Hyperbolic raises $30M to aggregate the world's heterogeneous GPU compute into one unified platform

Hyperbolic is building a unified platform that aggregates heterogeneous GPU compute from across the market and makes it accessible through a single control plane. The company has raised $30M and counts Intel CEO Lip-Bu Tan among its backers, investing in a personal capacity through his semiconductor venture firm, Walden Catalyst.

The platform

The core product is a system called Forge, a unified control plane that can attach any bare-metal cluster to Hyperbolic's platform and expose it to customers globally. Once connected, compute can be allocated flexibly, shared across workloads, and crucially, sold back into the market when the original customer isn't using it.

Zhang frames the target customer not as scrappy operators looking to offload idle gaming rigs, but as serious cloud tenants, including Neo Cloud operators and enterprises locking in multi-year contracts. The practical supported hardware ranges from A100s and RTX Pro 6000s for inference up through Blackwell. The pitch on economics is direct: spot sales of B300 capacity might clear at $4–5 per GPU hour, but Hyperbolic says it can help operators push that toward $10–20 per GPU hour by optimizing utilization and demand matching.

“At Hyperbolic we aggregate and turn heterogeneous compute into one unified platform. It's a platform where people can just come access compute in a very flexible way, manage and orchestrate the compute globally, and then sell the compute back when they don't need it... Our goal is to maximize the revenue per GB hour.”

The market logic

The GPU compute market has fragmented to the point where Zhang says there are now more than 300 Neo Cloud providers. That fragmentation has created an unfamiliar problem for technology founders, who suddenly have to manage sourcing, procurement, cash flow, and risk hedging alongside their core product work. Hyperbolic's aggregation layer is designed to sit above that complexity.

The idle-capacity use cases Zhang describes are varied: a company pausing between training runs, workloads shifting across time zones as demand drops overnight, inference capacity freed up while a cluster sits between jobs. All of it can theoretically route through Forge and generate incremental yield.

The Intel angle

Lip-Bu Tan's involvement is financial rather than strategic, at least for now. Zhang notes that Tan founded Walden Catalyst, a semiconductor-focused VC, and is investing from that vehicle rather than through Intel's corporate balance sheet. That said, Zhang doesn't rule out a strategic angle, given Hyperbolic's stated goal of aggregating compute across chip architectures, including non-Nvidia silicon.

The longer ambition is straightforward: put the world's compute behind anyone bold enough to build on it. Whether that vision supports a durable business at scale depends on how much pricing power an orchestration layer can hold as the Neo Cloud market consolidates around it.