A small lab that rents its own compute
Sundancæ Research Inc. operates a 312-GPU estate across multiple sites. We rent it by the hour, we build custom software and advertising systems on top of it, and we reinvest what both earn back into the fleet.
Twenty-one people. No resold capacity, no channel partners, no blog.
On-demand instances, reserved monthly nodes, bare metal and per-frame render jobs. A one-GPU account pays the same rate as a sixty-four-GPU one, billed per second, with no charge for data in or out.
We own the hardware and hold the power contracts, so there is no reseller margin in these numbers. What we cannot do is absorb a thousand-GPU training run; if that is your requirement, a hyperscaler is the right answer and we will say so.
Custom systems, built from nothing
No templates, no page builders, no reselling someone else's platform. We take on software a business cannot buy off the shelf: internal tooling, data pipelines, pricing engines, inference services. Where the workload needs GPUs we run it on our own fleet, which is usually the reason the economics work at all.
Sundancæ previously operated as a general B2B agency serving companies across every sector. That book of business, the full client list, was sold to a firm in Germany. We now take on custom engineering and advertising work only, and only where we can build the system ourselves.
Raising to build compute.
Deploying into what's next.
We are continuously raising
Capital goes into GPU inventory, colocation and power. We raise on a rolling basis rather than in fixed vintages, and every tranche is underwritten against specific hardware with a known payback period from rental revenue already contracted.
We invest across industries
Alongside infrastructure we hold minority positions in companies well outside it. The common thread is a business past its first million with one function it cannot build alone. We supply capital and, where it helps, the engineering.
Toward hybrid quantum-classical compute
We do not build quantum hardware and we are not going to. Our group works one layer up: how a classical GPU fleet schedules around, verifies and absorbs work from quantum processors reached through partner providers.
Four people, part-funded by rental margin. Results are folded into the scheduler when they hold up. Nothing here is sold as a product.
Enquire about the programmeTell us what you need to run
Capacity quotes, service scoping and investor materials all start in the same place. Pick a department below, or just send the form.