The Mirage of the Impenetrable Shield

Every dispute in this record begins with a definition. The US Army distinguishes a warhead kill, destruction or disabling of the incoming warhead, from a mission kill, which can mean merely diverting the missile. In 1991, an engagement was scored a success if the Scud caused no significant ground damage, even when the warhead landed intact in empty desert. The absence of damage was treated as evidence of interception. That definitional elasticity, not fraud in any single measurement, is how a near-zero warhead-kill record became a 96% success story delivered to the President of the United States.

This briefing uses the strictest available standard in each case: forensically verified warhead destruction where the data exists, independently compiled launch-versus-intercept datasets where it does not, and explicit flags wherever the numbers remain genuinely disputed.

On 15 February 1991, President George H.W. Bush stood at Raytheon’s Andover plant and declared the Patriot “41 for 42: 42 Scuds engaged, 41 intercepted.” The Army’s official figure then fell in stages: 96% in March 1991, 69% by May, 59% by April 1992, where it officially remains.

The GAO’s 1992 report carried its conclusion in its title: Data Does Not Exist to Conclusively Say How Well Patriot Performed. Outside the 9% of engagements with the strongest evidence, the Army could demonstrate only that “the Patriots came close to the Scuds, not that they destroyed them.” MIT’s Theodore Postol testified to Congress that the intercept rate “could be much lower than ten percent, possibly even zero.” The independent American Physical Society review panel later judged it “extremely unlikely” that the Postol-Lewis finding of very poor performance would ever be overturned.

The Israeli record was harsher still. Defense Minister Moshe Arens and IDF Chief of Staff Dan Shomron stated that Patriot intercepted at most one Scud over Israel; Shomron called its success “a myth.” Israeli analysis found that ground damage in Israel was greater after Patriot’s arrival than before: falling interceptors caused damage of their own.

The mechanics of failure

Iraq’s Al-Hussein, a Scud stretched for range, routinely broke apart on reentry, and the debris cloud confused Patriot’s radar into chasing harmless booster sections while warheads flew on. On 25 February 1991, a different failure mode killed 28 American soldiers at Dhahran: a 24-bit floating-point truncation in the range-gate software, compounded over 100+ hours of continuous operation into a 0.34-second timing error, shifted the tracking gate roughly 687 meters. The battery never engaged. The Scud struck the 14th Quartermaster Detachment barracks, the single most devastating attack on US forces of the entire war. The software patch arrived the next day.

When Houthi forces fired Iranian-derived Burqan-2H missiles at the Saudi capital, Riyadh claimed near-total interception. Jeffrey Lewis and colleagues at the Middlebury Institute’s Center for Nonproliferation Studies mapped every debris field and interceptor impact point for the November and December 2017 strikes. The pattern was unambiguous: the Burqan’s warhead separates from its airframe after boost. Saudi Patriots destroyed the large, radar-bright, harmless fuselage, while the warhead flew on, one landing within a few hundred meters of Terminal 5 at King Khalid International Airport.

For the seven-missile salvo of 25 March 2018, all claimed intercepted, Lewis found “no evidence that Saudi Arabia intercepted any missiles at all.” Video showed one interceptor executing a U-turn and detonating inside Riyadh. His verdict on the export of the system: the United States “seems to have sold them, and its own public, a lemon of a missile defense system.” Twenty-six years after Dhahran, the same discrimination failure, warhead versus debris, was producing the same false victories, now confirmed by open-source geolocation rather than classified telemetry.

Ukraine’s Patriot PAC-3 batteries opened with genuine triumphs: the first-ever combat interception of a Kh-47M2 Kinzhal in May 2023, and a reported 34-for-34 record against ballistic missiles targeting Kyiv that spring. A Ukrainian operator noted the “hypersonic” Kinzhal actually arrived at roughly Mach 3.6, a third of Moscow’s advertised speed. The hit-to-kill mechanics worked. Then Russia adapted.

The September 2025 collapse is the single most important data point in this record. Per the Financial Times and US Defense Intelligence Agency assessments, Russia upgraded the Iskander-M and Kinzhal with terminal-phase maneuvers, steep dives and lateral jinks that break the Patriot’s radar track in the final seconds. Interception fell from 37% to 6% in one month even as Russia launched fewer missiles. No hardware changed hands. The shield was defeated by a software update on the other side.

Russia paired adaptation with attrition: decoys that mimic the warhead’s radar and thermal signature, salvos timed to empty launcher canisters, and Iskander strikes, cued by electronic intelligence off the Patriot’s own radar emissions, that destroyed launchers and radars in place. A battery’s billion-dollar sensor is also its beacon. Russian industry now produces an estimated 840 to 1,020 Iskander/Kinzhal-class missiles per year; no Western interceptor line comes close.

The twelve-day Israel-Iran war of June 2025 was the largest ballistic missile defense engagement in history: roughly 550 to 574 Iranian ballistic missiles against the layered Arrow-2/Arrow-3, David’s Sling, THAAD, and ship-based SM-3 architecture. The headline number was genuinely impressive, approximately 86% intercepted overall, with 28 to 29 killed in Israel. The trend line underneath it was not.

Iran adapted inside a two-week war. Interception fell from 92% in the first week to 75% by the end as Tehran varied salvo timing, geometry, and missile mix; its best single strike put 10 of 27 missiles onto Israeli soil. Meanwhile the defenders burned interceptors faster than any production line can replace them: more than 150 THAAD rounds and roughly 80 SM-3s, the SM-3 expenditure alone exceeding the largest annual delivery of the Block IIA variant, of which the US buys about a dozen a year.

The economics of the exchange

This is the arithmetic that no radar upgrade fixes. The attacker adds warheads, decoys, and drones for six figures; the defender answers with eight-figure interceptors from years-long production queues. Saturation is not a tactic that sometimes works against missile defense. It is the condition under which missile defense always eventually operates.

Ground-based Midcourse Defense, the only US system defending the homeland against intercontinental missiles, has hit its target in roughly 55 to 57% of intercept tests, at a cumulative cost near $63 billion. Those tests are scripted: the defense knows the target’s timing, trajectory, and signature in advance, an advantage no real attack will grant. Real-world kill probability is unknown and presumed lower.

The deeper problem is physics, not engineering. In the vacuum of midcourse space there is no atmospheric drag, so a lightweight balloon decoy and a heavy nuclear warhead follow identical trajectories. Early GMD tests failed to distinguish mock warheads from simple balloons; the American Physical Society’s studies concluded the midcourse discrimination problem remains unsolved, and that a boost-phase space constellation against solid-fuel missiles would require on the order of 1,600 orbiting interceptors.

Into this record arrives Golden Dome, the space-based national shield announced in 2025 at a claimed $175 billion. The Congressional Budget Office’s May 2026 assessment priced it at $1.191 trillion over twenty years, nearly seven times the announced figure, with the 7,800-satellite interceptor layer alone consuming roughly $743 billion. The CBO’s own language: the system “would not be an impenetrable shield or be able to fully counter a large attack” by Russia or China. The one point on which contractors, critics, GAO, CBO, and the physics community agree is that the attacker’s next warhead is always cheaper than the defender’s next interceptor.

Section 219 of the House FY2027 National Defense Authorization Act establishes a United States-Israel Defense Technology Cooperation Initiative, directing the Secretary of Defense to appoint an executive agent to expand integration of the two nations’ defense-technology and industrial bases: missile defense, counter-drone, AI, cyber, autonomy, directed energy. The Massie-Khanna amendment to strip it was denied a floor vote by the Rules Committee; the provision has, to date, received no debate. Rep. Thomas Massie called it “an unprecedented escalation of foreign involvement in our military.” Senator Bernie Sanders said it would grant Israel “more military integration than any NATO ally.”

Here the two halves of this briefing meet. The June 2025 strikes on Iran were undertaken by leaderships that believed the missile shield could absorb the answer. It nearly could, at the cost of a quarter of the American THAAD arsenal, a rising leakage curve, and a demonstration to every adversary on Earth of exactly how long the magazine lasts. Confidence in the shield lowered the threshold for war; the war then drained the shield. That feedback loop, not any single intercept percentage, is the strategic finding of thirty-five years of data. A nation that binds its defense industrial base ever more tightly to a partner engaged in continuous regional war, without debate, without a vote on the merits, is compounding the same wager.

What would change this assessment

This briefing states its falsification conditions, because analysis that cannot be wrong is advocacy. The thesis fails if any of the following is demonstrated:

Until then, “impenetrable shield” remains a marketing claim, and every dollar and every policy built on it inherits the gap between the brochure and the battlefield.