GPUs you can count on: Hyperbolic’s On-Demand GPUs

Dear Hyperbolic members,

We've evolved our approach based on your feedback. While we initially focused on Spot offerings, we're now prioritizing On-Demand GPUs with even greater reliability. This shift wasn't sudden for us; it was a natural progression. After carefully analyzing a year's worth of Spot usage data, we consistently observed that while cost-effective GPUs are attractive, we also want to provide the reliability, availability, and performance needed for workloads.

With the acceleration of AI, most data centers are experiencing capacity constraints, leaving fewer resources available for spot instances. This shift has led us to focus on our on-demand offering, which better serves our customers' needs for consistent performance and reliability.

Over the past few months, we re-architected Hyperbolic for one goal: GPUs you can count on. We are prioritizing working with data centers with the most reliable GPUs available. On-Demand is the product we’ve been building towards: fast to start, predictable to scale, and stable enough for production. It’s designed for flexibility and reliability, with clean billing and clear SLAs. H100s are available today; B200 and other next-gen parts will join the lineup as supply stabilizes and they meet our reliability bar.

Now for the practical details: We're gracefully sunsetting our Spot offering and providing a seamless transition path for all active Spot projects to On-Demand with one-click upgrades. This ensures continuous operation with enhanced reliability for your critical workloads. Looking forward, we're actively expanding our hardware portfolio beyond H100s to include cutting-edge H200s, B200s, and next-generation accelerators as they become available.

Thank you for building with us, and for pushing us to build a new era of on-demand offerings that will reliably support your needs.

Onward,

The Hyperbolic Team

About Hyperbolic

Hyperbolic is the on-demand AI cloud made for developers. We provide fast, affordable access to compute, inference, and AI services. Over 195,000 developers use Hyperbolic to train, fine-tune, and deploy models at scale.

Our platform has quickly become a favorite among AI researchers, including those like Andrej Karpathy. We collaborate with teams at Hugging Face, Vercel, Quora, Chatbot Arena, LMSYS, OpenRouter, Black Forest Labs, Stanford, Berkeley, and beyond.

Founded by AI researchers from UC Berkeley and the University of Washington, Hyperbolic is built for the next wave of AI innovation—open, accessible, and developer-first.