How Cheap Is Starship V3 Really? A Plain-English Look at the 2026 $/kg-to-LEO Math

2026-06-09 · by Priya Raman · News Editor · Curatorial · Article

Starship V3 flew its first integrated flight test in May 2026. The vehicle is taller than a Saturn V, lifts more than 100 metric tons to LEO in its expendable configuration, and runs Raptor 3 engines that SpaceX says are designed for "airliner-style" rapid reuse. Every commercial number in the launch business is being quietly rewritten around it. Here's the math, in plain English.

The headline number, and why it isn't the whole story

SpaceX has talked about a target of <$200 per kilogram to LEO for Starship since 2019. In May 2026, after Flight 12, Musk put the "fully reusable" target at "somewhere between $100 and $200 per kg," with a 100+ ton payload. The industry spent the next week arguing about whether that was aspirational, achievable, or marketing. All three are probably true in different proportions.

To put that in context, here is what a kilogram to LEO has actually cost over the last four decades:

|---|---|---:|---|

| 1980s | Space Shuttle | ~$54,000 | Partially reusable, government program |

| 2000s | Delta II / Atlas V | ~$15,000-25,000 | Expendable, government-heavy manifest |

| 2010s | Falcon 9 (expendable) | ~$5,000-6,000 | First commercial re-entry pricing |

| 2020s | Falcon 9 (reused) | ~$1,500-3,000 | The current SpaceX list price |

| 2025 | Falcon Heavy (reused) | ~$1,200-2,000 | Side boosters reused, center expended |

The thing to notice is the slope of the line on a log scale. Each generation of vehicle took a real step down. Starship V3 isn't just another step — it would be a step that lands on a different floor entirely.

But here's the part that gets lost in the headlines: $/kg is not the same as cost to you. A kilogram of payload on Starship V3 is a kilogram in a 9-meter-diameter fairing. Most commercial customers aren't paying SpaceX's marginal cost — they're paying whatever SpaceX is willing to charge, which is set by the next-cheapest alternative. The actual price per kg won't be $200 for years, even if the actual cost gets close.

What's actually new in V3

Starship V3 is not a marketing refresh. The configuration that flew in May 2026 has three structural changes from V2:

Raptor 3 engines. The new closed-throttle, full-flow staged combustion engine is designed for double-digit reuse counts and rapid turnarounds. Each Raptor 3 is rated at ~280 tf at sea level, up from ~225 tf for Raptor 2.
Increased propellant mass. V3 carries more than 100 metric tons to LEO in expendable mode, against V2's ~75 tons. The stretch comes from longer propellant tanks, not from a wider diameter.
Pad 2 at Starbase. The new orbital launch mount is the first one designed for high-cadence reuse from day one. Pad 1 is being converted to a test article.

There is also a quieter change: the cost of a Raptor 3 has come down versus Raptor 2. SpaceX's internal numbers (leaked, then semi-confirmed in a 2025 industry talk) put Raptor 3 production cost at roughly 60% of Raptor 2 at the same thrust. The engine is the most expensive single component of the vehicle. This is where most of the marginal cost reduction is actually coming from.

The reuse math

Starship's economic case rests on a single word: reuse. Every flight that uses the same hardware twice halves the amortized hardware cost of that flight. Every flight that uses it ten times brings the amortized cost to 10% of new-build.

This is not a new idea — Falcon 9 has been doing it for a decade. But Falcon 9 reuses a first stage that costs about $30 million of a $67 million launch. The fairing, second stage, and integration are still mostly expendable. The savings from reuse top out around 30-40% of the total launch price.

Starship is fully reusable by design. Both stages, the heat shield, even the flaps are meant to be reflown. That is the structural difference. Theoretically, the entire vehicle is a $5 million asset amortized over many flights, with the marginal cost of a flight being mostly propellant, ground ops, and refurbishment.

How cheap is propellant? Starship V3 holds about 3,400 metric tons of methalox (liquid methane + liquid oxygen). Methane is roughly $0.50/kg industrial scale, LOX is around $0.20/kg. Total propellant cost for a flight: under $2 million. Add another $1-3 million for ground ops, range support, and refurbishment, and you have a marginal cost floor somewhere around $3-5 million per flight.

Divided by 100 tons to LEO (expendable), that's $30-50/kg. Divided by 50 tons (a realistic reusable number for an early-2020s V3), it's $60-100/kg. Add SpaceX's profit margin and corporate overhead, and you get to the "$200/kg list price" range SpaceX has been quoting.

The math works. It also requires SpaceX to hit the operational cadence and reuse counts they say they're targeting. Neither has happened at scale yet.

What this means for everyone else

The launch industry is being repriced around Starship in real time, and almost no one is happy about it. Some examples:

ULA Vulcan is selling launches for $110 million. The vehicle has a payload class of ~27 tons to LEO. That's roughly $4,000/kg. To compete, ULA has been talking about SMART reuse (sensible, modular, autonomous return technology) for the BE-4 engines, but Vulcan will never match Starship on price.
Blue Origin New Glenn is targeting 45 tons to LEO and reuse on the booster. Even at full reuse, the price floor is going to be in the $1,000-2,000/kg range. New Glenn is positioning for national-security and Blue Origin's own internal manifests (Kuiper / "Amazon Leo" launches, plus Blue Moon) rather than competing head-on with SpaceX on commercial cost.
Rocket Lab Neutron — Rocket Lab's mid-lift vehicle, first flight expected late 2026. Neutron is targeting 13 tons to LEO at a price of "well under $1,000/kg" by the mid-2020s. This is the strongest near-term commercial competitor in the 1-15 ton class.
Ariane 6 and the Chinese Long March family are priced $4,000-10,000/kg. They will continue to have sovereign, government, and Asia-Pacific customers, but they cannot compete on cost for commercial LEO constellations.

The knock-on effect: launch is no longer the limiting factor for almost any LEO mission. If you can build a satellite that fits in Starship's 9-meter fairing and weighs under 50 tons, launch is suddenly a small line item in your budget. That has implications for:

LEO broadband constellations. Starlink and Amazon Leo both gain. But so does every regional competitor (Telesat Lightspeed, China's Guowang, EU's IRIS²).
Space stations. Axiom, Vast, Voyager — all the commercial station developers assumed launch was a constraint. It is becoming one less constraint.
In-orbit servicing. If you can launch 50 tons at a time cheaply, the case for refueling a satellite in orbit weakens, but the case for launching large servicing vehicles to GEO strengthens.
Lunar and beyond-LEO. Starship is NASA's choice for the Artemis human landing system. Every other deep-space architecture is now being mentally repriced.

The Starlink multiplier

There is a wrinkle most coverage misses. The single largest customer for Starship V3 is SpaceX itself — for Starlink. As of mid-2026, Starlink has ~7,500 satellites on orbit. Version 3 Starlink satellites are larger and heavier than V2. With reuse, SpaceX is targeting a launch cadence of one Starship flight every week to ten days for Starlink replenishment and growth.

At that cadence, the marginal cost of a Starlink launch is not $5 million. It is the cost of one Starship vehicle amortized over maybe 10-20 flights plus a few days of pad time. The internal transfer price SpaceX charges Starlink is rumored to be on the order of $15-25 million per launch for 50-70 tons of payload. That is $200-500/kg to LEO, internal transfer price.

It is very hard to compete with that. It is also very hard to verify until SpaceX eventually IPOs Starlink, which is rumored for late 2026 or 2027.

What could go wrong

The case above assumes Starship V3 hits its reuse and cadence targets. Several things could derail that:

Raptor 3 reliability. The engine is new. Each Raptor 3 flight today is instrumented to the teeth, and SpaceX is deliberately over-engineering for the first 30-50 flights. A high-profile failure (a crewed Raptor 3 anomaly, a payload loss) could push the schedule by 6-12 months.
Heat shield refurbishment. The ceramic tile system is the part of Starship that nobody has a good operational model for yet. Flight 12 flew with a redesigned tile pattern; refurbishment time per flight is the single largest unknown in the marginal-cost calculation.
Regulatory cadence. FAA launch licensing is the bottleneck on actual flight rate. Even if SpaceX is ready to fly every 5 days, the FAA's range schedule and environmental review pace sets the ceiling. Expect this to be the subject of significant political friction in 2026-2027.
The competition surprises us. Blue Origin's New Glenn has flown successfully. Rocket Lab's Neutron is a real vehicle. Relativity's Terran R is still a question mark. China's Long March 9 is a question mark. The cost curve may not bend as cleanly as SpaceX hopes if a serious competitor emerges in the 50-100 ton class.

The number to watch

If you only track one metric in the launch industry in 2026, it is the average number of Starship flights per quarter. As of mid-2026, that number is 1-2. The implicit cost math in this article assumes it becomes 8-12 by mid-2027. If it does, the entire LEO economy reprices. If it doesn't, Falcon 9 reuse remains the practical ceiling for another 3-5 years.

Try it yourself

If you want to see how the math changes when you flex the assumptions — payload mass, reuse count, refurbishment cost, propellant cost — the Orbital Market sim lets you model launch economics against constellation revenue. Set Starship V3 to "$200/kg, 50 tons reusable, one flight per week" and watch the LEO broadband market become one of the most profitable industrial sectors on Earth.

Orbital Market — model the launch economics yourself →