When Iran began striking Naval Support Activity Bahrain on February 28, 5th Fleet headquarters evacuated personnel into Manama hotels that were struck as well. Fleet logistics displaced involuntarily to Diego Garcia. Navy Secretary Hung Cao described the result in September without softening it: "They blew the hell out of Bahrain". The New York Times reported that many of the 13 regional bases American troops use have become all but uninhabitable, and to add to that damage, an Iranian missile destroyed a $500 million E-3 command-and-control aircraft parked on the ramp at Prince Sultan Air Base. Because the force anchoring American naval power in the Gulf never believed anything would arrive overhead, it had limited hardened shelters, did not execute its dispersal plan, and lacked defense deep enough to counter a UAS threat from the air.
That belief in American dominance of the skies has a well-defined construction record, with nearly every expeditionary base the United States built after Desert Storm using it: HESCO barriers, T-walls, and containerized housing, armor against fragments and small arms, nothing against penetration from above. The FOBs I deployed to in Afghanistan and Iraq followed that pattern exactly, because we controlled the air above us. The T-walls gave us adequate protection from inbound Taliban mortars. American engineers have poured that assumption into gravel and wire for thirty years. A former CENTCOM communications director confirmed most US shelters in the Middle East region are T-wall constructions from the Iraq invasion era that no one ever seriously fortified. The Air Force's own infrastructure strategy now concedes its facilities "can no longer be considered a sanctuary". American bases stayed safe from air attack for thirty years because the weapons able to reach them were either expensive, scarce, or visible coming from a long way out - the advent of cheap autonomous drones has removed all three conditions at once.
Our adversaries spent decades working underground - digging is what you do when the other side owns the sky. Kuribayashi honeycombed Iwo Jima's rock with miles of tunnels, weeks of preparatory bombardment barely touched the garrison, and the Marines paid the price with 26,000 casualties across five weeks digging it out. Planners predicted the Umurbrogol ridges on Peleliu would fall in days, and the bloody fight ran more than two months. Viet Cong ran logistics, hospitals, and staging through the Cu Chi network and attacked into Saigon from beneath total American air dominance above. In Afghanistan, al-Qaeda's leadership escaped Tora Bora through its cave networks under weeks of B-52 strikes in 2001. Sixteen years later the US dropped the MOAB, its largest conventional bomb, on a single ISIS-K tunnel complex in Achin district. Across twenty years, the United States has employed that weapon against tunnels exactly once, from a cargo plane in uncontested airspace, and the district still fought on for years afterward. Hezbollah, North Korea, and Iran's missile cities continue the practice today. The Marine Corps still teaches the historical lessons from Iwo Jima and Peleliu, but has never dug tunnels of its own, because no enemy aircraft have truly attacked Marines on the ground since Korea.
Striking a dug-in enemy today shows the hefty price tag as well. When the United States hit Iran's enrichment site at Fordow in June 2025, the facility's depth forced the mission onto B-2 bombers dropping 30,000-pound Massive Ordnance Penetrators, the only aircraft and munition in the world that could attempt it, and the strike expended fourteen MOPs across Fordow and Natanz. The battle-damage debate over destroyed versus severely damaged was contentious, and B-2s struck Iranian underground missile facilities again in this year's war. Boeing's contract to replace those bombs does not begin delivery until January 2028. Twenty years of employing the heaviest conventional weapons ever built against underground networks has produced suppression and denial but rarely has resulted in complete destruction. Nine months after Iran's buried site demanded the most specialized weapon in the American inventory, Iran struck America's unburied facilities in Bahrain with Shahed-136s that cost a few tens of thousands of dollars each.
The United States built its air defenses to intercept small numbers of expensive weapons; large numbers of cheap ones quickly can exhaust them. Ukraine demonstrated this against Moscow in June, mapping the seams in the densest defended airspace on earth and sending enough drones that the batteries ran out of intercepts, while debris from the missiles' intercepts fell across the city's own neighborhoods. Both militaries are using ground and air drones across the battlefield laboratory that is Russia and Ukraine, and both adapt their penetration tactics and defenses weekly. In every aerial raid, some number of aircraft reach their targets no matter how well the defenders shoot.
While the exact numbers are hotly contested, the truth is that the Iran war has consumed significant quantities of American defense interceptors - CSIS analysts estimate the United States has spent roughly two-thirds of its pre-war Patriot inventory, leaving fewer than 1,000 interceptors worldwide and about 250 THAADs, with three or more years needed to rebuild either. Iran and its partners can rebuild their drone and missile inventories every month.
The Pentagon's answer to this problem so far leans heavily toward offense. The Army alone plans to buy roughly one million small drones over the next two to three years. Whether that mass fits any American war plan is the true question, since FPV drones dominate trench lines that sit a few kilometers apart in Ukraine. However, there are few Pacific scenarios looking anything like that. Iran forced the evacuation of Fifth Fleet headquarters with a few hundred Shaheds, not flying FPVs across the Persian Gulf.
China has moved on this earlier than the US as well. The PLA built more than 400 hardened aircraft shelters in the last decade, roughly 800 total, plus 20 new runways and 49 new taxiways within a thousand nautical miles of Taiwan, while the United States built 22 shelters across the entire Indo-Pacific; Guam has none, with bombers rotating through Andersen parked in open rows. Not everyone agrees hardening is the answer. Gen. Wilsbach said in 2023 he was not a big fan of hardening infrastructure, and Marine leadership in the Pacific favors dispersal over concrete, arguing that static defenses anchor a force that needs to maneuver. Dispersal and hardening work as layers rather than rivals - agile combat employment moves what flies while shelters protect what parks - and the PLA, which faces this problem across the same water, can do both at once.
Hardening has its own distinct limits. Fuel, munitions, command and control, and maintenance for high-value aircraft can go underground where geology allows. A hardened shelter costs roughly $30 million and protects a $600 million B-21. A well-placed warhead can wreck an entire ramp of unsheltered aircraft, while sheltered aircraft force the adversary to spend a precision weapon on each one. RAND reached the same recommendations this year: blast-resistant shelters, rapid runway repair, and a serious cost-benefit look at the Air Force fielding its own air-base defense interceptors rather than relying on Army units too few in number to cover a dispersed posture. Shelters defeat cluster submunitions and near misses, but not direct hits from penetrators. The soft underbelly of aviation remains the runway, which cannot go underground - an exposed strip that provides the exact argument for Shield AI's VTOL X-Bat and other distributed-launch concepts. Shelters also do nothing against attacks launched from inside the perimeter - Ukraine's Operation Spiderweb launched 117 drones from trucks parked inside Russia and destroyed roughly a third of the strategic missile-carrier fleet on open ramps. A base without the sensors to detect a launch from a truck inside its own fence line will have little to gain from the concrete over its aircraft. The shelter program and the detection layer must be seamless architecture for true protection. The US currently has no way to defend against this type of domestic attack.
The offensive mission of finding and clearing enemy tunnels belonged to Marines with flashlights at Iwo Jima and tunnel rats at Cu Chi, and it is now a drone mission. DARPA's Subterranean Challenge spent three years pushing autonomous robots through tunnel systems, caves, and urban undergrounds with no GPS and no communications link, and quads like Shield AI's former Nova 2 proved they could map interiors entirely on their own. The same capability can map a cave complex without sending a human in first. The defensive implication runs the other way: any facility the US digs could potentially face the same mapping robots - a buried command post still needs its entrances sensed and defended against systems small enough to fly through a door. Digging deep can buy protection from what falls from above, but the detection architecture has to watch more than just the fence line too.
DoW is learning these lessons, with CENTCOM and the services issuing solicitations for prefabricated bunkers deliverable within 30 days after getting hit, hardened underground facilities at Al Udeid, and counter-drone netting at stateside flightlines. Seven months into the Bahrain campaign, on September 30, SecWar Hegseth signed the memo directing site selection for the Next Great American Base, with founding language committing to a highly hardened joint installation in the homeland, the first new domestic base since the 1980s. During the same speech, he announced the Autonomous Warfare Command that will own much of the acquisition behind future c-UxS base defense. While the FY27 request puts $75 billion toward drone technology, there are no funded programs to fund hardened shelters on Guam, despite facing roughly two thousand Chinese ballistic missiles.
Congress can fund hardened shelters at Andersen and the priority Pacific airfields through the same MILCON process now building the Next Great American Base. Bases domestic and overseas need sensors covering launches from inside their own perimeter. Spiderweb's trucks parked legally inside Russia before launch. The FY27 conference is deciding the drone accounts right now, and hardened shelters, base sensors, and penetrator production can ride in the same bill.