20M free inference tokens, double the typical big-cloud free trial. Credit applies automatically to inference, storage and networking; CPU and GPU compute hours are prepaid separately.
Never. Prepaid CPU and GPU time only counts while your instances are running, and anything unused returns to your compute balance for your next workload.
Yes. Scale and Sovereign customers with predictable workloads can reserve capacity at tailored rates. Talk to our team for a proposal.
Claim $600 in launch credit and 20M free inference tokens. CPU and GPU compute in every region, no commitment.
Double the typical big-cloud free trial · Per-second metering · Live in minutes