Skip to main content

​
Reserved Clusters FAQ

Find answers to common questions about reserved clusters, contracts, and enterprise features.

​
Pricing & Contracts

Typically 8 GPUs for 1 month, though we can accommodate specific needs. Contact our sales team to discuss options for smaller or custom configurations.
No setup fees. You only pay for the compute time at your negotiated rate.
Reserved clusters typically start at $10K/month. This varies based on GPU type and configuration. Contact sales for a custom quote.
Discounts are based on your specific configuration and contract terms—typically ranging from 5-10%. Factors include GPU count, commitment length, and configuration complexity. Contact sales for a custom quote tailored to your requirements.
Billing terms are tailored to each customer. Typically, we require a down payment upfront, followed by recurring payments. We support ACH, wire transfer, and credit card payments. Contact sales to discuss billing arrangements for your specific situation.

​
Modifications & Scaling

Yes, you can scale up at any time. Scaling down requires 30-day notice. Changes to your configuration (GPU types, networking) can be discussed with your account manager.
Reservations cannot be paused, as the hardware is dedicated to you. However, we offer flexible scheduling options for batch workloads that don’t run 24/7.
You can use our On-Demand GPU service for burst capacity alongside your reserved cluster. Talk to your account manager about hybrid arrangements.
Yes, we support heterogeneous clusters. This is useful for workloads that benefit from different GPU types for different stages (e.g., Blackwell for training, H100s for inference).

​
Reliability & Support

We maintain spare capacity and guarantee 99.5% uptime. Failed hardware is automatically replaced, typically within hours. For critical workloads, we can configure automatic failover.
Our standard SLA guarantees 99.5% uptime, response within 2 hours for critical issues, and service credits if we miss targets. Custom SLAs are available for enterprise customers.
You get a dedicated Slack channel with direct access to our engineering team. Response times vary by severity—critical issues get sub-15-minute response, while general questions are answered within 24 hours.
Phone support is available for critical issues (cluster down, major incidents). For day-to-day support, Slack is the primary channel for fastest response.

​
Technical Questions

Yes, you have full control over the software stack. You can install any frameworks, libraries, or tools you need. We provide a base image or you can bring your own.
Yes, we support self-managed Kubernetes deployments. You have full control over your Kubernetes configuration, and we provide the underlying infrastructure. Our team can assist with setup and optimization for GPU workloads.
We support InfiniBand (400Gb/s), RoCE (200Gb/s), and NVLink for GPU-to-GPU communication. See our Configuration guide for details.
Yes, we support VPN connections, private peering, and custom network configurations to integrate with your existing cloud or on-premise infrastructure.
We offer SOC2 and HIPAA-ready configurations. Contact sales to discuss specific compliance requirements for your organization.

​
Getting Started

Standard deployment is 3-5 days from contract signing. This includes hardware provisioning, network configuration, and software setup. Rush deployments may be possible—ask your account manager.
We’ll need to understand your workload requirements (GPU type, count, networking), commitment length, and any special requirements (compliance, custom configurations). See our Getting Started guide for what to prepare.
We recommend using our On-Demand GPU service to test your workloads before committing to a reserved cluster. This helps ensure the configuration meets your needs.

​
Still Have Questions?

Contact Sales

Get answers from our sales team