Work that meets the requirement.
Measure a defined workload at a specified quality, latency and availability. For inference, token throughput only becomes meaningful with those conditions attached.
Project Vela / The defining objective
Most Useful
Compute
per Dollar-Kg.
Turn solar photons into useful work. Optimize the cost to deliver it and the mass required to put it in orbit—together.
Explore the Vela familyAspirational architecture · Ground validation ahead
COMPUTE / ARCHITECTURERESEARCH NOTES — 01
From photons
to useful inference.CONCEPTUAL
ILLUSTRATION
to useful inference.CONCEPTUAL
ILLUSTRATION
Primary focus and exploratory work
Orbital compute systemsNovel inference architecturesWatch the overview
Project Vela in a few minutes.
Φ / A joint cost-and-mass objective
Useful work is
the measure of success.
Evaluate useful compute per dollar and useful compute per launched kilogram side by side. The preferred design must balance both.
Count the cost to deliver it.
Compare useful output against cost over the same operating period. The concept comparison below uses deployed capital cost; the full evaluation must include operations and service delivery.
Count everything that flies.
Include the compute surface, hosts, storage, networking, deployment structure, thermal systems, spacecraft bus and propellant.
“Dollar-Kg” expresses a joint optimization, not a literal multiplication of cost by mass. A final design choice requires workload requirements and explicit cost and mass constraints.
01 / Orbital compute research
Project Vela.
From photons to useful work.
A family of solar-compute spacecraft concepts, developed around a common modular surface.
01
Physical systems
Project Vela
Concept & engineering studies
The solar-compute tile.
A common starting point.
The aspirational building block is a palm-sized tile of roughly 12 × 12 cm, pairing solar cells with local, low-power inference silicon on the same surface.
~3 WSunlit solar power per tile / design target
Tens of gramsTile mass / design target
Low-powerInference silicon / candidates under evaluation
Candidate inference chips are under evaluation, from commercial accelerators to purpose-built silicon. The shared goal is useful output from the complete system, not peak chip specifications.
From tile targets to system measurements
Most of each tile’s power budget is intended for compute, with the remainder for power conversion and data links. Host, memory and network support power still need to be closed at system level. Raw arithmetic capacity does not establish model compatibility or sustained throughput. Tile mass and power figures are design targets awaiting hardware validation.
The Vela family / Aspirational sizing
Three scales.
One optimization problem.
Vela-Scout is the proposed single test satellite. Vela-Optima and Vela-Max 2333 are later, larger concepts.
Tier 1 / Proposed first flight demonstrator
Vela-Scout
210 kg
The potential single test satellite: a proposed first flight demonstrator, not yet committed. A larger share of its mass goes to the central core and bus.
- Solar-compute area
- 40.3 m²
- Raw sunlit solar power
- 8.4 kW
Tier 2 / Later concept · baseline balance study
Vela-Optima
1,180 kg
The candidate balance point: distribute core and bus costs across a larger compute surface.
- Solar-compute area
- 314.0 m²
- Raw sunlit solar power
- 65.6 kW
Tier 3 / Later concept · max-bay benchmark
Vela-Max 2333
2,333 kg
A larger surface to study the trade between total output, structural demands and shared resources.
- Solar-compute area
- 625.0 m²
- Raw sunlit solar power
- 130.6 kW
Concept targets, not measured performance. These values reproduce the aspirational sizing brief. Complete-system mass and cost remain under reconciliation with engineering rollups; solar power is beginning-of-life sunlit power, not a continuous service budget.
| Consideration | Vela-Scout | Vela-Optima | Vela-Max 2333 |
|---|---|---|---|
| Role in the study | Proposed first flight demonstrator | Baseline balance study | Upper-scale benchmark |
| Share of mass in core & bus relative, concept-level | Highest | Moderate | Lowest |
| Total compute surface | Smallest | Mid-scale | Largest |
| Expected output per dollar to be validated | Lower | Candidate best balance | Similar to Vela-Optima |
| Main open question | Is overhead too large? | Where is the balance point? | Do structural demands erode gains? |
Early sizing suggests Vela-Optima as a reasonable balance point between cost and launched mass. It does not yet establish an optimum: workload, quality, duty cycle, overhead and cost assumptions must be validated together.
How to read these concepts
Masses, areas and solar power are aspirational sizing targets, not measured results. Throughput and cost per spacecraft depend on the inference silicon ultimately selected and will be published only once they can be supported by testing. Launch pricing, volume manufacturing, environmental qualification and full operating costs remain inputs for further study.
02Novel
Exploratory
Novel
Architectures
Early exploration · future work
New ways to run
inference efficiently.
Alongside Vela, we are exploring new system and software architectures that could make inference more efficient wherever it runs. This work is early and exploratory; it is not a product and makes no performance claims.
Get involved
Looking for builders.
Project Vela is assembling a small team of builders in power electronics, thermal, semiconductors, and spacecraft systems.
If that’s you, or you know someone, get in touch: jason@project-vela.com
Our approach
Start small.
Measure the whole system.
Progress means replacing assumptions with evidence, one experiment at a time.
- 01 / DEFINE
Make the question testable.
Choose a concrete workload and specify what success—or failure—would look like.
- 02 / MEASURE
Build the smallest useful test.
Include memory, power, data movement, and recovery costs. Compare against a credible baseline.
- 03 / REFINE
Let evidence set the direction.
Keep what demonstrates value. Revise or stop approaches that do not survive testing.
Current stage Preliminary engineering and research design. Hardware validation and architecture studies are the next milestones.
A few useful distinctions
Questions,
answered plainly.
Is this an operating satellite service?
No. Project Vela is an early-stage research and engineering initiative. It has not demonstrated an operating orbital compute service. Vela-Scout is proposed as a single test satellite, but no flight is committed.
Which chips does Vela use?
Candidate low-power inference chips are under evaluation. Selection will be driven by measured useful output per dollar and per kilogram for the complete system.