Products
AI infrastructure services for training and inference.
Compute access, private clusters, managed operations, and production serving. Each product answers a different question, and they work together as your program grows.
- Training
- GPU Cloud · Dedicated AI Clusters
- Inference
- Production Inference
- Operations
- Managed AI Infrastructure
- Supported by
- The ARC AI Platform
Products
What you can buy from ARC.
Every product runs on infrastructure ARC plans and engineers, and on the same AI Platform, so moving between them doesn’t mean starting over.
01
GPU Cloud
GPU compute without building the infrastructure.
Access to NVIDIA GPUs without building or operating the underlying infrastructure.
GPU instances suit development and variable demand. Bare metal servers give your team the whole host.
Best forTeams that need GPUs soon and want to choose their level of host control.
02
Dedicated AI Clusters
A private cluster, designed around your models.
A complete multi-node cluster engineered for your workloads and dedicated to your organization.
Tenancy is single-tenant: the servers, network, and storage serve only your organization.
Best forLarge training runs and organizations that need hardware isolation.
03
Managed AI Infrastructure
ARC operates the fleet. You run the models.
An operating service: ARC runs the infrastructure layers you hand over, under agreed responsibilities.
Applies to clusters ARC builds and to environments you already run. Responsibilities are agreed in writing first.
Best forTeams that want production reliability without staffing a fleet operations group.
04
Production Inference
From trained model to production service.
Everything needed to serve models in production, from GPU capacity to the APIs your applications call.
Includes integration with multiple model providers’ APIs alongside models deployed on ARC.
Best forTeams putting models in front of users.
Where to start
Find your starting point.
Pick the need closest to yours. Start with one product and add others as your program matures.
Recommended path
Start on GPU instances or bare metal servers. When demand becomes predictable, reserve capacity. When jobs span many nodes, move to a dedicated cluster.
Recommended path
We design the network, storage, and scheduler around your parallelism strategy, then validate the cluster with your own jobs before handoff.
Recommended path
ARC takes on monitoring, maintenance, recovery, and expansion under a written shared-responsibility model.
Recommended path
We size and tune the serving environment to your latency targets, concurrency, and cost per token.
Deployment options
How and where your environment runs.
These choices define access, tenancy, location, and who operates what. They apply across products.
| Choice | Options | What it determines |
|---|---|---|
| Access | GPU instances or bare metal GPU servers | Whether you work in a virtualized environment or control the whole host, including the operating system and drivers |
| Tenancy | Shared infrastructure or single-tenant clusters | Whether servers, network fabric, and storage are dedicated to your organization |
| Location | Set for each engagement | Where capacity runs, based on availability, data requirements, and your teams |
| Operating boundary | Self-managed, ARC-supported, or ARC-managed | Which layers ARC operates and which stay with your team |
Commercial terms
How you commit to capacity.
Commercial terms are separate from access and tenancy. Either can apply to instances, bare metal servers, or dedicated clusters.
Flexible terms
Capacity on terms that fit early work.
Shorter commitments suited to development, experiments, and demand that is still taking shape.
Reserved capacity
Hold capacity for an agreed term.
A commitment to a specific quantity, configuration, term, and start date, planned around expected usage.
Pricing and availability depend on system, quantity, location, and term, and are confirmed in each proposal.
Service stack
How the products fit together.
IaaS, PaaS, and MaaS describe layers, not separate catalogs. Infrastructure products sit at the base, the AI Platform manages resources and workloads, and model services run on top.
About the AI PlatformOperations
Managed AI InfrastructureARC can operate any of these layers, or your team can.Not sure which product fits?
Tell us the workload, the control you need, and when you need it. We’ll recommend a starting point and a path forward.
Support: support@arccloud.com
For product issues, service disruptions, or technical assistance.
Security: security@arccloud.com
To report vulnerabilities, suspicious activity, or other security concerns.