SSP, Strategic Supply Partners Live inventory
The desk
GPU procurementNodes and clusters, landed to your rack ● Live inventoryVerified lines, indicative pricing
By platform
B300 and GB300Current-generation Blackwell H200 and H100New stock and fleet rotation All platformsThe full catalogue
Rest of the stack
Memory, servers and storageThe kit around the accelerators Networking and opticsTransceivers, switching, cabling
Capacity without capex
GPU-as-a-serviceReserved, committed spend, on demand Bare-metal leasingDedicated clusters on a term
Decide first
Cloud vs bare metalWhere each one actually wins Reserved vs on demandThe break-even, calculated Cluster cost calculatorThe whole bill, not the GPU line
Space and power
Colocation and powerSingle racks to 100 MW Above 40 kW per rackWhere the shortlist collapses
Where
United States Europe Asia Pacific Middle East and Africa All regions →
Supply side
Have capacity to fill?Tell us what you hold
Analysis
InsightsWhat things cost, with the numbers Sizing guidesMemory, nodes, power, cost ComparisonsBuy, lease or rent
Start here
How an engagement worksThe first week, held to dates Find your fitThree questions, the right desk Glossary
Talk to the desk
Blog · From the desk

Reserved or on-demand: the break-even nobody calculates

Reserved capacity is cheaper per hour and on-demand is cheaper when you are idle. The line between them is a number you can work out in ten minutes, and most teams sign without doing it.

Strategic Supply Partners7 September 20264 min read
Four hundred GPUs translated into fifty HGX nodes and what each node carries

The whole argument is utilisation. A reserved commitment bills whether or not you run; on-demand bills only when you do. So the question is how many hours you genuinely consume, not how many you hope to.

The arithmetic

Take a 3-year reserved term on B300 at $4.00 per GPU-hour. That is 26,280 hours, $105,120 per GPU, $840,960 for an 8-GPU node, committed regardless of usage.

Now suppose on-demand for the same silicon sits at a 40% premium, so $5.60 per GPU-hour. The break-even is the utilisation at which the two cost the same: 26,280 hours at $4.00 equals your actual hours at $5.60, which puts break-even at about 71% utilisation.

Above 71% the reservation is cheaper. Below it you are paying for hours you did not use, and the further below, the worse it gets. At 40% utilisation the on-demand bill is roughly $588,672 against the reservation's $840,960: a $252,288 penalty for committing.

Why the premium matters more than the discount

Change the on-demand premium and the break-even moves sharply. At a 25% premium the break-even is 80% utilisation, which very few workloads sustain. At an 80% premium it drops to about 56%, which many do. So the first number to establish is not the reserved rate everyone quotes, it is the spread between reserved and on-demand for the same hardware in the same place.

The floor is the trap

Committed-spend agreements frequently carry a utilisation floor, billed at 80% to 100% of allocated capacity whatever you actually use. A floor above your realistic average converts a reservation into a subscription. Model your honest average across a full quarter including weekends and maintenance, not your best week.

What most teams should do

Split it. Reserve the baseline you are confident of, and burst on-demand above it. The reservation covers the hours you will certainly use at the better rate, and the premium only applies to genuinely variable demand. It is more work to negotiate and it is almost always the cheaper answer.

We run this calculation against your real usage before you commit, including the floor, the burst rights and the exit. The rental side is the GPUaaS desk, and the owned alternative is priced on the live inventory.

Straight answers

Asked first, answered straight.

When is reserved GPU capacity cheaper than on-demand?

It depends on the spread. If on-demand carries a 40% premium over a reserved rate, break-even is about 71% utilisation: above that the reservation wins, below it you pay for hours you did not use. At a 25% premium break-even rises to about 80%.

What is a utilisation floor in a GPU contract?

A clause billing you for 80% to 100% of allocated capacity regardless of actual usage. If the floor sits above your realistic average it turns a reservation into a subscription for hours you will never burn.

Should I reserve all my GPU capacity?

Usually not. Reserve the baseline you are confident of and burst on-demand above it, so the better rate covers certain hours and the premium only applies to variable demand.

Working through this on a real requirement? Twenty minutes with the desk, no pitch, no quote at the end of it. We run your numbers, not ours.

Talk to the desk
Keep reading

The GPUaaS desk →

What GPUaaS actually costs →

Cloud vs bare metal →

Talk to the desk

Working through this on a real requirement?

Twenty minutes with the desk, no pitch and no quote at the end of it. Tell us roughly what you need and we will come back within one business day.

Acknowledged within one hour, first sourcing pass within one business day.
✓

Sent. We are on it.

Your enquiry has landed with the desk. Acknowledged within one hour.