Hyperbolic

Hyperbolic is an AI cloud platform from Hyperbolic Labs that gives teams self-serve access to GPU compute. Users can launch on-demand GPU instances, reserve dedicated clusters, or set up private cloud infrastructure for training, fine-tuning, inference, and production workloads. The homepage states it serves 250,000+ builders and offers an OpenAI-compatible API for serving models.

What It Is

Hyperbolic is a GPU cloud and compute marketplace. Teams create an account, add a payment method, choose a GPU configuration such as H100, H200, or B200, and connect over SSH. Workloads range from single-GPU experiments to multi-GPU and multi-node deployments. Compute is sourced through a global network of providers, and compute providers can list capacity through Forge and Hyperbolic's managed platform.

Three Ways to Get Capacity

• On-Demand: pay-as-you-go GPU instances with no commitment, suited to experiments, evaluations, burst workloads, and short training runs.
• Reserved Clusters: dedicated capacity at discounted rates with commitments under one year, aimed at steady production workloads.
• Private Cloud: dedicated long-term infrastructure with single-tenant hardware, isolated networking, SSO, audit logging, and compliance support, according to the FAQ.

Use Cases and Isolation

The site lists foundation-model training, fine-tuning and LoRA, batch compute jobs, long-running training, agent automation by programmatically launching GPU resources, and private infrastructure. On-demand instances run on multi-tenant infrastructure with network isolation and instance cleanup between sessions, while reserved and private tiers provide stronger isolation and control.

Recent Updates

The blog lists the introduction of Forge, an infrastructure layer for GPU provisioning (June 2026), on-demand GPU reservations of up to 30 days (February 2026), dedicated model hosting (January 2026), and Hyperbolic Organizations (December 2025).