NASA paid SpaceX only on flown milestones and got two new rockets for $800 million

In the mid-2000s NASA faced an uncomfortable arithmetic: the shuttle was retiring and nothing else could resupply the International Space Station, while developing a replacement in-house the traditional way would eat the budget that exploration needed. An early attempt to hand Kistler Aerospace $227 million without competition collapsed after Elon Musk protested and the GAO prepared to rule against NASA — which pushed the agency into holding an open competition instead. That accident of process gave the agency the chance to design the procurement from scratch.

The design was unlike NASA's usual contracts. In August 2006 the Phase I winners were SpaceX, with a $278 million funded Space Act Agreement, and Rocketplane Kistler with $207 million — payable only against demonstrated milestones. When Kistler kept missing them, NASA terminated in October 2007, having paid out just $32.1 million; Orbital Sciences took the vacated slot in February 2008. The structure had teeth: augmentation funding raised SpaceX's total to $396 million and Orbital's to $288 million, and both companies kept ownership of what they built.

The bet paid off on a schedule cost-plus contracting rarely matches. In December 2008 NASA bought the service outright: $1.6 billion for at least 12 SpaceX Dragon flights and $1.9 billion for 8 Orbital flights. The first contracted cargo mission flew in October 2012, closing in November 2013 with two new launch vehicles and two automated cargo spacecraft for roughly $800 million — against the $12 billion cost-plus Orion contract. A NASA cost study put Falcon 9 launches at $2,720 per kilogram against the shuttle's $54,500 — a twentyfold reduction.

Pay only on flown milestones; fixed price, not cost-plus

Fixed ceilings and milestone payments put overrun risk on the builder, so the incentive to be cheap and fast finally sat with the party that controls the cost.

Payment followed demonstration, not effort: NASA bought outcomes — a launch, a berthing — so process failures cost the contractor, not the taxpayer.

Terminating Rocketplane Kistler for missed milestones proved the structure had teeth; under cost-plus, that money would simply have been sunk.

Because the companies kept their vehicles, SpaceX could sell Falcon 9 commercially and private demand helped carry the development bill.

Procurement design decides who carries risk, and that decides what things cost. Pay for demonstrated outcomes, not reimbursed effort, and whoever controls cost finally has a reason to control it.

The model became NASA's default for getting out of the launch business: fixed-price service contracts spread to Commercial Crew, where NASA later estimated the approach saved tens of billions versus traditional development, and cargo missions became routine. The lesson came with an asterisk: under fixed price, profits are earned and losses are real, and in later crew and cargo rounds several companies declined to bid or took heavy losses. The asymmetry that made COTS cheap for the government is exactly what makes it demanding for everyone else.