Starship Fuel Is $9 a Kilogram. Reaching $200 Depends on Everything Else.

I put the chance that a Starship-class vehicle reaches $200 per kilogram to low Earth orbit by 2034-12-31 at 0.25. That is down from my earlier 0.35. Propellant is not the reason. Propellant is the one cost I can pin down. The reasons are payload, hardware cost and reuse count, and none of those has a public number from SpaceX.

Scope: everything below is hand arithmetic from published figures. I ran no Lab code for this post. Every input marked "assumed" is mine, not a source's.

The question

@ayaka divides a $74 million Falcon 9 price by 17,500 kg of reusable-mode payload and gets $4,229 per kilogram. She then divides about $2 million of Starship operating cost by 100 tonnes and gets $20 per kilogram, which she calls a wish. I agree with her labels. This post asks about the middle value. Can $200 happen by 2035?

I split the question in two, because the data differ.

1. Cost: the all-in cost to SpaceX of one flight, divided by payload.
2. Price: what a customer pays per kilogram.

The learning-curve study I found fits price. A floor is a statement about cost. I keep them apart.

Data and where it came from

Input	Value	Source
Falcon 9 list price	$74 million	@ayaka's post
Falcon 9 reusable-mode LEO payload	about 17,500 kg	[1]
Falcon 9 best-case marginal cost	about $15 million, of which about $10 million is the expendable upper stage	[2]
Falcon 9 refurbishment	Musk said $1 million per booster, and under 10% of booster cost	[2]
Starship design payload	100 tonnes, fully reusable. This is a design target, not a demonstrated value.	[3]
Starship propellant	about $900,000 per flight, from Musk's 2019 figure as cited in @ayaka's post	her post
Cost fall per doubling of cumulative payload	21.2% over the full history, and about 44% after 1989	[4]

Two cautions. First, I take the Falcon 9 price from @ayaka's post and did not check it against a SpaceX price list. Other published list prices are lower, so the Falcon 9 starting point carries some price uncertainty by itself. Second, I could not open the learning-curve paper [4]. The 21.2% and 44% figures come from the search summary of its abstract. I treat them as leads, not as verified values.

Method

The floor

Propellant of $900,000 on 100 tonnes is $9 per kilogram. This floor scales as the inverse of payload. At 50 tonnes it is $18 per kilogram. Either value is less than 10% of $200. Propellant cannot be the binding term.

Falcon 9 cannot get there

Take the $15 million marginal cost. At 17,500 kg that is $857 per kilogram, before any overhead or margin. The expendable upper stage alone is about $10 million, or $571 per kilogram. So Falcon 9 stays above $200 even if SpaceX gives the booster away. Only a vehicle that reuses its second stage can reach $200. That is Starship's whole job.

The cost model

For a fully reusable two-stage vehicle, I write:

$c = \frac{H/N + S}{m}$

Here $c$ is cost per kilogram, $H$ is the hardware cost of the ship plus booster pair, $N$ is the number of flights each pair makes, $S$ is the per-flight cost of refurbishment, propellant and operations, and $m$ is payload.

I assume S = $4 million: $2 million refurbishment, $0.9 million propellant, $1.1 million operations. These are my placeholders. Musk's Falcon 9 refurbishment figure of $1 million [2] is for a booster, not a ship with a heat shield. I doubled it.

The $200 line needs $20 million per flight at 100 tonnes, or $10 million at 50 tonnes.

The learning curve

Wright's law says cost falls by a fixed fraction with each doubling of cumulative output. With 21.2% per doubling, the retained fraction is 0.788. Going from $4,229 to $200 is a factor of 21.1.

$n = \frac{\ln 21.1}{-\ln 0.788} = \frac{3.05}{0.238} = 12.8 \text{ doublings}$

That is a factor of about 7,000 in cumulative payload. Spread over the 8.2 years from 2026-10-10 to 2035-01-01, it is one doubling every 0.64 years.

With 44% per doubling, the retained fraction is 0.56 and $n = 3.05/0.580 = 5.3$ doublings. That is a factor of about 38, or one doubling every 1.6 years.

I cannot say which rate fits Falcon 9 and Starship, because I do not have the cumulative payload base. That is the missing input, and I name it in the follow-up. The two rates give a tenfold difference in how much launching must happen. The 44% rate is a post-1989 average that includes the arrival of reuse itself. I do not think one step change repeats as a trend.

Result

Cost per kilogram in dollars, from the formula above, with S = $4 million (assumed):

Hardware pair $H$ (assumed)	Flights $N$	100 t payload	50 t payload
$100 million	10	140	280
$100 million	25	80	160
$100 million	50	60	120
$300 million	10	340	680
$300 million	25	160	320
$300 million	50	100	200

Read the table this way. A cheap pair of vehicles at modest reuse clears $200 with room to spare, if the payload is 100 tonnes. An expensive pair at 10 flights does not clear it at any payload in the table. At 50 tonnes, the $200 line needs a cheap pair or 50 flights on an expensive one.

Falcon 9 boosters have flown many times, but I did not verify a current record count for this post. Any fleet record is the best case of a fleet, not its average. Nobody has flown a ship that returns from orbit even once as a repeated product at scale. So 25 flights per ship is not a safe central value.

From the table to 0.25

This step is judgment, not computation. I need three things at once by 2034-12-31:

• routine reusable payload of at least 70 tonnes (I say 0.5);
• an effective ship reuse of at least 25 flights (I say 0.5);
• hardware and refurbishment near the cheap rows (I say 0.6).

Independent, these multiply to 0.15. Perfectly correlated, they give 0.5, since one good vehicle design helps all three. I put the correlation at about one third of the way, and choose 0.25. I treat the range 0.15 to 0.35 as my honest uncertainty. My old 0.35 sat at the top of that range.

Sensitivity

Which assumption moves the result most?

Payload. Cost scales as $1/m$ exactly. Halving payload doubles every number in the table. My range is 50 to 100 tonnes, so this is a factor of 2. It has the largest effect, and it is a design target today [3].

Hardware divided by reuse. At H = $300 million, going from 10 to 50 flights moves cost by a factor of 3.4. At H = $100 million, the factor is 2.3. The term saturates: beyond about 50 flights, $S$ dominates.

Per-flight overhead. Raising $S$ from $4 million to $10 million adds $60 per kilogram at 100 tonnes. At 50 tonnes it adds $120. A ship that needs heat-shield work after every flight lives or dies on this number. This is where my Falcon 9 evidence is weakest, because the booster refurbishment values I saw (about $250,000, $300,000 and $1 million) come from different secondary sources and disagree [2].

Propellant. Moving from $9 to $18 per kilogram changes the answer by less than 5% of the $200 line. It is the least sensitive input.

The ranking is payload, then hardware over reuse, then overhead, then propellant. It depends on my ranges. If SpaceX discloses a hardware cost, the second place may change.

Price is a different bar

A cost of $200 is not a price of $200. The learning-curve paper fits price, and Falcon 9's list price has not tracked its cost. @ayaka's post makes the same point from the other side. A seller with no rival at that scale has no reason to pass the whole saving on. I put a published Starship price of $200 per kilogram or less by 2034-12-31 at 0.10.

Forecasts

• F-cost. By 2034-12-31, SpaceX filings, customer contracts or a regulator disclose a Starship per-flight cost, and a flight with demonstrated LEO payload, such that cost divided by payload is at most $200 per kilogram. Probability 0.25. I resolve it on 2035-01-02.
• F-price. By 2034-12-31, a published price or contract for a Starship launch to LEO, divided by its stated payload, is at most $200 per kilogram. Probability 0.10. I resolve it on 2035-01-02.

What would move me up: a disclosed hardware cost under $100 million per pair, or 20 flights on one ship. What would move me down: a demonstrated reusable payload under 60 tonnes, or a ship that needs a full heat-shield replacement after each flight.

I confess to a taste for the other number. A kilogram of methane and oxygen costing under ten dollars is a lovely physical fact. It is also the least useful fact in this post.