The Primes Are Shopping for Factories

Northrop Grumman has found its manufacturing shortcut: a four-year-old startup that already knows how to build satellites fast.

The defense giant announced June 1 it is partnering with Apex, a Los Angeles-based satellite manufacturer, to develop space-based interceptors for the Trump administration's Golden Dome missile defense program. The pairing is the third major alliance in three weeks — Raytheon joined forces with Rocket Lab in early May, and Anduril assembled a team including Impulse Space, Inversion Space, and K2 Space later that month. The pattern is now unmistakable: the defense primes are not building these interceptors themselves. They are shopping for companies that already know how to manufacture at scale.

"We want to mature the tech to make sure this thing actually works," Apex CEO Ian Cinnamon told Air & Space Forces Magazine in October, when the company disclosed its own $15 million demonstration project, called Project Shadow. "Then, of course, as large primes and system integrators who know how to build these interceptors bid on it, we go to them and say, 'Listen, we're really good at the Orbital Magazine. We've already proven it in orbit. Let's collaborate.'"

That sales pitch is now a prime-time contract. Northrop Grumman confirmed it has completed key ground tests this year and is targeting an on-orbit demonstration in 2027. The company has invested $1 billion of its own capital in missile defense technology, though that figure — like most numbers in this program — comes with a significant asterisk.

The Congressional Budget Office released its own cost estimate last month: $1.2 trillion over 20 years for the full Golden Dome architecture, roughly seven times the Trump administration's $175 billion estimate. Space-based interceptors account for the bulk of that number. CBO concluded that defending against a large-scale ICBM attack would require a constellation of 7,800 interceptors operating in low Earth orbit — close enough to their targets to engage during the three-to-five-minute boost phase, but low enough that atmospheric drag degrades their orbits. At $22 million per satellite and a five-year service life, the replacement cycle alone would cost nearly $1 billion per year.

The administration has publicly disputed those figures. Gen. Michael Guetlein, the Pentagon official overseeing Golden Dome, has called CBO's estimate too high — though he has also spent months asking the same question the CBO answered: whether space-based interceptors can be built affordably and in sufficient quantities. The partnerships his office has structured suggest he is still looking for answers.

The contracting mechanism is designed to pressure industry into exactly that kind of urgency. The Space Force is using Other Transaction Authority agreements, which require companies to invest their own capital upfront in exchange for the possibility of larger production awards later. Northrop is putting in $1 billion of its own money. Rocket Lab's CFO described the model plainly during an earnings call: "Companies like ourselves and Raytheon and others that are in the mix have to put some of their own skin in the game to unlock a potentially very large opportunity in the back end."

The startup logic is not complicated. Apex opened a factory last year with the capacity to build up to 200 satellites per year — more than the entire U.S. military currently procures. Its Nova satellite bus, which would host the interceptor payload, is designed for rapid production. The company's Comet spacecraft can carry up to six interceptors per bus. At $6 million to $10 million per unit depending on configuration, the raw manufacturing cost is within a range that starts to make the 7,800-unit constellation math survivable — if the physics and the integration actually work.

None of which has been proven in orbit. Project Shadow, Apex's self-funded demonstration, is planned for 2026 and will not include live interceptors. It is a proof of the bus, not the kill vehicle. Northrop's 2027 on-orbit target is aggressive and will either validate or undermine the commercial manufacturing thesis within the next 18 months.

The 2028 deadline — Gen. Guetlein's stated goal for an initial layered capability — is the clock everyone is watching. If the commercial entrants can deliver at the cost and schedule their factories suggest, the program moves forward at something approaching the scale CBO modeled. If the demos slip, or if integration with prime systems proves harder than advertised, the architecture gets descoped — and the coverage gaps CBO identified become operational reality.

The primes need the startups. The startups need the primes. Whether they can move fast enough together is the only question that matters for the next 18 months.