Two astronauts spent 286 days on the International Space Station. Their spacecraft had flown with a helium leak and thrusters that underperformed, and the ground environments behind both faults were never qualified for what the vehicle met in orbit. On Monday, NASA and Boeing put that capsule back on a flight manifest. The first launch is uncrewed. The crew launch is dated 2028.
Sixty-one recommendations came out of the investigation. Most of the fixes are scheduled after the next flight, not before it.
That ordering is the whole story.
What the 2024 flight test actually found
Butch Wilmore and Suni Williams launched on 5 June 2024 on a mission billed as ten days. Starliner reached the station on 3 July 2024. The crew stayed 286 days.
Two things went wrong in transit. Helium leaked. Reaction control thrusters underperformed, traced to a poppet seal that extruded under conditions nobody had qualified on the ground. NASA sent the capsule home without anyone in it and reclassified the mission as a Type A mishap, the most severe rating in its scheme, the one reserved for crew-endangering events.
Two uncrewed test flights had come before it. Both had problems too.
It's the underlying pattern. It was inadequate systems engineering and integration, creating gaps in hardware qualification.— Warren "Woody" Hoburg, NASA astronaut, commander of Starliner-2
Hoburg will command the flight that puts astronauts back on the spacecraft. He named the cause as a pattern rather than a part, and pointed at the ground: the environments behind the helium leak and the thruster problem were never qualified for what the vehicle met in flight.
That is an expensive kind of finding. A single bad part gets replaced. A qualification regime that does not match the flight environment has to be rebuilt, and every subsequent test inherits the doubt.
The propulsion fix is named, narrow, and scheduled after the next flight
Here is the part worth reading twice. Starliner-1 is uncrewed. It launches as early as December 2026 or January 2027, carrying cargo. Its job is to gather data on the thermal environment around the propulsion system, using thermal modifications Boeing has already made to the service module.
Then, after the capsule comes home, the thruster valve redesign happens.
Program Investigation Team actions
Recommendations issued to NASA after the 2024 crewed flight test forced an uncrewed return. · NASA, February 2026
The specific defect has a name. Poppet seal extrusion, degrading thruster performance. The fix is a service module thruster valve design modification, agreed jointly by NASA and Boeing and explicitly deferred until after Starliner-1 returns. Beyond it sit new crew module thrusters, new batteries, and small changes to the parachute system.
Read as an engineering sequence, the logic holds. You do not validate a thermal modification and a valve redesign in the same flight and call the result verified. You separate them, so that when a number moves you know which change moved it.
The deferred-thruster-window question
It does gate crewed flight, which is why Starliner-2 sits in 2028 rather than 2027.
What to watch: whether NASA reports the thermal data as nominal, partial, or inconclusive. An inconclusive thermal result moves the valve work into a second uncrewed demonstration flight and pushes crew beyond 2028.
The uncomfortable part is what "validate" means here. NASA's Dana Weigel framed Starliner-1 as the step that closes the loop on thermal performance, the element she called a key part of certification.
Starliner's next flight is a critical step on the path to achieving full system certification and ensuring a sustained human presence in low Earth orbit.
A single uncrewed demonstration does not certify a crew system. It produces the data that makes certification arguable. Between those two states sits the entire programme risk, and it now has a date attached to it: a window opening in December and running into January.
The certification gate moved from hardware to evidence. What failed in 2024 was a qualification regime, not a component, and only an on-orbit thermal result closes that gap.
The $359M buys two things at once: a capsule fix and a human-rated replacement launcher. The launcher half has no published schedule, which makes it the quieter risk in the announcement.
Sixty-one recommendations, $359 million
Two numbers frame the cost of the restart, and only one of them is new.
The sixty-one recommendations are the corrective actions NASA identified in February 2026 after the crewed flight test returned uncrewed. The second number is the increment Weigel announced on Monday.
Upgrades plus launcher certification
Incremental cost of the Starliner hardware changes combined with human-rating the replacement rocket. · Weigel, NASA, September 2026
Three hundred and fifty-nine million dollars covers the spacecraft upgrades and the Vulcan certification together, which keeps two separate programmes inside one number.
Neither figure is small. Both are smaller than the alternative NASA was implicitly weighing, which is a single-provider future for American crew launch.
A second rocket to certify
The launcher transition is the piece most likely to be underestimated. Starliner-1 through Starliner-4 use the remaining Atlas V rockets assigned to the programme. Starliner-5 and Starliner-6 move to Vulcan Centaur, and Vulcan is not yet certified to carry astronauts.
| Parameter | Atlas V, flights 1–4 | Vulcan Centaur, flights 5–6 |
|---|---|---|
| Certification state | ◐ Human-rated, programme rockets assigned | ✗ Not human-rated |
| Additional work required | ✔ None beyond the capsule | ✗ Full human-rating with ULA and Boeing |
| Funding source | ◐ Existing contract | ✗ Inside the $359M increment |
Launcher assignment and certification state as stated by NASA and reported by Aerospace Global News, September 2026.
So the agency that just admitted a qualification gap on the spacecraft is funding a second human-rating effort on the rocket, on a schedule that has already moved once. These are different engineering cultures inside the same programme.
Why NASA kept a capsule that failed its own test
The commercial crew contracts went to Boeing and SpaceX in 2014, on the premise of two independent American systems carrying astronauts to low Earth orbit. Starliner exists to be the second one.
NASA Administrator Jared Isaacman argued the demand case on Monday rather than the engineering case.
The agency issued a statement on Monday too, describing close work with Boeing on the issues found on previous flights and an intention to see the vehicle return to flight for the station and future commercial destinations.
Isaacman's own case was about scale, not hardware. He argued the low Earth orbit economy is opening, and that demand for vehicles carrying astronauts will grow with it. That is a market argument for redundancy, and a weaker argument for this particular capsule, in this particular decade, than it was in 2014. The gap between the two is exactly what the $359 million and the 2028 date are paying for.
Hoburg put the near-term version of the same problem more plainly. Correct the technical deficiencies, then return to crewed flight. The sequence is not in question. The date is.
We wrote about a different NASA programme earlier today, a quantum gravity gradiometer measuring groundwater from orbit. Different budget line, same instinct: own the capability you cannot outsource.
Does Starliner-1 actually fly before the year ends?
Probability: 70% — NASA and Boeing have both named the December–January window for the uncrewed flight, the capsule is already built, and the remaining thermal and propulsion work does not gate that launch. The crew date is far softer, because it sits behind a valve redesign with no published schedule.
✅ Arguments for
Boeing committed publicly to hardware enhancements and added human-spaceflight certification resources in the same statement NASA used to announce the plan.
Splitting the thermal validation from the valve redesign is the correct sequencing, and sequencing is what failed in 2024.
Confirmation criteria: Starliner-1 launches inside the named window and returns with thermal data NASA publishes as nominal.
❌ Arguments against
The corrective fixes land after the flight that is meant to validate the approach, so Starliner-1 can succeed and still not move the crew date.
Vulcan human-rating runs in parallel on a second schedule, with no published milestone dates.
Disconfirmation criteria: the December–January window closes with no launch attempt, or Starliner-1 returns with thermal or propulsion data NASA declines to characterise.
Scenarios
🟢 Clean validation (25%)
Consequences: NASA holds two certified crew systems before its replacement launcher is even rated, and the 2028 date survives with months of margin.
🟡 Partial result (55%)
Consequences: redundancy arrives about a year later than planned, and the Atlas V to Vulcan handover overlaps the crew certification instead of following it.
🔴 Second gate (20%)
Consequences: NASA holds a capsule it cannot certify, having spent $359M to arrive at the same gate twice, and ISS crew rotation stays dependent on a single provider for another three years.
Whether NASA characterises Starliner-1's thermal data as nominal, partial, or inconclusive
How many of the 61 recommendations NASA marks closed rather than deferred
The first published human-rating milestone for Vulcan Centaur
Whether the Atlas V flight count in the remaining programme matches the Starliner-1 to -4 plan
Any crewed-flight certification date Boeing commits to in writing rather than in a briefing