Today, we’re opening a new chapter for Hyperbolic with a new domain, a new look, and a new set of products designed for the developers building the future of AI.

You can now find us at hyperbolic.ai and spin up GPUs and AI models directly at app.hyperbolic.ai.

You’ll probably notice the plants, butterflies, and other wildlife on our site — yes, it’s a bit poetic for an AI infrastructure company. But it reflects a belief we hold deeply: GPU compute is the foundation — the rich soil of the AI ecosystem. It’s what empowers AI developers, researchers, and companies to build and grow artificial intelligence — the flora and fauna of this new era.

Before starting Hyperbolic, I worked in quantitative finance at Citadel Securities, where I learned that every microsecond and basis point mattered. Efficiency was everything. When I started looking at AI infrastructure, I couldn’t believe the inefficiency. 80% of enterprise GPUs were sitting idle, while my friends in AI research struggled to find available machines. It was insane. 

We started Hyperbolic because we believe that building more data centers isn’t the only solution – contrary to @sama.

If you’ve shopped around for GPUs, you’ve seen the spread: AWS charges $9+/hr for an H100. Niche providers go as low as $2.99/hr. At Hyperbolic, we’ve pushed that down to $0.99/hr – same hardware, just different economics and incentives.

So why doesn’t everyone switch?

No one wants to stake a multi-million dollar training run on a provider they’ve never heard of. Reliability and trust matters. That's why we've spent the past year doing the groundwork – benchmarking machines, validating throughput, checking network stability and bandwidth – so you don’t have to gamble to get low-cost GPUs.

Where we started: Spot GPUs and Serverless Inference

What if instead of building more data centers, we unlocked 80% of the GPUs that are being underutilized?

Our platform transforms otherwise idle compute into revenue for data centers, and in turn reduces compute costs for AI companies by 50-75% by offering spot instances. Lower reliability, but low cost for smaller workloads.

Through community feedback, we’ve grown this product to be a tool that 200,000+ developers from some of the best research teams around the country love. Now we’re onto part two of our quest.

What's New: On-demand and Reserved

For the past 4 months, we’ve turned our team’s full focus to reliability and scale so that developers can train and fine-tune larger models on Hyperbolic. Today, we’re opening a first look at our On-Demand and Reserved products. These aren't just cheaper alternatives, they're different approaches to AI infrastructure:

On-Demand GPUs and Clusters: For growing AI companies tired of choosing between burning through budgets and progress. Focused on startups and smaller teams with agile but targeted compute needs. Spin up thousands of GPUs in seconds and pay only for what you use with instant quotes. We offer bare-metal clusters in addition to VM instances, allowing you to rent multi-node clusters tailored to your workloads.  You can now access production-ready H100s for just $1.49/h at app.hyperbolic.ai.

Reserved Clusters: For organizations that need optimized large-scale, mission-critical workloads requiring extensive compute and network performance. Lock in dedicated clusters of GPUs for predictable performance and leverage great discounts on long-term reservations. Schedule your call here.

We built this for you. Tell us what you're building, we'll power it with 99.5% uptime, 70% lower costs, fast deploys, and a world-class team behind you. Connect with us.

What’s coming: The Future of AI is Accessible

The future of AI isn't about building more data centers. It's about building smarter systems. It's about recognizing that in a world of exponential AI growth, linear infrastructure scaling is a death sentence.

Real change starts with a simple realization: We already have enough GPUs. We just need to share them better. 

Today, whether you're spinning up H100s on-demand, reserving clusters for large-scale training, or using our serverless inference to access the latest models instantly - Hyperbolic gives you maximum flexibility at a fraction of the cost.

Welcome to the future of collaborative AI. Let's build it together.

— Dr. Jasper Zhang
Co-Founder & CEO, Hyperbolic

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.

Website | X | Discord | LinkedIn | YouTube | GitHub | Documentation