SpaceX has a stated launch ladder: 1,000 Starship flights in 2028, 3,000 in 2029, 8,000 in 2030, alongside orbital AI data centre goals of 1 GW in 2027, 10 GW in 2028 and 100 GW in 2029.
Where it comes from matters more than usual. These figures were given in a technical talk on June 8, 2026 about manufacturing, launching and operating AI satellites at scale — the day before the IPO wrapped — and they reach the market through an aggregator's summary rather than through a prospectus, a shareholder letter or an earnings call. There is no filing that carries them. A number stated at a technical event on the eve of a listing is not guidance, and this piece treats it as an ambition to be measured rather than a plan to be modelled.
One number in the same story is different in kind. On the Q2 2026 earnings call on August 4 — the company's first as a public company — Musk said SpaceX would "probably a year from now" be doing at least one flight a day, possibly more. That is a company statement on a results call, with a date attached, and it can be graded by August 2027. Everything else on the ladder can only be graded after the thesis has already paid or failed.
The argument is eight figures. Here is what each one is, and how firm it is:
- 1,000 / 3,000 / 8,000 flights for 2028 / 2029 / 2030 — Musk's stated targets from the June 8 technical talk. Claimed, not filed.
- 21.9 flights a day is what 8,000 a year means. The headline doing the rounds says "30 daily launches by 2030"; the ladder itself says 21.9, and the ladder is the number worth having.
- 170 flights in 2025, ~158 annualising in 2026 — all vehicles, from our cadence piece three days ago. 8,000 is 51× the 2026 run rate.
- 42 of 2025's flights carried an external payload — 165 Falcon 9 missions less the 123 that were Starlink. That is the number our model's Space segment is built on, and it is the honest denominator for a revenue launch business.
- 374 flights in 2030 is what our own model projects, against the 8,000 target. 4.7% of it.
- 91 flights a quarter — one a day — is our model's stated ceiling. It reaches 73 of it by 2031 Q2 and never touches it. Musk says one a day arrives by next August.
- $9.31 a share is what the entire orbital-compute vertical is worth in the model today, of a $187.12 base-case fair value against a $146.23 price as of August 19.
- $500.93 is fair value if the company's ladder is fed into the model in place of ours — a $313.81 swing, and 2.7× the base case.
Our ladder against theirs
The model already assumes Starship cadence ramping hard: the Space AI vertical opens in 2028 Q1 at 12 Starship flights a quarter and glides toward a ceiling of 91 — one a day — closing a tenth of the gap each quarter. It is not a conservative line. Here is where it lands against the company's:
| Flights a year | R40 model | SpaceX target | Model as a share |
|---|---|---|---|
| 2027 | 72 | — | — |
| 2028 | 203 | 1,000 | 20.3% |
| 2029 | 302 | 3,000 | 10.1% |
| 2030 | 374 | 8,000 | 4.7% |
The model's figure splits into external flights sold and Starship flights deploying compute — 115 and 259 respectively in 2030 — and it is a floor rather than a manifest, because internal Starlink launches are carried as a cost and not as a unit anywhere in the model. Even so, closing the 2030 gap does not mean adding a few flights. It means 21× the cadence our model already stretches to reach.
The ladder and the tonnage target are not the same claim
Three days ago we converted SpaceX's 2.25 million tonne mass target into cadence: 31 to 62 Starship launches a day, depending on whether a flight carries 200 tonnes or 100. The obvious question is whether the new ladder confirms that arithmetic. It does not, and the way it fails is the interesting part.
8,000 flights at the 200 t Block 4 design target is 1.6 million tonnes — 71% of the 2.25M figure. At the ~100 t Block 3 actually flying today it is 800,000 tonnes, or 36%. To reach 2.25M tonnes on 8,000 flights, each flight has to carry 281 tonnes, which is 1.41× a design target that has never flown.
So the two claims are consistent only if a payload well above the published design target arrives with the cadence. This is where an article normally picks whichever number flatters the story. Both belong on the page: either the mass target sits later than 2030, or the ladder implies a Starship nobody has described. Nothing in the public record settles it.
The rung that can actually be graded
One flight a day by August 2027 is the whole forecast reduced to something falsifiable, and our model has already put a number on it — as a limit.
The Space AI vertical's flightsCeiling is 91 a quarter, and the model's own rationale calls it a ceiling, not a waypoint. The projection reaches 39 flights a quarter by end-2028, 57 by end-2029, 69 by end-2030 and 73 by 2031 Q2. It never gets there inside the horizon. Musk's statement is that SpaceX gets there within twelve months of August 2026, four and a half years earlier, with a vehicle that has flown twice this year.
Set the model to one flight a day immediately — start and ceiling both at 91 — and fair value goes from $187.12 to $194.59. The rung is worth $7.47 a share.
Which sets up the only comparison this piece really needs. The ladder is worth $313.81 a share and cannot be graded until 2030. The rung is worth $7.47 and can be graded next August. The market is being asked to pay for the first on the evidence of the second — and the second is a 42-to-1 smaller claim.
The satellite is now disclosed, and our model is too generous
The same June 8 talk unveiled AI1: a compute satellite with a 150 kW peak and 120 kW average payload, 70 kW per tonne, and a 70-metre wingspan.
Our model assumes 40 satellites a flight at 160 kW each. Against the disclosure that is 6.7% above AI1's peak and 33% above its average — so the model's 6.4 MW a flight should be nearer 4.8 MW. Correcting the input to the disclosed 120 kW average costs $2.33 a share.
The satellite count runs the other way. At 70 kW per tonne, an AI1 masses 1.7 to 2.1 tonnes, so a 100 t payload is 47 to 58 satellites by mass and a 200 t payload is 93 to 117 — against the 40 a flight our model carries. That is not an error: the cadence piece already showed this line is volume-limited rather than mass-limited, and a 70-metre wingspan is exactly what fills a bay before it fills a mass budget. But the two inputs now have a disclosed anchor on one side and none on the other, which is a model revision worth making deliberately under define-thicker-model rather than inside an article.
The gigawatt ladder is the same shape, further out
The orbital data centre goals scale with cadence, because in this model orbital capacity is cadence: flights × 40 satellites × 160 kW.
| Orbital AI capacity | SpaceX goal | R40 model | Model as a share |
|---|---|---|---|
| 2027 | 1 GW | 0 | Our Space AI line does not open until 2028 Q1 |
| 2028 | 10 GW | 0.38 GW | 3.8% |
| 2029 | 100 GW | 1.03 GW | 1.0% |
For scale on the 100 GW: our model has SpaceX's terrestrial AI fleet at 13.0 GW energised by end-2029, after five years of compounding a 400 MW-a-quarter build rate. The 2029 orbital goal is 7.7× the entire ground business the model projects for the same date — in a business line that has not launched its first satellite.
What the model says, in both directions
Base case fair value is $187.12 a share against $146.23 as of August 19, a 28% upside, on a 20× exit multiple and an 11% discount rate.
| Change from the base case | Fair value | Effect |
|---|---|---|
| Space AI removed entirely | $177.81 | −$9.31 |
| AI1's disclosed 120 kW, not our 160 kW | $184.79 | −$2.33 |
| Half a turn off the exit multiple | $182.32 | −$4.80 |
| External flight growth 4% → 6% a quarter | $189.53 | +$2.41 |
| One flight a day, immediately | $194.59 | +$7.47 |
| SpaceX's ladder: 8,000 flights a year | $500.93 | +$313.81 |
Read the top and the bottom of that table together. Everything orbital that the model currently believes is worth $9.31 a share — about one turn of the exit multiple, since half a turn is $4.80. Everything orbital that SpaceX has said is worth $313.81. The entire distance between a $187 stock and a $501 stock is a launch rate, and the launch rate has one public checkpoint before 2028.
What to watch
- Starship flights in the next four quarters. Two so far in 2026. One a day by August 2027 requires roughly 90 in the fourth of those quarters alone; anything under about 20 in Q2 2027 makes the rung unreachable arithmetically, not just unlikely.
- Whether the ladder appears in a filing. A 10-Q, a shareholder letter or an earnings-call restatement of 1,000 / 3,000 / 8,000 converts it from a technical-talk figure into something close to guidance, and it changes how this model should treat it.
- A published Starship payload figure for the block that flies the ladder. 281 tonnes a flight is what the mass target and the launch target imply together. The design target on the record is 200.
- The first AI1 on orbit, and its measured average power. 120 kW is a specification. The model runs on 160 kW and should not.
- Falcon's decline. SpaceX's own commercial sales VP has said 2025 and 2026 are likely the peak of Falcon activity. Our Space segment grows external flights 4% a quarter through the horizon on the assumption Starship takes the manifest over; if Falcon falls before Starship arrives, that line breaks in the middle.
Why this is not the same article as the last one
Three days ago the question was whether a mass target was plausible. It is now something narrower and more useful: a public company has put a launch ladder on the record, and there is a four-and-a-half-year disagreement between that ladder and the model this site publishes. One of the two is wrong, and it is genuinely not obvious which — our ceiling of one flight a day may be the timid assumption, in which case the model owes the reader a revision.
What is not in doubt is where the value sits. A launch business that flies 42 paying missions a year is worth what it is worth. An orbital compute business that needs 8,000 flights is worth $314 a share on our own arithmetic, and rests on a figure given at a technical talk the day before the shares were sold.
Sources and provenance. The 1,000 / 3,000 / 8,000 launch ladder and the 1 GW / 10 GW / 100 GW orbital data centre goals are claimed — stated by Elon Musk at a technical talk on June 8, 2026 on manufacturing, launching and operating AI satellites at scale, reported by nextbigfuture and carried by Advanced Television on June 9; they appear in no SpaceX filing. AI1's 150 kW peak, 120 kW average, 70 kW per tonne and 70-metre wingspan come from the same talk and are claimed on the same basis. Musk's "at least one flight a day" within a year is disclosed on the record — SpaceX's Q2 2026 earnings call of August 4, 2026, reported by Space.com and Aviation Week. The 165 Falcon 9 flights in 2025, of which 123 were Starlink, and the commercial sales VP's remark that 2025–2026 is likely peak Falcon are press-reported by SpaceNews and Space.com. The 170 flights in 2025, ~158 annualising in 2026, 0.43 a day, the 2.25 million tonne target and the 31-to-62-launches-a-day conversion are ours, from the cadence piece of August 19, where the 2.25M figure's own provenance — an enthusiast post Musk agreed with — is set out. Every launch count attributed to the model, the 91-a-quarter ceiling, the 40 satellites a flight at 160 kW, the $187.12 base case, the $9.31, the $7.47, the $313.81 and every row of the sensitivity table are R40 model output on the assumptions published at /stocks/spcx/model/ — they are our estimates, not SpaceX figures, and the model discloses none of them as fact. The $146.23 price is the snapshot the model was struck against on August 19 and is not a live price. The 281 tonnes a flight, the 1.6M and 800,000 tonne totals, the 21.9 flights a day and every percentage in this piece are R40 arithmetic on the figures above.
