Starlink's hidden pieces: one Starship launch at a time

What Elon said

Near the end of the Q3 call on 20 October, Elon said "we do see a clear path to several hundred billion dollars a year in connectivity revenue" because "you are missing some of the pieces that we know about internally." Earlier he said "revenue for Starlink will be at least somewhat proportionate to the total bandwidth delivered." Gwynne Shotwell added that SpaceX expects "Starlink alone to represent a significant portion of global Internet traffic." These quotes are from an unofficial transcript of the replay, checked against the Motley Fool transcript.

If revenue follows bandwidth, the unit to study is the thing that adds bandwidth: one Starship launch.

The unit: one full Starship launch

Tbps (terabits per second) measures how much data a network can move. 1 Tbps is 1,000 Gbps.

Elon said Flight 14 deployed 26 Tbps of Starlink bandwidth, "equivalent to 10 Falcon 9 launches," while "about maybe half loaded." CFO Bret Johnsen said each Starship launch is ramping toward "over 20x the downlink capacity" of a Falcon 9 launch. Headline capacity is not what customers get, since satellites spend much of their time over oceans and empty land. ABI Research models usable capacity at about 21.3% of headline.

Step 3 uses a hobbyist tally of 864.7 Tbps of headline capacity today, from Jonathan McDowell's tracking, times 21.3%. The $1,360 in step 5 is $300B a year (our reading of "several hundred billion") spread over 100 times today's usable capacity, which Elon expects "probably within three years." It is 86% below today. Step 6 uses the S-1 goal to cut launch cost "by 99% or more" from the $18,500 per kilogram historical average, so about $185 per kilogram, times the 100 metric tons we assume for a full Starship V3.

The base rate: launch costs have fallen before

SpaceX has done a step like this once. Per the S-1, citing NASA, the first Falcon 9 cut launch cost to about $2,700 per kilogram in 2010, about 85% below the $18,500 historical average. The Starship goal is a second drop of about 93% from there (our estimate). The first drop made Starlink possible. The second is what would make 100x bandwidth affordable.

For scale, 100 times today's usable capacity is about 18.4 Pbps, roughly 8 times the 2,259 Tbps of international internet bandwidth TeleGeography counts. Different measures, so treat it as a sense of size.

The physical check

100x means adding about 85,600 Tbps of headline capacity. At 52 Tbps a launch that is about 1,650 launches. Step 5 gives the same count from the money side: $300B divided by $181M a launch is about 1,660. Over three years that is about 10 launches a week (our estimate). SpaceX's plan is weekly flights in the first half of 2027, then twice weekly or more by the end of 2027. So the money math works long before the rocket math does.

A CDN becomes worth building

If revenue tracks bandwidth but price per unit falls 86%, every byte that does not cross the satellite link is saved cost. That is what a CDN (content delivery network) does: it keeps copies of popular videos and updates close to users. Two live SpaceX job posts ask engineers to "Design, build, and scale a new Content Delivery Network (CDN) for Starlink." A third hires staff for "hardware at Starlink PoP colocation sites worldwide," the buildings where Starlink meets the wider internet. For scale, Akamai booked $4.208B of 2025 revenue ($1.257B from delivery) and Cloudflare $2,167.9M. Whether SpaceX would sell CDN service or only use it to cut its own costs is not disclosed.

The other pieces

Classified compute. Bret said enterprise and government revenue grew 91% year over year, "highlighted by growth in our Starshield program," the government version of Starlink. One job post leads business growth across the intelligence community "focusing on AI and classified compute that amplify overhead space systems." Another manages data center hardware with a TS/SCI clearance, and a third leads "Starshield Silicon" chip design. The US appropriated $73.3B for the National Intelligence Program and $27.8B for the Military Intelligence Program in FY2025, $101.1B in total.

GPS backup. In a May 2025 FCC letter, SpaceX called "next-generation low-Earth orbit satellite constellations that can deliver PNT as a service" (positioning, navigation and timing) "a particularly ripe, low-hanging fruit." UT Austin researchers say Starlink's timing weaknesses "appear to be a consequence of software design choices, not hardware limitations." Ars Technica reported in May 2026 that Starlink shut off the access some users had found. No product exists, so this piece is speculation.

Kits. The Q3 10-Q shows product revenue, all of it in Connectivity and mostly Starlink kits, at $771M, up from $461M in Q2 and $430M a year ago. Bret said SpaceX prioritized kit deliveries to retail "ahead of the holiday season."

The money math (our estimate)

Every input below is our assumption: capacity sold, revenue per piece, margins and multiples. Revenue is yearly, around 2030 to 2031. The bear case total is below today's $21B, so it assumes prices fall faster than volume grows.

Bull, 20% of $2,255B451

Base, 50% of $810B405

Bear, 30% of $96B29

Weighted total885$SPCX = ~$65

Weighted value: about $0.88T. Over 13,581,063,709 shares (10-Q cover, 15 Oct 2026), that is $SPCX = ~$65 before discounting for time, or about 45% of the $144.75 close (a market value of about $1.97T). Part of this is likely already in the price.

What to watch

• The longer Starlink presentation Elon suggested on the call.
• Starship cadence against weekly flights in the first half of 2027.
• Q4 subscriber adds. ABI expects over 15.6M by the end of 2026, against 13.6M at the end of Q3.
• A CDN or Starshield compute contract named in a filing.

What would make this wrong

The $18.5M is a target SpaceX has yet to reach, and it leaves out the satellites themselves, whose cost SpaceX does not disclose. The true payback from one launch could be far longer than our ratio suggests. The 864.7 Tbps starting point is a hobbyist tally, not a company figure.

Not financial advice.