Transparent compute pricing

Pay for the compute.
Not the infrastructure.

SolarCompute is designed around usage-based compute pricing. Workloads are matched with compatible capacity and measured by the resources allocated while the work runs.

Run a workload → Create account →
A simpler compute model

Use the resources your workload needs.

SolarCompute coordinates the machines, network allocation and approved runtime behind each supported workload. You submit the work rather than provisioning and managing the underlying machine.

When commercial billing begins, the applicable compute rate and estimated cost should be visible before paid work is submitted.

How compute is measured

Two core resource classes.

Different workloads require different combinations of accelerated and general-purpose compute.

01 · GPU

GPU compute

Used for supported AI inference, training, rendering and other GPU-accelerated workloads.

GPU-hour 1 GPU allocated for 1 hour
Two GPUs for 30 minutes also equal one GPU-hour.
02 · CPU

CPU compute

Used for supported processing, simulation and general computational workloads.

vCPU-hour 1 vCPU allocated for 1 hour
Four vCPUs for 15 minutes equal one vCPU-hour.
From resources to cost

A price connected to the work actually being performed.

A workload can require a particular resource type, quantity, runtime and compatible machine configuration. These inputs determine the applicable compute cost.

Specific GPU models or workload types may use an applicable rate different from the default network rate.

What determines the price

Compute is not one-size-fits-all.

01

Workload

Each supported workload has its own execution requirements, validation rules and compatible runtime.

02

Resources

GPU, CPU, memory and other resource requirements determine the capacity needed to execute the work.

03

Runtime

Billable resource use is measured over the time that capacity is allocated to the workload.

04

Machine capability

Some workloads require specific compatible hardware and may use a model-specific or workload-specific compute rate.

Public testing period

We are validating the network before commercial billing begins.

01

Workload submission and validation

02

Queueing and compatible-machine allocation

03

Runtime and resource measurement

04

Execution monitoring and result delivery

05

Billing policy and pricing controls

Pricing transparency

Know the applicable rate before you run paid work.

SolarCompute is being designed so users can review the applicable compute rate, workload requirements and estimated cost before committing to a paid workload.

01

Rate

See the applicable compute rate for the selected workload.

02

Resources

Understand the GPU or CPU capacity required to perform the work.

03

Estimate

Review the estimated compute cost before paid execution begins.

SolarCompute network

Bring the workload. We find the capacity.

Explore supported workloads and see how distributed compute is allocated across the SolarCompute network.

Explore workloads →