How / Updated Nov 2026
Our target is about 5 billion tokens per second from one 100-tonne Starship launch, on a 7B-class language model. That's roughly the work of 100,000 H100 GPUs. Here is why we think it's achievable, and what still has to be proven.
Read the paper: The Panel Is the Computer → (Paper v1 · November 2026)
No new physics. No new factories. Every part of the plan uses physics that is well understood and manufacturing that runs in high volume today. Building it is hard engineering work, but nothing in it is waiting on an invention.
Above the atmosphere, sunlight delivers about 1,361 watts per square metre. In the right orbit, it shines nearly all the time. On the ground, AI is limited by how much grid power a site can get. In orbit, the limit is how much panel you launch, and silicon solar cells are among the most mass-produced things on Earth: the world makes hundreds of gigawatts of them a year. One launch needs a tiny fraction of that.
Most of the energy in an AI answer goes into moving numbers between memory and the processor, not into the arithmetic. The shorter that trip, the less energy each token costs. We design the whole system around keeping data close to where it's used. The physical floor on energy per operation sits about 100,000 times below where chips work today, so the gains we're after come from good engineering, not from beating a physical limit.
A panel in space sheds heat from both faces. When the compute is spread across the same area that collects the sunlight, the panel settles at a moderate temperature, about 60 °C, well inside what ordinary chips are rated for. There are no pumps, no coolant loops and no separate radiators.
Our chips use established manufacturing processes, the kind that have been in volume production for over a decade. They don't need the newest, scarcest factories (EUV lithography), photonics or new materials. The chips for one launch are a fraction of one month's output from a single large factory. The packaging steps our later generations need are already in high-volume production in other industries.
| Generation | What it is | Per watt vs. an H100 | Status |
|---|---|---|---|
| V1 | Off-the-shelf low-power AI chips on a solar panel | On par | Building now. Measured on the bench |
| V2 | A custom chip designed for this one job | ~10–18× | In development with chip suppliers |
| V3 | The panel itself becomes the computer | ~30–65× | In design |
Our first generation, built from today's off-the-shelf parts, already works out to tens of millions of tokens per second per launch on paper. If every V2 and V3 gain fell through, a launch of Vela panels would still be a large AI data center that needs no grid, no water and no land.
We'd rather say this plainly:
Each item has a test or a study behind it.
V1 figures and our bench results are measured. V2 and V3 figures are design targets. Not yet built or flown.
Want the engineering? We share architecture, test data and the full physics and manufacturing case with researchers, partners and investors under NDA. Email us →
Project Vela. Infinite Sunshine. Effortless Intelligence.