Use the cluster you already own alongside rented cards
One place to see both, and one meter that understands the difference between hardware you are paying for by the second and hardware you already bought.
Everything below describes what this is being built to do. It needs Compute, which you can look around today but cannot yet use for the whole of this job. The problems are real now; the answers are the part still being built.
Hardware you already bought and hardware you are renting are the same workload and two different tools.
Why it is hard
Two control planes, one job
Owned capacity and rented capacity are managed separately, so the workload that spans both is stitched together by hand.
The comparison is impossible to make
Working out whether the owned cluster is cheaper than renting requires putting two incompatible cost models side by side.
Bursting is a manual decision
Reaching for rented capacity when the cluster is full is something somebody notices and does, usually later than they would have liked.
What answers it
Both appear in one place
The intent is that your own cluster sits beside rented capacity rather than in a separate tool.
One meter across both
The useful part is not the connection but the reporting: hardware you own and hardware you rent by the second in the same ledger.
Outbound, from your side
The connection is expected to be made out to us, which is the shape that does not ask you to open anything to the internet.
How billing works
per second, to the microeuro · no minimum · no egress fee · idle costs nothing
Metering starts when the container does and stops when it stops. The rate is frozen the moment you match, so a provider re-pricing cannot change a running rental, and an aborted run costs what it used.
What you use from Kracht
Related jobs
- ComputeTrain a model without buying hardwareTake an H100 for the length of a training run and give it back. No reservation, no minimum, and no card sitting idle between experiments.Read the job
- ComputeFine-tune on a scheduleA job that runs nightly needs a GPU for the hour it runs, not for the day around it. Per-second metering makes the difference the whole business case.Read the job
- ComputeServe inference when traffic spikesAdd capacity for the hours you need it and stop paying when the spike passes. Nothing continues to bill after the container stops.Read the job