Hyperbolic was founded in late 2022 by Jasper Zhang, a mathematics PhD and former International Mathematical Olympiad medalist, and Yuchen Jin, a former cloud infrastructure engineer. The company's mission is to make access to AI compute genuinely open and affordable, countering a market where GPU capacity is expensive, scarce, and concentrated among a handful of providers. Hyperbolic approaches this as both a marketplace problem and a serving problem.
On the supply side, Hyperbolic aggregates GPU capacity from data centers, miners, and other operators with underutilized hardware, creating a marketplace that brings idle compute online. On the demand side, it exposes that capacity to AI developers as rentable GPUs and as serverless inference endpoints for popular open models. By matching latent supply with developer demand, the platform can offer prices significantly below conventional cloud rates while keeping access self-serve and contract-free.
The product spans several layers: a GPU marketplace for renting raw compute for training and fine-tuning, an inference service that serves open LLMs and image models through a simple API, and AI services that developers can compose into applications. Hyperbolic positions itself as a one-stop, lower-cost alternative for teams that find hyperscaler GPU access either too expensive or too hard to obtain.
The company raised seed and pre-seed rounds before closing a $12 million round in December 2024 led by Variant and Polychain Capital, with participation from Chapter One, Lightspeed Faction, IOSG, Vertex, GSR, Wintermute Ventures, Blockchain Builders Fund, Alumni Ventures, and Ambush. That brought total funding to roughly $20 million. The investor mix reflects Hyperbolic's blend of AI infrastructure and decentralized-compute marketplace dynamics.
Hyperbolic competes with GPU marketplaces and serverless inference clouds alike. Its differentiator is the combination of aggregated, lower-cost supply with a developer-friendly inference layer, making it appealing to independent developers, researchers, and startups who want flexible, affordable access to both raw GPUs and ready-to-use model endpoints.