The race map — every possible world at once

Nobody knows what a satellite-year of orbital computing will turn out to be worth, or how much of it the world will buy. So instead of guessing, the study swept the whole space of possibilities. Every point on this map is a possible world.

In plain terms: across the map, a satellite-year is worth more; up the map, the world wants more of it. In the grey worlds (~69% of the swept box), the million-satellite paper queue never becomes hardware — filing was free, building never pays. In the blue worlds, fleets get built and stay under the physical threshold. In the amber worlds, the same incentives that build the fleet carry it straight past the cliff — and nothing in the machinery is calibrated to stop at the line. The model cannot say which world is ours; it says what kind of map we are walking on. Run your finger across it.

paper never becomes hardware fleets built, under the threshold built and carried past the cliff | dashed line: where building starts to pay

The thin middle — how narrow the safe-and-profitable window is (tap to open)

Where the cliff is reachable at all, the gap between "worth building" and "past the threshold" is usually thin: in half the reachable worlds the cliff sits within 1.35× of the build-it line, and in 74% it sits within 2×. A market that overshoots by "a bit" is, on most of this map, a market that overshoots the physical line.

MODELED · ASSUMED BOX Verbatim from the data file:

Grid: 200×200 at knob medians, additive D8 emission of the certified race model (seed 26716; byte-identity attested in-file; referee fold-in complete). Boundaries drawn from the grid's own region rows (regions are contiguous in value per demand row). Source: race_map_grid_v1.json; if a number here and a number in that file disagree, the file wins.

AS FILED · CLAIMED The official voice: both new entrants have formally asked the FCC to waive the deployment milestones and surety bonds — asking off the very clock that the middle region's "milestones honored" premise assumes. The queue's own paperwork contests IF #2. Meanwhile the filings' business cases, where stated at all, name only the latency-tolerant workload class. Sources: the waiver exhibits (primary), paper §6.3 and §7.4.

In the paper: §6.3 (the race model and its IFs) · §6.4 (the thin middle) · Fig. racemap