A single solar-power satellite is intermittent: in low orbit it spends up to 40% of every pass in Earth's shadow, so on its own it can only deliver continuous power if it carries heavy batteries. A constellation changes the picture. Spread enough satellites around an orbital shell and, at any instant, a predictable fraction are always in sunlight. This tool models that constellation as a single orbital power grid.
The key idea is eclipse-smoothing: sunlit satellites route power to eclipsed ones over inter-satellite links, so the fleet delivers steady baseload from intermittent nodes — and each satellite needs only a small buffer instead of enough battery to ride its entire eclipse. The model makes that trade concrete, from grid efficiency to the battery mass the network saves.
The model computes fleet peak power (every satellite in sun) and fleet average generation (peak scaled by the orbit's sunlit fraction). To turn that average into continuous, deliverable baseload, the power consumed by eclipsed satellites must travel over inter-satellite links, which lose a fraction at each hop. Grid efficiency is the ratio of deliverable baseload to average generation — it falls only as link efficiency drops, and only on the eclipsed portion of demand.
It then contrasts two battery strategies. Islanded, each satellite must store enough energy to ride its own eclipse, so fleet battery scales with eclipse duration. Gridded, satellites need only a short hand-off buffer because the network always has sunlit members supplying power. The difference is the battery mass the grid saves — often the single biggest mass and cost lever. Total mass sums thin-film arrays, structure, optical link terminals, and the small grid battery, which sets Starship flights and, with hardware, the capex. Cost of energy and cost per GPU-hour follow at 10%/yr capital recovery, and per-satellite fleet share shows how gracefully the constellation degrades when a satellite fails.
Fly 48 satellites of 2 MWp each in low Earth orbit with 85% inter-satellite links: the fleet peaks near 96 MW but averages less because each satellite is often eclipsed. The grid still delivers a strong continuous baseload at high grid efficiency, while cutting hundreds of tonnes of battery that an islanded fleet would need to survive its own eclipses — vividly showing why an orbital power grid, not a bag of independent satellites, is the architecture that makes space solar deliver baseload.
In a distributed constellation, some satellites are always in sunlight even when others are eclipsed, so the fleet's total generation is far steadier than any single satellite's.
So sunlit satellites can supply eclipsed ones, letting the constellation deliver continuous baseload without each satellite carrying enough battery to ride its full eclipse.
Inter-satellite link losses — but only on the eclipsed fraction of demand, so a well-lit shell with efficient links stays close to its average generation.
Enough to matter: islanded satellites size batteries to their full eclipse, while a grid needs only a short hand-off buffer, cutting fleet battery mass sharply.
No — it's a transparent first-order model with adjustable public assumptions, for planning and education, not detailed constellation design.
Yes — free, runs entirely in your browser, and available in 25 languages.