Washington's Hedge Has to Prove Itself on Amazon's Flights

Washington’s insurance policy against SpaceX has to prove itself on Amazon’s satellites before it gets near the Pentagon’s.

Two Ars Technica stories landed within a day of each other, and they describe the same shift from opposite ends. Eric Berger argues that SpaceX has gone from needing the US government as an indispensable customer to being the government’s indispensable contractor. He writes that SpaceX now seems increasingly uninterested in most government contracts outside data services such as Starshield. Stephen Clark reports that ULA’s Vulcan returns to flight on November 13, nine months after a solid rocket booster anomaly, carrying 40 satellites for Amazon Leo. Put them side by side and the picture is clear. The buyer has lost its leverage, and the hedge it built has to earn back trust on someone else’s manifest.

Start with how the buyer lost. For most of SpaceX’s history the US government was the market. Berger describes Elon Musk and Gwynne Shotwell treating Washington as El Dorado, cutting through a jungle of Boeing, Lockheed Martin, bean counters and regulators. They sued. They complained about tilted playing fields. Most of all, they delivered quality space products for less. So the government did what a sensible buyer does. It rewarded the better, cheaper supplier, again and again. That is how a monopsony ends. The buyer picks the winner, the winner compounds, and one day the buyer finds out it can’t leave. Chris Davenport, an Arizona State professor and author of Rocket Dreams, put it plainly to Berger: “NASA desperately needs SpaceX, and SpaceX knows it, and Elon knows it.” Lori Garver, NASA’s deputy administrator under Obama, is blunter about the mood: “I guess I am still in denial that there is a divorce pending.”

The Pentagon saw this coming. That’s why it split the work. According to Ars, the Pentagon has given ULA’s Vulcan roughly half of its major launches through the early 2030s. On paper, that’s assured access to space through two providers. In practice, a stated policy is only as real as the second rocket. Vulcan had nozzle failures on its solid rocket boosters on two of its first four missions, most recently in February. Both times the rocket recovered and finished the mission. The Space Force certified Vulcan in 2025 and flew two missions on it. After February, officials lost confidence and the rocket was grounded. The hedge has spent most of 2026 on the ground.

So how does the hedge get back in the air for the Pentagon? Read the manifest. The return-to-flight mission doesn’t carry a Space Force payload. It carries 40 Amazon Leo satellites, the first of 38 Vulcan flights Amazon has booked. And the military is waiting on those flights. Brig. Gen. Eric Zarybnisky, the Space Force’s chief rocket buyer, says the military wants data from commercial Vulcan flights before clearing ULA to resume national security launches. At least two Amazon Leo missions are scheduled before Vulcan flies another national security mission. Ars still calls the Space Force ULA’s primary customer, and that hasn’t changed. The sequence has, though. The way back for Washington’s protection against one company’s rockets runs through another company’s broadband constellation. No procurement document describes it that way. The flight schedule does.

Here’s the strongest case against me. Washington isn’t helpless. It still pays for Starshield, and SpaceX clearly still wants that business. It controls the range: even Vulcan’s November date is pending range approval. It licenses launches. It can make life slow and expensive for anyone it chooses. All true. But look at what kind of leverage those are. They’re permission levers, not purchase levers. A buyer with purchasing power can walk away and make the seller chase it. A buyer left with only permission power can delay things, and delaying an indispensable contractor hurts the buyer first. Starshield doesn’t weaken the argument either. A contractor that stays keen on recurring data services and stops signing many large new spaceflight contracts is behaving exactly like one that no longer needs the customer.

None of this is a disaster. It’s what a working market looks like when technology compounds faster than procurement does. A commercial customer flying real payloads on Vulcan gives the Pentagon something better than a test campaign: flight data at no risk to a national security satellite. The real failure would be pretending the two-provider policy is self-executing. A policy can assign half the launches. It can’t make the rocket fly.

The stakes are concrete. If Vulcan’s Amazon flights come back clean, the Pentagon gets the data it says it needs and the split holds. If they don’t, the US government moves closer to being a price-taker for access to space, buying from a contractor that has stopped needing it.

A second source isn’t a policy. It’s a business, and somebody has to keep it alive.