Missile Defense

The US is adding interceptors to its Ground-Based Midcourse Defense site in Alaska. The Pentagon calls it a game-changer. Most of the analysts I trust call it incremental. Both can be true.

What GMD actually is

Ground-Based Midcourse Defense is a system designed to hit ICBM-class warheads in the middle phase of their flight — when they're coasting in space, after boost, before reentry. It uses long-range radars (Cobra Dane in the Aleutians, sea-based X-band, the upgraded early warning radars in Greenland and the UK), a command-and-control network, and ground-based interceptors at Fort Greely, Alaska and Vandenberg, California.

It was built by Boeing as the prime, with Raytheon as a major subcontractor on the kill vehicle. There are around 44 interceptors deployed today, with planned growth to about 64.

What it doesn't do: short-range theater missiles (that's THAAD and Patriot), cruise missiles (that's a totally different problem), or hypersonic glide vehicles in their boost-glide phase (that's still a research question). GMD does one job: ICBM-class threats from North Korea, primarily. Not Russia or China — they have too many warheads for it to matter.

If you're trying to read up on this stuff, the two books worth your money are Joe Cirincione's Nuclear Nightmares and Fred Kaplan's The Bomb. Both written for an educated general reader.

The honest performance picture

GMD intercept tests have run roughly 12 hits in 20 attempts since 1999 — under controlled conditions, with cooperative targets, known trajectories, no countermeasures. That's the optimistic number. In a real launch with decoys, chaff, maneuvering reentry, and a fog-of-war timeline, no serious analyst expects 60 percent. The honest internal estimate hovers somewhere between 25 and 50 percent per interceptor — which is why the doctrine is "shoot two interceptors at each warhead, hope at least one hits."

The 2019 SM-3 Block IIA failure didn't help. The new Next-Generation Interceptor program is meant to fix the kill-vehicle reliability issues. It's running over budget and behind schedule. Surprising no one.

For the policy-curious, a subscription to Foreign Affairs or The Bulletin of the Atomic Scientists is the cheap way to follow this debate without wading through Defense Department press releases.

What the new technology actually buys you

The headlines talk about AI, machine learning, hypersonic propulsion. The reality is narrower. AI is mostly being used in threat discrimination — sorting decoys from real warheads in the seconds you have before intercept. That's a genuine problem and a genuine improvement, but it's not a fundamental shift.

Hypersonic propulsion in the interceptor sense means faster reaction time, more time on target. It doesn't fix the basic geometry problem: hitting a bullet with a bullet at closing speeds north of 15,000 mph, where a one-meter miss is a miss. That's hard physics, not a software problem.

If you want a clear-eyed overview of the technology, the Congressional Research Service publishes free reports — search "CRS Missile Defense" and you'll find the latest. More reliable than anything in the trade press.

The cost question

GMD has consumed somewhere around $50 billion over its lifetime. That's real money. Critics argue the same dollars buy more security in better radar networks, hardened command systems, or non-kinetic responses (cyber, sanctions, deterrence). Supporters argue that even a 30 percent shot at stopping a North Korean warhead is worth any number of zeros.

Both sides are partially right. The mistake is talking about missile defense as a binary "works/doesn't" question. It's a probability shift, and the question is whether the probability shift is worth the marginal dollar.

If geopolitics is your bag, Tim Marshall's Prisoners of Geography is the friendliest entry point to why Alaska matters in this picture. Spoiler: it's the great circle route from North Korea to North America.

Where this goes

The Pentagon will keep expanding GMD because the political cost of not doing it is high and the technical cost of doing it has been amortized. Whether it ever works as advertised against a real threat is an open question, and the honest answer from inside the program is "we don't know and we hope we never find out."

That's not a great answer. It's the only one available.