Long-term contracts give you a stable operation and a site a lender will finance. The short-term market pays more per hour than the same capacity under a multi-year term — when it buys at all. In an oversupplied quarter that premium narrows and the hour book thins out, which is exactly why the contract book underneath it keeps the site solvent. Genesis Grid lets you serve both from the same cluster at the same time, so idle time shrinks to what neither book wanted. We model both against your own rate card rather than an industry curve, and we will show you the quarter where the premium is not there.
A cluster leased to one tenant is paid for whether that tenant uses it or not — their idle hours are their problem, not yours. Your problem is the capacity nobody has contracted: the racks between two anchor deals, the tail of a fleet that only sells in units of eight, the cluster that comes online months before its customer does. Basis: published research, not our own operating records. Execution-idle is measured at 19.7% of in-execution time on a large academic AI cluster (arXiv:2604.04745, April 2026). The compute share comes from a production deployment of more than 20,000 GPUs, where SM activity rose from 16% to 33% (MuxFlow, arXiv:2303.13803). The revenue figure is a model calculation: two paying tenants per accelerator at the measured throughput, priced against published rental rates — we will walk the assumptions with you rather than ask you to take them. If your cluster is already shared and contracted end to end, this number is not yours and we will say so on the first call.
of the hours you already sell, the accelerator sits idle
real compute share per accelerator once one cluster serves many tenants
additional revenue per installed accelerator, no new accelerators or fabric
Metered by the hour or the minute against the rate card you publish. This is the entry point for tenants still sizing their workload, and in most quarters the book that carries your best average price per hour.
Fixed capacity for a fixed term, invoiced up front or in instalments. This is contracted revenue you can put in front of a lender or an investment committee, and it is what makes the on-demand book affordable to run.
Consumption billing for inference endpoints your tenants publish on your platform. The meter ships with the stack and records from the first request, so a new billing shape is a configuration decision rather than a project.
A cluster dedicated to a single tenant, or capacity resold under a partner brand. Which isolation the product carries is a property of what you sell and is set out on the multi-tenancy page.
Genesis Grid meters every tenant against your rate card and produces the usage records behind each invoice. You can invoice from the platform or export the metering data into the billing system your finance team already runs, with the tenant, project and cost-centre structure it expects.
.svg){kind=small}
Yes. Rate cards, discounts, commitments and contract terms are yours alone, and no list price of ours sits underneath them. Nothing your tenants pay passes through us.
No. Genesis Cloud is a software company and sells no capacity. Volta licensed Genesis Grid and is building its own business on it. The ownership and conflict question is answered in full on the about page.
Tenants onboard themselves, so the clock is set by your side of the work: 2 days to stand the stack up on a further cluster. From signed contract, pilot included, the first invoice follows in two weeks.
Two forms: a licence, or a revenue share. Under a licence the fee is charged against the capacity you put under management, not against what your tenants consume, so every point of utilisation you win stays on your side of the ledger. Both forms are set out on the pricing page; the figures come out of a conversation.
