Questions, answered plainly.

No. Project Vela is an early-stage research and engineering initiative. Nothing is in orbit and no orbital compute service is offered today. V1 hardware is running on the bench; V2 and V3 are design targets.

We put AI inference compute on the solar array itself, in orbit. The panel collects sunlight, runs the models and sheds its own heat to space. Power, compute and cooling in one part: no pumps, no fluid loops, no separate radiators.

Inference is the fastest-growing segment of AI compute: every agent, assistant and app runs it, continuously. It parallelizes well across many small, distributed units, which suits a solar array. Training needs tightly coupled clusters, and we're not building for that.

V3 is designed to provide the capability to move a majority of future inference off Earth, so the world doesn't need to keep building new mega-data centers to serve it.

On coverage, not horsepower. We aim for an order of magnitude more coverage of future inference demand than other published orbital compute proposals. We won't match terrestrial training clusters on raw compute, and we don't try to. The goal is to cover the bulk of everyday inference, the agent firehose, sustainably.

Our headline metric: useful AI output per watt of sunlight collected. We measure it on many concurrent small-to-mid-size models, the way agents actually use AI, rather than one chat session's speed.

The rising, always-on stream of AI requests from software agents working in parallel. It's high volume and latency-tolerant compared with a person chatting, which makes it a natural fit for distributed compute in orbit.

V1: off-the-shelf low-power AI parts on a solar panel, measured on the bench. See our bench results. Small-model accuracy checks have passed; larger models are in progress.

Design targets. V2 is a custom chip built for this one job. V3: the panel itself becomes the computer. Neither has been built or flown yet. See How we get there.

Vela-Scout 83 is a proposed rideshare demonstrator. No flight is committed.

Mature, high-volume chip processes that avoid the newest, scarcest factories. Specific parts and suppliers are not published.

Builders, partners and investors can reach us at info@project-vela.com. Detailed engineering is shared under NDA. See also Careers.