Generate power
Solar arrays supply electricity. Storage and power management support the selected orbit and workload.
Technology / The proposed platform
Solar power.
AI compute.
One orbital platform.
The goal is straightforward: turn solar electricity in orbit into useful AI inference. Delivering it requires the whole system to work together.
01System architecture
Four systems.
One engineering problem.
Run AI models
Compute hardware runs trained models within a defined power, memory and temperature budget.
Reject heat
A thermal system carries heat from the electronics to radiators that release it into space.
Connect to Earth
Ground links carry models, inputs, results and operating data, with security built into the design.
02Inference first
A focused
first workload.
Inference means running an already trained AI model. We plan to start with a selected batch job whose data and timing requirements fit the platform.
Upload
Transfer the model and its inputs when a ground connection is available.
Compute
Run the job within the platform’s power, temperature and memory limits.
Return
Send results back to Earth and measure performance, power use and reliability.
The first flight would be a research experiment. Customer workloads would follow technical and commercial validation.
03What needs to be proven
The engineering
defines the economics.
Space changes the constraints. The development program must show that useful compute can be delivered reliably at a competitive total cost.
Heat
Size and test the cooling loop and radiators under realistic heat loads and space conditions.
Power
Balance solar generation, storage, eclipses and the energy needed for useful compute.
Mass and deployment
Fit the compute, power and cooling hardware within launch limits and deploy it reliably.
Radiation and reliability
Evaluate shielding, error recovery, safe shutdown and the impact of component failures.
Communications and security
Measure data transfer limits and latency. Protect model weights, customer data and operations.
Full lifetime cost
Include hardware, launch, ground systems, operations, replacement and disposal in the comparison.
04Common questions
A clear view
of the challenge.
From architecture to evidence
Start small.
Measure what matters.
Our proposed development plan moves from requirements and a ground prototype to a qualified first flight experiment.
Explore the roadmap