Ten minutes. Forty satellites. Twenty years without a total loss, until Sunday. That is the Vulcan launch Ars Technica's Stephen Clark reported: United Launch Alliance's rocket failed partway through its first Amazon Leo mission and took 40 broadband satellites with it. It was ULA's first complete mission failure since Boeing and Lockheed Martin merged their rocket divisions in 2006. Vulcan will be grounded indefinitely, and it is on contract for roughly half of the U.S. military's national security launches over the next few years.
The obvious read is that the Pentagon now leans harder on SpaceX. True, and Ars says so. The more useful read is about how the government counts its backups. Federal law sets the floor at "at least two space launch vehicles (or families of space launch vehicles)" able to carry any national security payload. The Space Force went past that floor in 2025 and added Blue Origin's New Glenn as a third option. That makes three rockets. Count engines instead and there are two, because Vulcan's first stage and New Glenn's both burn Blue Origin's BE-4.
Sunday's telemetry points at the stage those engines push. Ars's sources put the anomaly in Vulcan's booster stage, not the Centaur upper stage or the solid boosters, and the twin BE-4s fired at least 30 seconds past their planned cutoff near the five-minute mark, which Ars notes can mean an underperforming stage was trying to compensate. Nobody yet knows whether the engines, the plumbing that feeds them or something else failed. But this isn't the first time the question has come up. In May a New Glenn blew up on its pad during a ground test, and investigators traced it to a faulty valve on one of its BE-4s. Ars spells out where that leaves Blue Origin:
If engineers determine a BE-4 was at fault on Sunday's Vulcan failure, it could stand in the way of Blue Origin returning New Glenn to flight early next year.Ars Technica
That sentence is what a vendor count hides. A backup is worth having only if the first option's failure can't reach it. Here, one rocket's failure review now sets the other's return-to-flight date.
Two airframes on one engine guard against a bad fairing or a bad upper stage. They do nothing about the engine.
The other engine family belongs to rockets SpaceX is retiring. Falcon 9 and Falcon Heavy are on their way out as SpaceX moves to Starship, and SpaceX hasn't said it will certify Starship for national security missions, a process Ars puts at multiple years. Earlier Vulcan delays had already moved six national security launches to SpaceX. Expect more swaps, onto vehicles with a shrinking runway.
The case for a shared engine is real, and it deserves a straight hearing. A large methane engine is the slowest, most expensive part of a new rocket to build. Two vehicles buying the same one give Blue Origin the production volume to justify the line, and every flight on either one adds data for both. Vulcan chose the BE-4 in the first place to replace the Russian RD-180 that Congress wanted off national security launches, which was a real resilience gain. And Sunday's cause may still turn out to be a feed line rather than an engine.
Grant all of that. It still doesn't make the two rockets independent. Shared volume and shared data are why commonality is cheap, and they are also how a defect travels. Before either rocket counts as a hedge against the other, someone has to answer whether the BE-4 was at fault. Even if the answer is no, the waiting is the coupling. The Space Force has three names on its list and, for the next few months, one family it can fly with confidence, which is the one being phased out.
The fix isn't to drop Vulcan or New Glenn. It's to change the unit. Assured access should be measured the way failures actually happen, by the components a fault can reach, and that means asking which engines, valves and suppliers the backups share. The law asks for two families of rockets. It should also ask what those families have in common.
Assured access is counted in rockets. It fails in parts.