SpaceX Spent $15 Billion on Starship. Falcon 9 Cost $400 Million. One of Them Can Land.
SpaceX invested 37.5 times more developing Starship than it spent building Falcon 9. After 13 test flights, the booster still cannot land, the rocket cannot reach orbit, and Starlink’s average revenue per user has fallen 33% since 2023. The S-1 filing says the entire growth strategy hinges on achieving full reusability. On Friday, the booster exploded on impact with the Gulf of Mexico. Again.
Fifteen billion dollars is how much SpaceX has spent developing Starship, according to the company’s own S-1 registration statement filed with the SEC in May, a figure first reported by Reuters that dwarfs the roughly $400 million SpaceX spent building Falcon 9, the rocket that has flown more than 400 missions and made commercial spaceflight boring in the best possible sense, and that, when you divide one by the other, reveals a 37.5× ratio of development capital consumed by the successor versus the vehicle it is supposed to replace.
On Friday evening at 6:51 p.m. Eastern, SpaceX launched Starship on its thirteenth test flight from Boca Chica, Texas. After stage separation, the Super Heavy booster executed a boostback burn and attempted a simulated landing in the Gulf of Mexico. It could not fire all its engines for the landing burn and hit the water too fast and exploded, which marks the second consecutive V3 booster failure after the booster encountered a different failure during stage separation in May. The pattern is unmistakable: SpaceX’s largest-ever rocket keeps destroying its largest-ever booster.
The upper stage fared better, deploying 20 third-generation Starlink satellites, surviving atmospheric reentry, performing a simulated landing in the Indian Ocean, and, for the first time in the program’s history, not exploding when it tipped over into the water. A drone examined the heat shield tiles while it floated, capturing images that SpaceX engineers will study for months as they work toward controlled upper-stage recovery and the fully reusable architecture that the entire business case depends on.
Twenty minutes later, all 20 V3 Starlinks burned up in the atmosphere because Starship still cannot reach Earth orbit, which means SpaceX burned approximately $24 million in satellite hardware and an entire booster to collect flight data on a rocket that, thirteen flights in, has yet to deliver a single payload to a stable trajectory.
The $37.50 Ratio
Falcon 9 cost $400 million to develop, has flown more than 400 commercial missions generating an estimated $20 billion in launch revenue across government contracts, commercial satellite deployments, and Starlink missions, and when you divide the development cost by the flights the amortized R&D per mission comes to roughly $1 million, which by aerospace standards is an extraordinary return on engineering investment that no other launch vehicle in history has matched.
Starship has cost $15 billion and completed zero commercial flights, which means the amortized development cost per flight is currently undefined because the denominator is zero. When commercial operations begin, the number will start absurdly high and decline toward something defensible only if SpaceX achieves the launch cadence Elon Musk has described: thousands of flights per year, a figure that Morgan Stanley analyst Adam Jonas has noted would exceed all of humanity’s 330 orbital launches in 2025 by nearly an order of magnitude.
At 167 fully reusable flights, Starship’s development cost amortizes to $89.8 million per flight, at 1,000 flights it drops to $15 million, and at 10,000 flights $1.5 million, a progression that is simple to calculate and ferociously difficult to engineer.
The S-1 Says It All Depends on This
SpaceX’s S-1 contains language that investors will eventually learn by heart: “Our ability to execute our growth strategy is highly dependent on the successful development and scaling of Starship and the ability to increase our launch cadence, both of which are subject to challenges and uncertainties inherent in the development and deployment of new and complex technologies.” If Starship does not work, the growth story collapses.
The filing revealed a company with three segments producing starkly different financial outcomes in 2025: Starlink generated $11.4 billion in revenue with $4.4 billion in operating income and a 63% EBITDA margin, the Space segment produced $4.1 billion in revenue against a $657 million operating loss dragged down by Starship R&D, and the AI segment, absorbed after SpaceX acquired xAI in February 2026, contributed $3.2 billion in revenue against a staggering $6.4 billion operating loss that, in a single line item, explains why a company generating nearly $19 billion in revenue still posted a $4.9 billion consolidated net loss for the year. Starlink is the sole profitable business.
The ARPU-Reusability Scissors
Here is the math that should concern anyone holding SPCX shares at $115, down 15% from the IPO price and nearly 43% from the post-listing peak above $200.
Starlink’s average revenue per user has been falling steadily as SpaceX expands into lower-priced international markets:
| Period | ARPU | Change |
|---|---|---|
| 2023 | $99 | — |
| 2024 | $91 | −8.1% |
| 2025 | $81 | −11.0% |
| Q1 2026 | $66 | −18.5% |
Source: SpaceX S-1 filing. Q1 2026’s drop is the steepest in the series, coinciding with aggressive expansion in Southeast Asia, sub-Saharan Africa, and South America where broadband pricing cannot support $99 monthly plans.
Simultaneously, Starlink’s capital expenditure is enormous and growing: SpaceX has invested $11.4 billion in satellite infrastructure since the beginning of 2023, a sum that precisely matches Starlink’s 2025 annual revenue and that reflects the punishing reality that the constellation needs to replace approximately 20% of its satellites each year just to maintain service as older units deorbit, meaning every dollar of Starlink revenue has been matched by a dollar of satellite capex in what TechCrunch’s Tim Fernholz aptly called a treadmill that never stops.
These two lines form scissors. Revenue per subscriber falls while the cost of maintaining and expanding the constellation does not, and reusability is the mechanism that is supposed to close them by dramatically reducing launch costs, and without it they keep opening until the business model snaps.
What Reusability Is Actually Worth
Satellite market analyst Tim Farrar of TMF Associates estimated in a May 2026 client note that an expendable Starship launch would cost approximately $100 million, or about $1,000 per kilogram, and wrote that at that price the vehicle would be “not much lower than Falcon 9, even if the full 100 ton capability is realized,” a conclusion that guts the economic rationale for the entire program if reusability remains elusive.
Musk’s target for a fully reusable Starship is $10 million per launch, roughly $100 per kilogram, which means the gap between expendable and reusable is a factor of ten, and that factor of ten determines whether Starship represents a revolution in spaceflight economics or an extraordinarily expensive way to do what Falcon 9 already does at a fraction of the development cost.
We can quantify the difference in capacity deployed per dollar using publicly available figures, and the results reveal just how much of SpaceX’s future value is concentrated in a single engineering outcome. Each V3 Starlink satellite delivers approximately 1 Tbps of downlink capacity and costs roughly $1.2 million to manufacture, according to SpaceX disclosures, and each Starship flight carries 60 of them, while a Falcon 9 carries approximately 23 of the smaller V2 Mini satellites at roughly 50 Gbps each:
| Configuration | Satellites | Capacity per launch | Total cost per launch | Gbps per $M |
|---|---|---|---|---|
| V2 Mini on Falcon 9 | 23 | 1,150 Gbps | ~$33.8M | 34 |
| V3 on expendable Starship | 60 | 60,000 Gbps | ~$172M | 349 |
| V3 on reusable Starship | 60 | 60,000 Gbps | ~$82M | 732 |
Reusable Starship deploys 21.5 times more bandwidth per dollar than the current Falcon 9 configuration, and even expendable Starship delivers 10.2 times more, owing entirely to the V3 satellite’s vastly greater capacity per unit. But V3 satellites are physically too large for Falcon 9 and can only fly on Starship, which means SpaceX has locked its next-generation constellation behind a rocket that, as of July 25, 2026, cannot deliver payload to a stable orbit.
The Break-Even Calculation
If the cost difference between an expendable and a fully reusable Starship launch is $90 million ($100M minus $10M), SpaceX needs 167 fully reusable flights to recoup the $15 billion development investment on savings alone, which at 24 flights per year, a cadence SpaceX has discussed for 2027, pushes break-even out roughly seven years, while the 2030 target of 100 flights per year compresses it to 20 months and the aspirational thousand flights per year brings it down to just two months, though all three timelines exclude the time value of money, ongoing R&D spending, and infrastructure maintenance that continues to burn cash regardless of flight rate.
They also exclude the $29.1 billion in total debt SpaceX disclosed in its S-1, including a $20 billion bridge loan that matures just 15 months after the IPO, a capital structure that leaves no room for a patient, multi-year development timeline at a company whose investors priced patience out of the deal by setting the offering at $1.75 trillion.
What Flight 13 Actually Proved
Strip away the stock reaction and the headline and Flight 13 delivered a mixed but informative signal. The upper stage performed its best flight yet, and surviving reentry, deploying payload, and floating intact after splashdown represent genuine engineering milestones that bring controlled upper-stage landing closer, while the fact that SpaceX communicated with all 20 V3 Starlinks during their brief time in space validated the deployment mechanism and satellite bus electronics even though the payloads burned up minutes later.
Booster failure is more concerning because it is the second V3 booster failure in two attempts, suggesting a systemic issue specific to the V3 variant rather than random bad luck. The V1 Starship booster was caught by the tower’s mechanical arms on Flights 5 and 7, demonstrating the concept works on an earlier design with different engines and a different flight profile, and the fact that two consecutive V3 flights have failed to land the booster means SpaceX is, in some sense, regressing on the capability it most needs to prove. The July 17 abort introduced a separate failure mode when frozen water seized four engine turbopumps, a problem SpaceX addressed by replacing six engines before the July 25 attempt, which then failed during the landing burn for an apparently different reason.
SpaceX’s “fly, fail, fix” development philosophy absorbs these failures by design, but the question is whether public markets will absorb them with equal equanimity, because as a private company SpaceX could fail a dozen boosters without consequence while as a public company trading at 40 times estimated 2026 sales with quarterly earnings disclosures two weeks away, every failed landing is a data point for short sellers and a drag on investor confidence that translates directly into cost of capital.
The Strongest Case for SpaceX
SpaceX has earned the benefit of the doubt in ways that deserve full acknowledgment: Falcon 9 boosters now land routinely after hundreds of missions, the company invented commercial booster reuse and proved skeptics wrong when the consensus position was that reuse would never be economical, and Starship V1 boosters were caught at the tower on two separate flights demonstrating the mechanical catch concept on an earlier design. V3 booster failures may be version-specific teething problems that SpaceX resolves within the next few flights, just as Falcon 9 required four years and multiple failures before achieving its first booster landing in December 2015, and the company’s track record of iterating through failure is the strongest evidence that this particular failure mode is temporary rather than architectural.
If reusability arrives at Musk’s target cost, everything flips. At 732 Gbps deployed per million dollars, V3 Starlinks on a reusable Starship would deliver 21.5 times more bandwidth per dollar than the current system, transforming Starlink from a business running on a capex treadmill to one with operating leverage that could justify a $1.75 trillion valuation and make every orbital ambition, from AI data centers to Artemis contracts to point-to-point travel, not only viable but fiercely profitable at scale.
The honest assessment is that SpaceX has more institutional knowledge about rocket reusability than any organization on Earth, and the counterargument against the company is not that they cannot do it but that they have not done it yet on this vehicle, and their own filing says the business depends on it.
Limitations
Several figures in this analysis rely on estimates rather than disclosures. SpaceX does not publish its internal Falcon 9 marginal launch cost; the $28 million figure used in the capacity table is derived from industry estimates that range from $15 million to $28 million. V2 Mini satellite capacity of approximately 50 Gbps is an industry approximation, as SpaceX does not publish per-satellite throughput figures. Tim Farrar’s $100 million expendable Starship estimate is one analyst’s projection; actual costs could differ materially. The Starship development total of $15 billion may include launch infrastructure with independent utility, meaning the per-flight amortization is somewhat overstated for the vehicle itself. Starlink’s ARPU trajectory may not continue its current decline if enterprise contracts, including airline and government deals, stabilize the mix.
The Bottom Line
SpaceX built Falcon 9 for $400 million and turned it into the most flown rocket in history, generating an estimated 50× return on development investment. It has spent $15 billion on Starship and has zero commercial flights to show for it. The S-1 filing makes the stakes explicit: without full reusability, the growth story falters, Starlink’s capex treadmill accelerates, and the V3 satellite constellation that is supposed to deliver 21.5 times more bandwidth per dollar stays locked behind a rocket that cannot reach orbit. At $66 ARPU and falling, SpaceX cannot afford to run the replacement treadmill at expendable-Starship prices for long. The ARPU-reusability scissors are opening. Every failed booster landing widens them. SpaceX’s first quarterly earnings report as a public company arrives in roughly two weeks. Investors will want to know a date for the first successful V3 booster landing. SpaceX will give them a Musk timeline. Both sides know the difference between those two things.
What You Can Do
If you hold SPCX shares, the first earnings call will be the most informative disclosure event in SpaceX’s history. Listen for three specific numbers: total Starship-related R&D spending in Q2, the projected date for the first booster catch attempt on V3, and Starlink subscriber additions in markets with ARPU below $50. If R&D spending is accelerating while subscriber growth in high-ARPU markets is decelerating, the scissors are widening faster than the public narrative suggests. If you are evaluating SpaceX as a competitive threat to terrestrial broadband, Amazon’s Kuiper, or other satellite operators, the capacity-per-dollar table above is your benchmark: V3 on reusable Starship delivers 732 Gbps per million dollars of launch-plus-construction cost. No competitor is within an order of magnitude of that number, but the reusable qualifier is doing all the work. Until that qualifier becomes a demonstrated fact rather than an engineering aspiration, discount accordingly.