Four point two billion dollars. That is what NASA agreed to pay Boeing in September 2014 to develop and fly a crew vehicle, and the significant word in the sentence is not the number. It is fixed.
A fixed-price contract means the price does not move. If the work costs more, the contractor absorbs it. Boeing has recorded charges against this program repeatedly since, and continues to.
Starliner launched with two astronauts on 5 June and docked with the space station the following day, having lost thrusters and developed helium leaks on the way. The argument here is that the contracting model, not the engineering, is the thing this flight is actually testing.
First: what happened on the way up
Be precise, because the reporting has been loose in both directions.
The vehicle launched with one known helium leak in the service module propulsion system, which had been analyzed before flight and judged acceptable. Two more appeared in orbit, then more. Helium is not a propellant; it pressurizes the tanks that feed the thrusters, and losing it eventually means losing the ability to fire them.
Separately, five of the small reaction control thrusters on the service module dropped out during the approach to the station. Flight controllers recovered four, and the crew flew parts of the approach manually. The vehicle docked about an hour later than planned.
Neither of those is trivial and neither prevented the mission. The crew is on the station and the vehicle is attached to it.
Second: why fixed-price was chosen
The contracting model is the reason this program exists in the form it does, and it was a deliberate reform.
NASA’s historical approach was cost-plus: the agency pays documented costs plus a fee, owns the design, and directs the work. That model built everything from Apollo to the shuttle and it has a known failure mode, which is that a contractor’s incentive runs toward more hours rather than fewer.
Commercial Crew inverted it. The agency stated requirements, ran a competition, and bought a service at a fixed price from two providers. The contractor owns the design, keeps the savings if it finishes cheaply, and eats the overrun if it does not.
That was expected to do two things: reduce cost to the government, and produce two independent vehicles so that neither could strand the program.
Third: what the model actually produced
On cost to the government it worked, unambiguously. NASA paid a fixed sum, has not paid more, and the overruns have fallen on the contractor’s shareholders rather than on the appropriation.
On schedule it did not. Both vehicles were contracted for 2017. One flew crew in 2020. The other is flying its crewed test now, in 2024, seven years late.
And here is the part that complicates the reform’s own case. A fixed-price contract transfers risk to the contractor, and a contractor absorbing losses has an incentive structure that is not simply “work harder”. It has an incentive to find the cheapest path to meeting the stated requirement, to contest what the requirement means, and to avoid discretionary work not strictly required.
None of that is wrongdoing. It is what the instrument is designed to produce. The question the reform never answered is whether the cheapest path to a stated requirement is the right way to build something where the failure modes are unknown in advance.
What the other provider did with the same instrument
The comparison is available inside the same program, which is unusual and makes it worth more than most.
NASA contracted two providers in 2014 under the same model, with the other receiving a smaller sum for the same service. That provider flew crew in 2020 and has flown routinely since.
The difference is not the contract. It is that one organization had been building and flying its own vehicles continuously, iterating in public and absorbing failures on uncrewed flights, while the other was a traditional prime contractor whose recent experience was in cost-plus programs where the customer carries the risk.
Put a fixed-price instrument into an organization built around cost-plus assumptions and the result is not automatically discipline. It can equally be an organization doing the same work under different accounting, discovering the difference only when the charges appear.
The verdict, such as it is
Fixed-price appears to work well where the thing being bought is understood: a launch to a known orbit, a resupply run, a satellite bus with heritage. The requirement can be stated precisely because everybody knows what it involves.
It works less well for development, where the point is that nobody yet knows what the work involves. A price fixed against an unknown scope is a bet by both parties about how much unknown there is, and when the bet is wrong the contractor is the one holding it.
The evidence is on both sides of one program. Two contractors, the same instrument, opposite outcomes, which suggests the model is not decisive on its own and that what it does is amplify whatever the organization already was.
What the second vehicle was for
Worth stating plainly, because it has been vindicated in a way nobody enjoys.
The reason NASA bought two vehicles was that a single provider means a single point of failure, and the agency had just spent nine years unable to launch its own astronauts after the shuttle retired. Redundancy was the entire justification for the extra expense.
That redundancy has run in one direction for four years, and the program’s current position depends on it continuing to.
A third party to the arrangement is worth naming: the crew. Two astronauts flew a vehicle whose first uncrewed attempt failed to reach the station in 2019 and whose second succeeded in 2022, and they flew it knowing the propulsion system had a known leak. That is a normal test-pilot judgment made under abnormal commercial pressure, and the contracting model is part of the pressure whether or not it entered anybody’s reasoning.
What is not resolved
Whether this vehicle comes home on schedule, and what the leaks and thrusters mean for the return.
The service module carrying the affected hardware is discarded before re-entry and burns up, which means the components that failed will never be examined. Whatever is learned about them has to be learned now, in orbit, through telemetry and testing, by people working against a docked vehicle they cannot take apart.
That is an unusual engineering position and it is a direct consequence of the design. A test flight exists to find problems, this one found some, and the part that would answer the questions is the part that does not come back.
