Sentinel ICBM: how an 81 percent cost jump tripped the ...

In January 2024 the U.S. Air Force told Congress that its new intercontinental ballistic missile program had blown through a legal cost limit that is supposed to force a program to be cancelled. Six months later the Pentagon estimated the cost had grown roughly 81 percent, from about $95.8 billion to about $140.9 billion. Then it decided to keep going anyway.

That is the whole story of the LGM-35A Sentinel in one paragraph, and almost every part of it is easy to get wrong. The overrun was not driven by the missile. The $141 billion is not money already spent. And the program did not get cancelled despite tripping a "critical" breach, because the same law that threatens cancellation also gives the Secretary of Defense a certified exit. This post walks through what Sentinel is, what happened, what the numbers actually mean, and the honest case for and against continuing.

What Sentinel is

Sentinel, formerly called the Ground Based Strategic Deterrent (GBSD) and renamed in April 2022, is the Air Force and Northrop Grumman program to replace the LGM-30 Minuteman III and modernize the land leg of the U.S. nuclear triad. Northrop Grumman is the prime contractor. The Congressional Research Service documents the LGM-35A identity, the Northrop prime role, and the Minuteman III replacement mission in its defense primer on the program.

The thing most people picture when they hear "new ICBM" is the missile. Sentinel is much more than that. It is the missile plus the vast ground infrastructure that fields it: silos, launch control centers, and the cabling that ties them together across the northern plains. The system it replaces spans more than 600 facilities across five states, according to the Government Accountability Office. Commonly cited planning figures put that at roughly 450 silos and launch facilities spread over about 40,000 square miles in Wyoming, Montana, North Dakota, Nebraska, and Colorado. GAO calls Sentinel "the most significant and complex infrastructure project in the [Air Force's] history." That framing is not decoration. It is the key to understanding where the money went.

The Minuteman III entered service in the early 1970s. It is decades past its intended service life. The land leg it anchors is a fleet of dispersed, hardened, always-alert missiles whose purpose is to make a disarming first strike against the United States impossible to plan. That mission is real, and it is the reason this program exists.

What happened, in dates

The timeline matters because two different cost figures come from two different moments, and treating them as a contradiction is the easiest mistake to make.

September 2020: Sentinel passes Milestone B, the formal start of engineering and manufacturing development. It is the first new U.S. ICBM effort in about 50 years. Officials would later concede the program was not at full design maturity when it cleared this gate.
December 2023: Reported per-missile unit cost has risen from about $118 million (2020) to roughly $162 million, per compiled Selected Acquisition Report figures. Treat these unit numbers as illustrative rather than precise. Note that this is the missile's unit cost, even though the missile is not what drove the breach.
January 18, 2024: The Air Force notifies Congress that Sentinel has incurred a critical Nunn-McCurdy breach. The initial notification, as reported by the Arms Control Association, cited at least a 37 percent overrun, from a roughly $95.3 billion baseline toward more than $125 billion. That triggers a mandatory review.
July 8, 2024: The completed Nunn-McCurdy review lands. Under Secretary of Defense for Acquisition and Sustainment William LaPlante certifies Sentinel as essential to national security. The Pentagon rescinds the September 2020 Milestone B approval and orders a restructuring with a new baseline.
February 18, 2026: GAO's report on the program puts Sentinel at "at least $141 billion," with the first flight test slipped to March 2028 and Minuteman III potentially kept operational until 2050.

The money, kept straight

Here is the part that rewards careful reading.

The two percentages are two snapshots, not a dispute. The January 2024 notification cited about 37 percent over a roughly $95.3 billion baseline. That was the early, partial picture that triggered the review. The fuller figure came out of the completed July 2024 review: an increase of roughly 81 percent, from about $95.8 billion to about $140.9 billion, widely rounded to $141 billion. That is an increase of roughly $45.3 billion. The 81 percent is measured against the roughly $95.8 billion Milestone B estimate. Both numbers are correct; they just describe different moments in the same review.

These are estimates, not cash out the door. The $140.9 billion is a base-year program-cost estimate, not money already spent or obligated. Presenting $141 billion as "already spent" is the single most common error in casual coverage. GAO's February 2026 report uses the "at least $141 billion" language precisely because it is a forward-looking program estimate that could still move.

What a "critical" breach means. The Nunn-McCurdy law sets tripwires on unit-cost growth. A "critical" breach is roughly 25 percent or more over the current baseline, or 50 percent or more over the original. Sentinel cleared that bar decisively. By statute, a critical breach requires the program to be terminated unless the Secretary of Defense certifies it against four criteria, including that it is essential to national security and that there is no alternative of lower cost. That certification is the accountability lever, and it is exactly what fired here.

The overrun is on the ground, not in the air. This is the load-bearing, counterintuitive point. The Defense Department attributed the majority of the cost growth to the command-and-launch segment: the launch facilities, the launch control centers, and the enormous job of converting Minuteman III sites to Sentinel, not the missile itself. As reporting quoting LaPlante and other officials laid out, the review found that existing silos need new construction rather than the reuse the program had assumed. Rebuilding hardened underground infrastructure and its cabling across five states, at scale, turned out to be far harder and costlier than the 2020 estimate assumed. Officials acknowledged the program was not at the design maturity it should have had when it passed Milestone B. If you take away one fact from this article, make it this one: the silos, not the rockets, drove the breach.

The schedule slipped, and kept slipping. Early 2024 estimates put the delay around two years. The July 2024 review projected at least three years. By GAO's February 2026 report, the first flight test had slipped about four years, to March 2028. The knock-on effect is that the Air Force may have to keep the 1970s-era Minuteman III operational until about 2050, roughly 14 years longer than planned, layering sustainment cost for the old system on top of the cost of the new one.

What continuation actually looked like

It would be wrong to describe the July 2024 decision as a shrug. The Nunn-McCurdy review examined roughly four to five options, including extending Minuteman III toward as late as around 2070. It found none of them cheaper or adequate to the requirement. On that basis LaPlante certified the program essential to national security, with no alternative able to meet the requirement at lower cost, per the Pentagon's announcement of the review results.

Continuation also came with teeth. The Pentagon rescinded Sentinel's Milestone B approval, the formal marker that development was properly under way, and directed a restructuring with a new cost and schedule baseline, a process expected to take 18 to 24 months to re-approve. In acquisition terms, that is not nothing. The program survived, but it was sent back to redo the homework that the 2020 milestone had certified as complete.

The honest critique and the honest defense

The critique. By GAO's own account, this is a documented cost and schedule failure. A cost estimate that grows about 81 percent, from roughly $95.8 billion to $140.9 billion, and a first flight that slips about four years, is not a rounding error. The most damning detail is that the biggest driver, the ground infrastructure, was scope the department admitted it did not fully understand when it passed Milestone B in 2020. The program committed to a baseline before it had done the design maturity work that would have exposed the true cost of rebuilding 450 silos and their launch centers. GAO continues to flag software and schedule risk going forward. And the delay is not free: keeping Minuteman III credible until around 2050 means paying to sustain a 1970s system for another generation while its replacement catches up.

The defense. The mission is genuine, and the review took it seriously rather than rubber-stamping it. Minuteman III is decades past its designed life, and the land leg it anchors, hundreds of dispersed, hardened, always-alert missiles, is a deterrence function that an aging system cannot indefinitely guarantee. The Nunn-McCurdy process is precisely the mechanism meant to test whether spending more is justified, and here it did examine alternatives, including extending the legacy missile, and formally concluded none met the requirement at lower cost. The breach also produced real accountability mechanics: Milestone B was rescinded and the program was ordered restructured against a new baseline. On the defenders' account, the system caught the problem, forced a re-examination, and continued only after certifying against statutory criteria that there was no cheaper way to keep the land leg viable.

Both verdicts are true at once. Sentinel is a real, GAO-documented cost and schedule breach, and it is a recapitalization the government formally judged to have no cheaper adequate alternative. The interesting question is not which of those is the "real" story. It is whether the oversight machinery that caught the breach, and the design discipline that should have caught it three years earlier, are the same institution learning at two different speeds.

Fact-check notes and sources

Program identity (LGM-35A, formerly GBSD, renamed April 2022), Northrop Grumman as prime, and the Minuteman III replacement mission: Congressional Research Service, Defense Primer IF11681.
"More than 600 facilities across five states" and "the most significant and complex infrastructure project in the [Air Force's] history," plus the "at least $141 billion" figure, the roughly four-year slip to a March 2028 first flight, and Minuteman III potentially running to 2050: GAO-26-108755 (Feb. 18, 2026). The 450-silo and 40,000-square-mile figures are commonly cited planning numbers drawn from CRS and other secondary sources.
January 18, 2024 critical Nunn-McCurdy breach notification, the initial roughly 37 percent overrun over a roughly $95.3 billion baseline, and the mechanics of a "critical" breach: Arms Control Association.
The roughly 81 percent growth from about $95.8 billion to about $140.9 billion (near $141 billion) versus the original forecast, from the July 2024 review: Stars and Stripes (July 8, 2024).
The ground command-and-launch segment (launch facilities, control centers, Minuteman III to Sentinel conversion) as the majority driver of cost growth, the finding that existing silos need new construction rather than reuse, the July 8, 2024 certification by USD(A&S) William LaPlante, and the review of roughly four to five alternatives: Breaking Defense (July 2024).
The official Defense Department announcement of the Nunn-McCurdy review results and the decision to continue: Department of Defense release. This release now lives on the war.gov domain following the department's renaming; the URL is authoritative even though it returns a 403 to automated fetch tools.
Per-missile unit cost rising from about $118 million (2020) to roughly $162 million (December 2023): compiled from DoD Selected Acquisition Report figures via secondary reporting. Treat these unit figures as illustrative; they are the missile's unit cost, not the ground segment that drove the breach.

Sentinel ICBM: how an 81 percent cost jump tripped the Nunn-McCurdy limit and the program survived anyway