The Navy Ran a $420 Million a Year Program Whose Only Deliverable Was Waiting

The Navy spent an estimated $4.2 billion over ten years supporting attack submarines that were delayed in maintenance or sitting idle while waiting for shipyard capacity. That figure comes from an August 27 Government Accountability Office report, as reported by WorkBoat. Averaged across the decade, it works out to $420 million a year, every year, for ten years. I read it differently: the Navy has been running a ten-year, $420-million-a-year readiness program whose only deliverable was waiting time. It bought a queue, paid for it annually in cash, and got no availability back for the money.

A Queue Is Not A Calendar Problem

When shipyard slots and skilled trades are the scarce resources, hulls form a queue, and a nuclear attack submarine is an expensive thing to hold in one. The crew still draws pay. The pier services still run, the engineering and logistics tail still bills hours, and the material condition of the boat still demands attention whether or not an availability ever starts. Hull-level support costs accrue continuously while availability does not accrue at all. That asymmetry is the whole mechanism: it converts schedule variance, which sounds like a planning nuisance, into recurring cash burn, which is a budget line.

The trap is in the framing. A slipped availability gets briefed as a calendar problem: the boat enters the yard late, the plan shifts right, the chart gets updated, and everyone moves on. But calendars do not consume money and queues do. Every month a hull waits, the meter runs, and because the cost is spread thin across accounts, nobody is forced to treat the meter as a decision variable. Somewhere a comptroller sees pieces of it, but nobody sees the queue itself as a purchase, and the waiting became normal because no one ever had to buy it deliberately.

And the queue never sleeps.

What $420 Million A Year Would Have Bought

Put the number next to the fix: the Navy is now investing close to $1 billion to improve how its submarines are built, because delays and worker shortages continue to slow production. About $900 million of that goes into automated manufacturing facilities that can produce submarine parts faster. The broader effort uses automated factories, artificial intelligence and supply-chain support to increase Virginia-class and Columbia-class submarine output. The queue cost an estimated $4.2 billion over the decade. The Navy is now making a construction-improvement investment of close to $1 billion.

I am not arguing the $1 billion is misdirected; aiming money at throughput is exactly the right instinct, and it is overdue. I am arguing about the ten years in which the queue cost was treated as background noise rather than as capital that could have justified exactly this kind of investment much earlier. Any factory manager who watched $420 million a year evaporate into work-in-process would be asked for a throughput plan within the week. The difference is that a factory measures its queues in dollars, and a fleet has mostly measured its queues in months.

The Same Economics On The Build Side

The construction side shows the identical shape. The Navy expects its new 20,810-ton nuclear-powered submarine to be delivered by 2028. The boat is around 70 percent complete and will replace the Ohio-class submarines. Construction complications revised the delivery date, and the program has also been affected by issues in the timely supply of crucial parts and by delays in assembling major sections of the vessel. A bottleneck somewhere in parts or trades becomes schedule variance, and schedule variance becomes money, whether the hull is waiting for a maintenance slot or waiting for a module.

Notice what the public reporting does not contain. It does not name the binding constraint, state the throughput of that constraint, or price what one month of slip costs at the hull level. Those numbers exist somewhere inside the program; someone is living with them daily. They are simply not the numbers the plan is argued with in public, and what a plan is argued with is what gets managed. Everything else gets absorbed, and absorbed costs are the ones that quietly compound.

What Pricing A Bottleneck Looks Like

The commercial industrial base offers a contrast. GE Aerospace made a $12 billion purchase to contain shortages of precision jet-engine parts and to secure valuable technology by breaking into the inner circle of castings manufacturers. The CPP deal targets a jet-engine supply bottleneck. Agree or disagree with the price, the logic is legible: identify the constraint, estimate what it costs you, and pay to control it. The bottleneck was treated as an asset with a market value, not as overhead to be endured.

Northrop Grumman is doing similar arithmetic in the open. Its 57,000-square-foot Propulsion Innovation Center will be operated by approximately 250 engineers building advanced systems for U.S. and allied defense programs. The site is part of a broader $100 million investment at the company's existing 550-acre Elkton campus. That investment will increase local rocket-motor design and production capacity by 25 percent and enlarge the workforce by over 30 percent. Capacity, headcount, percentage change: those are the units of an organization that treats production capacity as the product it is selling, not as an expense to be minimized.

Capital spending alone proves nothing, though. A $533 million Army ammunition factory intended to boost artillery production failed to produce a single usable shell. The factory, built to help replenish 155 mm shell supplies, was instead plagued by robot fires, malfunctioning equipment and production failures. Buying a building is not buying throughput, and automation is not a throughput plan until someone writes down what the machines must produce, at what rate, and how that rate will be verified; the same holds for artificial intelligence, which appears in the Navy's submarine investment as a general ingredient rather than a task with a specification. Whether the $900 million becomes an Elkton or becomes that ammunition plant depends on that discipline.

Price The Queue In Every Plan

Here is what I would change. Stop treating maintenance capacity as overhead, and require every force-structure plan to carry three priced quantities before it is approved: expected queue time per hull with its annual carrying cost, throughput of the constraining facility, and the identity and headcount of the bottleneck trade. Then tag every one of those figures with its maturity, so the reader knows whether it is modeled, specified, calculated or objective. A tagged number invites scrutiny at the level it deserves. An untagged number invites nothing, which is precisely how $4.2 billion accrues over a decade without a single deliberate decision to spend it.

This is not hypothetical discipline. At Kibernan we have produced six complete, costed engineering programs: AI Managed Missile Factory, AEGIS-ARGUS-KRONOS, IRON HIVE, DRONE WALL, FANGS and VES. Every published figure in those programs carries one of those four maturity tags, and they are proposals; no Kibernan hardware has been built, flown or fielded, and we say so in writing, because a number without its provenance is a way of hiding what you do not yet know. If a small firm can hold every figure across six programs to that standard, a force-structure plan can certainly hold three.

The $4.2 billion is not a scandal. It is a price, and prices are useful when you can see them before you pay them. The scandal would be paying it again for another decade because queue time never appeared in any plan as a cost to be traded against capacity. Waiting is a program. The only choice is whether you fund it deliberately.