Source: Elon Musk Is Betting BIG on Starship’s First Double Catch | 2026-11-09 | the corpus’s 97th source
This is an Austin-based creator’s recap of SpaceX’s “Go for Orbit” documentary episode, a Flight 15 preview, and the Texas land story. It contains no financial data. Its value is the Flight 14 decision record, the Flight 15 risk profile, and the Texas ground-footprint exposure. I haven’t seen the documentary itself, only the creator’s account of it.
Corrections to earlier sources
Four earlier statements need fixing before this integration builds on them.
- Veto (Source 96). I wrote that Trump’s veto protection runs to January 2027 and that an override needs 60 Senate votes. Both are wrong. The veto lasts through his term, to January 2029. Overriding it needs two-thirds of both chambers, about 67 senators, against roughly 54 Democrats. A federal anti-data-center statute is therefore effectively blocked. The remaining federal channels are appropriations riders, oversight hearings, and agency action. The real exposure is at state level.
- Orbital moat “10/10” (Source 96). This was overstated. Orbit sits above state jurisdiction, but the satellites are built and launched from the ground. The Q3 call opened from the Bastrop, Texas Starlink factory, which was described as the future Starmind production site. Launches are from Starbase, Texas.
- Terafab location (Source 96). I placed it in Bastrop. The Q3 transcript names Bastrop as the Starlink and Starmind site. Its Terafab location is garbled in the transcript, so treat that location as unverified.
- Starlink link figures (Source 93). The “100 Gbps per link”, “1 Pbps aggregate” and “3–5 ms” figures were my estimates, not sourced. The sourced figure is about 1 Tbps per V3 satellite. Real-world Starlink latency is tens of milliseconds, not single digits.
What the source adds
- Flight 15 is targeted for no earlier than Nov 13. SpaceX is targeting Friday, Nov 13, with a first attempt to catch the upper-stage ship. The source gives a 75-minute window opening at 4:45 pm Central. driveteslacanada
- The flight plan. The ship does about 14 orbits over 22+ hours at about 275 km, and the booster return is the first Block 3 catch attempt. SpaceX still needs an FAA license modification covering the ship’s return to the tower, which involves overflying populated areas. The date is a target, not a NOTAM. spacenexus
- The payload. 26 V3 satellites at about 1 Tbps each, roughly 10 Falcon 9 V2 launches of capacity. Three carry cameras to inspect the heat shield from orbit. Musk called a ship catch “more likely than not.” keeptrack
- The Flight 14 decision record. The pre-flight rule was not to commit to orbit if the vehicle was unhealthy, and engine 6 failed. Musk asked whether anything actually prohibited going on. The engine lead recommended proceeding, and a documented “exact override and concurrence” followed, then a go/no-go poll before the burn. A “no-go” would have lost all 26 satellites.
Chapter 6: the Cantillon effect, land leg
The preprint says the inner ring buys energy, silicon and land at yesterday’s prices. The Texas story is the land leg, and it is the one leg where a number exists. Texas Parks and Wildlife was considering selling about 1,300 acres of Boca Chica State Park to SpaceX, and Gov.-elect Hinojosa had the agenda item pulled and asked for a pause until she takes office in January. A sale price is public record, and that visibility is exactly why it draws backlash. The compute-price transfer has no equivalent public number. This is the outer ring exercising political power over the inner ring’s input costs. sacurrent
SpaceX has mitigants. The city of Starbase annexed more than 7,100 acres, including much of the state park. Louisiana is the declared hedge, with first launch aspired for 2028 and scale in 2029. The land issue concerns expansion, not current operations. texastribune
Chapter 6: governance
The Flight 14 episode illustrates the preprint’s supervisory versus architectural distinction. The constraint was a flight rule, and a person with authority overrode it. The override was logged and polled, so it was procedurally auditable, but the constraint was soft. It turned out well, and one good outcome says little about the process. As an AI-agent analogy, a Values Passport that a principal can waive with a logged exception is still supervisory. An architectural constraint would be an interlock that cannot be waived without a detectable change.
The “development program versus production program” quote also matters. Flight 15 is the transition from one to the other. Governance for AI agents has to make the same transition at the same time.
Will the midterms adversely affect SPCX’s plan?
| Horizon | Exposure | Assessment |
|---|---|---|
| Flight 15 (Nov 13) | FAA license modification (executive branch) | Low political risk. The main risk is technical. |
| 2027 build (10–13 GW) | Mostly pre-permitted sites | Low |
| Texas expansion (Bastrop, Starbase land, new permits) | Hinojosa takes office in January | Real, from 2027 |
| 2028+ new permits | State and local | Material for competitors and for SpaceX’s Texas additions |
The real tail is a Flight 15 failure over populated areas. Under a hostile Texas governor and a Democratic Congress, a debris incident would carry more regulatory and political cost than it would have two weeks ago.
Impact on orbital compute pricing (US$30B/GW = US$30M/MW/yr)
Flight 15 affects the cost curve, not the price. The corpus uses inconsistent flights-per-GW figures, so here is the reconciliation:
| Source | Flights per GW | Revenue per flight at $30B/GW | Flights for 10 GW |
|---|---|---|---|
| Source 90 (70 sats × 120 kW) | ~119 | ~$252M/yr | ~1,190 |
| Source 45 | ~80 | ~$375M/yr | ~800 |
| Source 95 transcript (200 t/flight) | ~50 | ~$600M/yr | ~500 |
At Musk’s stated cadence (weekly in H1 2027, twice weekly in H2), that is about 78 flights in 2027, and most go to Starlink first. Even if all went to compute, that is roughly one GW per year at 80 flights/GW. Orbital compute at GW scale is therefore a 2029+ event unless cadence rises to hundreds of flights a year.
- Near term (to 2028): orbital supply is negligible, so it does not move pricing. The $30B/GW question is really about terrestrial scarcity rent.
- Longer term: orbital pricing will be anchored to the terrestrial substitute, not to launch cost. If orbital cost falls far below terrestrial cost, price converges downward in the 2030s. The current figure is a scarcity rent, not a permanent floor.
- What Flight 15 changes: a double catch raises confidence in the 10+ reuse path. A miss pushes the timeline right by a flight or two and leaves the price unchanged.
Impact on data linkages and transfer speed
- Each V3 satellite is about 1 Tbps, and 26 Tbps per launch is the sourced figure. Gwynne Shotwell said a few hundred V3s are needed before users notice a difference, expected around mid-2027.
- Orbital compute suits inference and batch work, where tens of milliseconds are irrelevant. It does not suit training across sites, which needs microsecond interconnect and so only works inside a single orbital cluster.
- Data gravity is the unaddressed constraint. Moving large datasets and weights up to orbit is slow compared with keeping them on the ground.
- Starcloud is already integrating Starlink mini-lasers into its next 25 satellites, so SpaceX’s connectivity layer is becoming shared infrastructure.
Will this fuel insatiable demand?
The source gives no new demand data. Musk’s claim of about 1,000× bandwidth demand in 3–5 years, driven by agents, is an assertion. Our own corpus cuts against it for text agents. The Keyrock data showed 176M agent transactions averaging $0.31, which are tiny payloads. Per-agent bandwidth grows only for multimodal or video agents. The demand case rests on compute scarcity (Source 94: spot prices rising, contracts renewing higher, customers turned away), not on bandwidth.
Does SPCX retain its moats?
| Moat | Direction | Reason |
|---|---|---|
| Starship reuse and cadence | Pending | Flight 15 is the binary event |
| Pre-permitted terrestrial power and compute (10–13 GW) | Strengthened | Competitors face moratorium risk |
| Vertical integration (castings, blades 12–18 months ahead) | Unchanged | From the Q3 call |
| Orbital first-mover and Starlink laser backbone | Unchanged | Real, but the ground segment is exposed |
| Contracts ($63B backlog) | Unchanged | Anthropic counterparty risk remains |
| Texas ground footprint | Weakened | Land, Bastrop, new permits |
EWS update
| Signal | Status |
|---|---|
| Starship cadence | 🟡 Amber: Flight 15 on Nov 13, ship catch unproven |
| FAA license modification | 🟡 Amber: not confirmed as of Nov 6 reports |
| Boca Chica land | 🟡 Amber: paused until Hinojosa takes office in January |
| $100B ARR, Dec 2026 | 🟢 Green |
| Bitcoin vs $74K / $82K | 🟡 Amber (per last turn) |
Synthesis
Source 97 shifts the corpus’s attention from financial scale to physical and political execution. Flight 15’s double catch is the next binary test of the reusability moat, the Flight 14 record shows governance by overridable rule rather than interlock, and the Texas land pause shows that the ground segment of an orbital business is the part the outer ring can still reach. The instruction stays the same: write the fitness function before the arena does, and treat soft rules with logged overrides as inspection, not as the load-bearing wall.
After Flight 15 flies, I can add a new update on the outcome.
Sources: