benjamin aaron advisors · research

dazzling data

starlink users from design downlink · assumptions toggled below

0
users
calibrated · ~60 Mbps of design per subscriber
0 Tbps design
jan 2020
v1 0
v1.5 0
v2 mini 0
v3 0

presets

mbps per user · busy-hour rates (1 to 10) or calibrated design share (~60)

usable share of design

estimated users = design Tbps × 1,000,000 × usable share ÷ Mbps per user. default is calibrated: 12.0m subscribers at 06/30/20261 on about 719 Tbps of design capacity works out to about 60 Mbps of design per subscriber. busy-hour presets (3 Mbps) only match reality at a 5 to 10% usable share, since most capacity sits over oceans and empty land at any moment. stellar times uses stifel’s method (2 Mbps on about 30% of capacity)2, which reads higher. both are estimates.

bear - users base - users bull - users end-2028 · capacity-implied base needs ~58 full v3 starship launches by end-2028 (3,500 v3 ÷ 60 per launch, baa est.)

footnotes · sources and assumptions

Design downlink per sat (BAA encoding): V1 17 Gbps (low end of the 17 to 23 Gbps in SpaceX’s first-generation FCC filing)3; V1.5 24 Gbps (implied by SpaceX’s statement that V2 Mini carries ~4× the capacity of earlier satellites3, vs V2 Mini 96); V2 Mini / Optimized 96 Gbps (SpaceX 2024 Progress Report)4; V3 1,000 Gbps / 1 Tbps downlink (SpaceX V3 specs as reported 2024 to 2026)4.

Fleet path is a BAA reconstruction from public sat counts (Jonathan McDowell’s Starlink statistics5) and press coverage (end-2020 ~15 Tbps; end-2024 ~327 Tbps in our rebuild. spacex says it added more than 300 Tbps in 2024 alone4, more than our per-satellite math gives, so treat early years as rough; mid/late 2025 to Sep 2026 climb to ~700 to 800 Tbps). Forecast uses V3 Starship cadence: Flight 14 on 09/28/2026 (+26 V3, in service a few weeks later); later loads up to 60 V3. Base: ~50 to 80 V3 end-2026, ~1,000 mid-2027 (Elon Musk’s “critical mass” estimate on SpaceX’s Q2 2026 earnings call6), ~3,500 end-2028. Bear ~⅓ and bull ~2× base V3 adds. Pre-V3 fleet flat-to-slowly rising. Vertical markers: v2 mini, optimized, flight 14 v3s in service (est.), actual | forecast.

Design capacity ≠ sold throughput. Users shown are capacity-implied estimates from toggleable busy-hour Mbps and usable-share assumptions, not SpaceX guidance. Full write-up: model note.

sources

  1. spacex · q2 2026 results (12.0m starlink subscribers at 06/30/2026) · 08/04/2026 · link ↩
  2. stifel · satcoms’ muskonomics challenge (2 mbps per user, sellable share) · 06/2024 · link ↩
  3. ars technica · spacex’s 2nd-generation starlink satellites (v1 17 to 23 gbps, v2 mini ~4×) · 02/2023 · link ↩ ↩
  4. starlink · progress 2024 (v2 mini 96 gbps, v3 1 tbps, more than 300 tbps added in 2024) · 12/2024 · link ↩ ↩ ↩
  5. jonathan mcdowell · starlink statistics · 09/28/2026 · link ↩
  6. btw.media · flight 14 and ~1,000 v3s by mid-2027 · 09/2026 · link ↩

public sources only. nothing here uses or reflects confidential information from spacex, any former employer, or consulting work. not investment advice. the author may hold spcx or tsla.