AIARCO Atlas — Compute Cloud
GPU + CPU + bare metal compute, parallel storage, managed Kubernetes and a batch scheduler. Per-second billed.
What it is
Atlas is the AIARCO compute cloud. It runs your inference, training, batch, and long-running agent workloads on dedicated GPU and CPU pools in apac, use and euw regions. Atlas exposes 13 primitives, each addressable through the same control plane API.
When to use it
- Pick Atlas when you need GPUs (H100 / H200 / B200 / B300 / GB200 / GB300 / A100 / L40S), bare metal, or PB-scale parallel storage.
- Pick Edge when you need a globally-distributed, request-driven workload (CDN, DNS, edge functions).
- Pick Shield when you need a security layer in front of an existing origin.
Quickstart
asc atlas regions list
asc pods run --gpu h100 --image ghcr.io/me/train:latest --gpus 8client.pods.run(image='ghcr.io/me/train:latest', gpu='h100', gpu_count=8)await client.pods.run({ image: 'ghcr.io/me/train:latest', gpu: 'h100', gpuCount: 8 });Limits & quotas
| Limit | Default | Burst | Notes |
|---|---|---|---|
| Default GPU quota (per region) | 32 GPUs | 256 | Lift via sales@aiarco.com |
| Default object-storage egress to other clouds | 1 TB/day | — | Egress to Atlas compute in same region is free |
| Pod max lifetime | 30 days | — | Use Agents for persistent workspaces |
Pricing
See pricing. Pay-as-you-go, billed monthly via Stripe.
API surface
/v1/pods— long-running container workloads/v1/storage— object storage/v1/volumes— block volumes/v1/atlas/k8s— managed Kubernetes- See per-primitive pages below for full surface
Security
All access is authenticated and scoped. See Auth & scopes and Network controls.