The Gulf Already Won the Detection War. It Is Losing the Allocation War.

What 2026 taught the Gulf about air defense, and why the decision layer, not the sensor or the interceptor, is now the binding constraint.

Adamant · Riyadh

---

The number everyone quotes is the wrong number

On 23 June 2025, Qatar's Ministry of Defence reported that 18 of 19 Iranian ballistic missiles fired at Al Udeid Air Base were intercepted (Xinhua). Ninety-five percent. By any historical standard, an extraordinary result.

Eight months later, that standard stopped mattering.

Between 28 February and mid-March 2026, Iran struck all six GCC states (Reuters). As of 18 March, the UAE had been targeted by 342 missiles and 1,699 drones, Bahrain by 132 and 234, and Qatar by 203 and 87 since 28 February; Kuwait's cumulative count stood at 237 and 445. Saudi Arabia saw interceptions over Prince Sultan Air Base at Al-Kharj and an attempted drone strike on Ras Tanura (compiled event record, corroborated by ACLED's March 2026 regional update). ACLED assessed the UAE as the most heavily struck state in the opening four days, and recorded three US F-15s lost to friendly fire.

A note on those figures, because they are load-bearing. They are launch counts, not impact counts. The UAE, Kuwait and Bahrain numbers all trace back to a single Anadolu Agency tally of 19 March 2026, so they should be read as one wire service's compilation rather than three independent confirmations. The UAE subsequently revised its own count upward, to 378 ballistic missiles, 15 cruise missiles and roughly 1,835 drones by 27 March. The order of magnitude is what matters here, not the third significant figure.

Run the arithmetic with a generous intercept rate. At 95% effectiveness against 1,699 drones, roughly 85 leak through. Emirates Global Aluminium's Al Taweelah plant was hit on 28 March and faces repairs estimated at up to a year. Bahrain's Bapco refinery took damage. Thirty-two civilians were injured in a drone attack on Sitra.

This is the structural insight of the 2026 campaign season, and it is the same one Ukraine arrived at eighteen months earlier: percentage interception is a vanity metric. What does the damage is absolute leakage, and what governs absolute leakage is not how good your interceptors are. It is how many you have left, and whether you spent them on the right targets.

Ukraine's own analysts have already made this shift explicit. A 2026 study of Ukraine's air defense campaign describes exactly this pivot, "the shift from measuring air defense success by interception rates to managing system throughput and sustainability under wave-based attacks," and analyses air defense "as a living system shaped by throughput limits, interception economics, command-and-control resilience, and the ability to recover and reconfigure quickly" (Holding Back the Sky: Ukraine's Air Defense Campaign, 2022–2025, Snake Island Institute, since renamed the Serpens Center for International Security, with the Colonel Yevhen Konovalets Military School). Ukraine's interception rate climbed from 80.2% in December 2025 to roughly 90% by March 2026 (Defense News). Over the same window, Russian launches hit a record 8,161 in May 2026, an average of 263 per day (Institute for Science and International Security). Better percentage, more damage.

The magazine is the real constraint

The Gulf's exposure is not sensor coverage. It is inventory arithmetic.

Within days of the February 2026 strikes, The Economist estimated that Arab countries using US systems "may have burned through 800 PAC-3 MSE or THAAD interceptors" (cited by CSIS), a hedged figure, but the only public one. Set that against the supply side. Lockheed Martin delivered 620 PAC-3 MSE in all of 2025, about 52 per month, against a nameplate capacity near 600 per year, and a seven-year framework announced on 6 January 2026 that would lift annual capacity "from approximately 600 to 2,000 per year" by the end of 2030 (Lockheed Martin briefing transcript). That framework was announced, not finalised: CEO Jim Taiclet's words were that "the contract will reflect the exact numbers when it's finalized."

Days of firing consumed more than a year of production.

THAAD is worse. Defending Israel during the June 2025 exchange, US forces "blew through about a quarter" of the American THAAD interceptor stockpile, according to sources cited by CNN. JINSA's independent estimate is lower, roughly 92 rounds, 14.3% of inventory, but adds that replenishment would take three to eight years (JINSA cost estimate). Whichever figure is right, a single twelve-day campaign consumed a double-digit percentage of a global inventory. US procurement was 11 interceptors in 2024 and 12 in FY2025, rising to a planned 37 in FY2026 (CSIS).

The cost exchange behind those numbers is now well documented.

Item	Unit cost
PAC-3 MSE	~$4.2M; FY2027 budget $5.3M (Army Recognition)
THAAD	~$12.7M (JINSA)
Ukrainian interceptor drone	~$3,000–5,000 (NSDC)
Shahed-136 (the threat)	~$35,000 (CSIS)

On those published unit costs alone, a PAC-3 fired at a Shahed is a 120:1 exchange and a THAAD round is 360:1. Campaign-level estimates are looser and worth treating as such: drone-industry executive Arthur Erickson, quoted by The New York Times, put the aggregate ratio at "at best 10 to one" but possibly "60 or 70 to one in terms of cost, in favor of Iran." No plausible version of the arithmetic favours the defender.

And a meaningful share of what arrives is not even a weapon. In Ukraine, Gerbera decoys cost roughly one-tenth of a Shahed-136, and in a typical nightly wave of 150 drones only 20 to 30 carried explosives; the rest were wood and foam (Bojor, Vasile, Pintili & Mihalache, "Mini-Drones, Maximum Threat: C-SUAS Lessons from Ukraine," Proceedings of the International Conference Knowledge-Based Organization, 2026). And the supply of them is not the constraint: Russian industry is now producing over 6,000 one-way attack drones a month (Breaking Defense).

Spend a $4.2 million interceptor on a $3,500 piece of plywood and you have not failed to intercept. You have been successfully attacked.

Buying more magazine does not solve a targeting problem

The region's procurement response has been fast and large. The DSCA cleared 730 PAC-3 MSE for Saudi Arabia on 30 January 2026 at $9.0 billion (DSCA notification 26-13), followed on 15 July by "up to ten thousand (10,000) Advanced Precision Kill Weapon System (APKWS-II) air-to-air guidance sections; and up to ten thousand (10,000) APKWS-II air-to-ground guidance sections" worth $1.96 billion, the air-to-air half being the AGR-20F FALCO round, explicitly a counter-drone weapon (State Department, Transmittal 26-85). Kuwait, two days after the 3 June strike that damaged its international airport, was cleared for a $1.98 billion Anduril counter-UAS package covering Roadrunner-Munition and Anvil-Kinetic interceptors, Pulsar electromagnetic warfare, Long Range and Maritime Sentry towers, Menace tactical operations centres and Lattice command and control (Defense News), plus $1.02 billion of NASAMS.

Globally, announced government counter-UAS contracts exceeded $29 billion in the first quarter of 2026 alone (Unmanned Airspace). Commercial forecasters, whose numbers are vendor estimates and should be read as directional rather than authoritative, put the counter-UAS sector at $6.64 billion in 2025 growing to $20.31 billion by 2030, a 25.1% CAGR (MarketsandMarkets), with the Middle East compounding near 19.7% (Fortune Business Insights).

The cheap layer is arriving too. Ukraine produced 100,000 interceptor drones in 2025, at unit costs of $3,000–$5,000 and success rates above 60%, a Shahed kill for more than 25 times less than a Western air-defense missile (Ukraine NSDC). By February 2026 those interceptors accounted for over 70% of Shahed kills over Kyiv (Defense News). That capability is now flowing to the Gulf: Zelenskyy signed ten-year defence cooperation agreements with Saudi Arabia on 27 March and Qatar on 28 March 2026, with a parallel UAE agreement, covering co-production of interceptors, EW and maritime drones (FDD, Al Jazeera). Reported order intentions run to 5,000 units for the UAE and 2,000 for Qatar (Small Wars Journal), and Saudi Aramco is in talks with Ukrainian manufacturers for oil-field protection (Militarnyi).

Here is the part that should concentrate minds. Abu Dhabi secured access to Ukrainian Shahed interceptors on an exclusive basis relative to Saudi Arabia and Qatar, with first deliveries arriving in late April 2026, and their deployment, according to the trade intelligence outlet that broke the story, "is hampered by a lack of operators and incomplete integration into local command systems" (Intelligence Online). That report is single-sourced and paywalled, and we have found no independent corroboration of the integration detail, but it is consistent with everything else in this analysis.

The effectors arrived. The decision layer that would use them did not.

Soft kill is losing its share, and that changes the math again

For a decade the standard answer to cheap drones was cheap defeat: jam the control link and let the airframe fall. That answer is eroding fast.

Russia is now launching roughly 10,000 FPVs a day with about 60% fiber-optic control in key sectors, and Chinese optical-fiber shipments to Russia rose from 119,000 miles in May 2025 to 328,000 miles by August 2025 (Euromaidan Press). In parts of Kharkiv oblast, roughly 90% of Russian attack drones are fiber-controlled, with spools reaching 50 km in 2026 (Gwara Media). Fiber FPVs are already bypassing substation defenses in Sumy (Reuters).

Where the link cannot be cut, autonomy closes the gap. Drones with autonomous navigation raise target engagement success from 10–20% to around 70–80% (CSIS analysis via Breaking Defense). The Hague Centre for Strategic Studies puts it plainly: "the proliferation of unjammable fibre-optic drones has drastically decreased the effectiveness of Counter-UAV EW systems" (HCSS, February 2026).

Jamming is not obsolete. Ukrainian EW still neutralises nearly half of drones on some nights in the deep-rear Shahed fight (Reuters). But it has become a conditional effector rather than a default one. Which means a defender now needs something it did not need before: the ability to judge, per track, in seconds, whether jamming will work at all, and to route to a kinetic or directed-energy option when it will not. That judgement is a software problem.

Command and control is where defenses actually fail

The failure mode of modern air defense is not a missed intercept. It is a misclassification under time pressure with a degraded picture.

The 152-page US Navy investigation into the USS Gettysburg's shootdown of a friendly F/A-18 on 21–22 December 2024 reads as a specification for what an air-defense C2 system must prevent (Military Times reconstruction). Link 16 was down from about 6 p.m. to just after 9 p.m., on a network the investigation described as having "noticeably degraded" with regular outages beforehand. IFF "casualties" went unaddressed. The Hornets were classified "unknown," not "friendly," on a track that had earlier been falsely identified as a foreign anti-ship cruise missile. The battle watch captain took nearly five minutes to issue a cease-fire. Two SM-2s were fired; the second missed by one or two plane-lengths. The investigation called it "entirely avoidable."

The signature is not new. The 2003 Patriot fratricides produced the same forensic profile: positive electronic identification of "low reliability," jammed communications, non-responding transponders, atrophied crew skills (Army Center for Lessons Learned, via ABC). Twenty-one years apart, degraded IFF plus a degraded datalink plus a compressed timeline produced the same outcome.

Now compress the timeline further. US planners protecting the 2026 World Cup worked to roughly 90 seconds from detection to decision (CSIS). Reviewing NATO's C-UAS TIE24 trials, Romanian military researchers characterise sensor coverage below 300 metres as "fragmented across national architectures," and find soldiers engaging FPV drones with small arms achieve approximately 25% engagement efficiency, a 4:1 performance deficit (Bojor, Vasile, Pintili & Mihalache, 2026).

The GCC's own diagnosis lands in the same place. Following the extraordinary Joint Defence Council in Doha in September 2025, the Unified Military Command was directed to transmit the air situation across all GCC operations centres and accelerate the ballistic-missile early warning system. GCC Secretary-General Jasem Mohamed Albudaiwi said in December 2025 that the Joint Gulf Missile Defence Shield still requires resolving "many technical aspects," with a lack of interoperability between GCC militaries identified as a major challenge, and hardware and software differences described as "subtle but significant" (ORF Middle East).

Subtle differences in software are exactly what turns six national air pictures into six incompatible ones at the moment they need to be one.

---

What "good" looks like: allocation as a computed problem

There is a working reference for what the decision layer should do, and it is nearly two decades old.

Iron Dome's battle management system "constantly plots the ballistic trajectories of each threat, assessing if it is presenting a threat to the protected areas." Only those threats are marked red and trigger the computation of intercept solutions, which is to say the system's first act is to decide which tracks not to engage. For each threatening target it computes over 800 intercept solutions within a few seconds, evaluating combinations of target, interceptor, launcher and launch unit against no-fire zones, collateral-damage minimisation and deconfliction with friendly aircraft. It then proposes a preferred solution and requests operator permission to launch, with manual override available (Defense Update).

Two things in that description matter more than the rest. First, the system's most valuable function is deciding not to shoot. Second, the human authorises rather than calculates.

The academic literature on weapon-target assignment reaches the same conclusion from the other direction. In simulated air-defense scenarios pitting seven long-range and five short-range units against mixed waves of cruise bombs, UAVs, fighters and jammers, untrained agents adopting a random policy waste "too much ammunition in defending against the first attacks and eventually failing because they run out of resources," while an architecture that "divides the air defense operations process into two parts, resource scheduling and mission execution" reaches higher win ratios faster than conventional baselines (Liu, Wang, Guo, Wang & Fu, Frontiers in Neurorobotics, 2022).

That is the whole Gulf problem in a simulation abstract. Magazine depletion, not intercept probability, is the binding constraint, and the variable that governs it is allocation policy.

The market has already repriced accordingly. In March 2026, JIATF-401 awarded Anduril an $87 million task order for tactical counter-UAS command and control, the first order under an Army enterprise agreement worth up to $20 billion over ten years, described as the largest counter-UAS contract ever placed (Breaking Defense, Defense One). The $20 billion figure deserves a caveat Anduril itself supplied: "There's no money attached to it, this is just a contract vehicle." The $87 million task order is the real money. What matters is what it bought. Not an interceptor. Not a radar. A fire-control software layer, replacing the incumbent FAAD C2 (Breaking Defense).

The lesson from Ukraine's institutional analysts is identical: "software integration and information management — not platform performance — are the primary drivers of operational tempo and combat effectiveness" (Kostiuk, Patiuk, Shapochkina & Tenenbaum, Mapping the MilTech War, IFRI, February 2026).

---

Why this is a sovereignty question in the Gulf specifically

Every other region can treat the decision layer as a procurement line item. The Gulf cannot, for three reasons documented in the last eighteen months.

First, the layer is where localisation targets actually bite. GAMI reported 24.89% localisation of military spending at end-2024, against a Vision 2030 target above 50% (GAMI), up from 4% in 2018, with 572 establishment approvals and licences covering 311 establishments by end-2024 (GAMI) and the workforce growing from 25,000 in 2020 to 34,000 (Arab News). Saudi Arabia's 2025 military expenditure was $83.2 billion, the world's eighth largest at 6.5% of GDP (SIPRI), with SAR 240 billion allocated for FY2026 (Ministry of Finance). Closing a 25-point localisation gap through metal-bending alone is arithmetically hard. Software is where domestic value-add scales without a foundry.

Second, the export-control ceiling is real and repeatedly demonstrated. MTCR Category I imposes a "strong presumption of denial" on Reaper-class systems. As the Mitchell Institute's Heather Penney put it, "the US starts at 'no' when allies and partners ask for this capability" (FlightGlobal). The UAE suspended its $23 billion F-35 package in December 2021 over what officials called "sovereign operational restrictions" (The Washington Institute). A September 2025 MTCR reinterpretation eased drone transfers (Middle East Eye), which demonstrates the point rather than refuting it: the ceiling moves when someone else decides to move it.

Third, buyers have already priced sovereignty into commercial terms. Technology access, escrow, localisation and data-sovereignty requirements are now core contract terms in GCC defence procurement, with export-control classification built into bid strategy from day one (DLA Piper). Lockheed Martin's response was to announce a C2 software factory in Saudi Arabia in January 2026, explicitly framed around software that "can be Saudi-owned intellectual property" (Breaking Defense). The UAE's economy minister frames the category directly: "AI is a state of sovereignty. Like you spend budgets on defense, on cyber, you have to spend budget on AI" (INSS).

The region is not short of effectors. Saudi Arabia inaugurated three THAAD brigades in FY2025 (Ministry of Finance). EDGE unveiled 42 new products at the 2025 Dubai Airshow (Euro-SD) and closed 2025 with $4.9 billion of revenue and a $21.1 billion backlog (EDGE). World Defense Show 2026 in Riyadh produced 220 agreements and over SR 33 billion in deals (Arab News). Red Sands 2025 put 20 counter-UAS platforms and 300+ personnel on the Shamal-2 range, running everything from passive RF geolocation to a Patriot live fire against Group 3–5 UAS through a single Combined Base Defense Operations Center (CENTCOM, Air Defense Artillery Journal).

Sensors and effectors are becoming commodities, sourced from many countries. The scarce thing is the sovereign layer that decides which one fires.

Five design requirements the last eighteen months imply

For anyone specifying a counter-UAS command-and-control capability in this region, the evidence points to a fairly specific requirements list.

1. Optimise for magazine, not for probability of kill. The objective function is not "intercept this track." It is "survive the campaign." That means explicit cost-aware assignment, deliberate non-engagement of tracks that do not threaten a protected area, and reserve management across waves: the logic Iron Dome's battle management system has run since 2009 (Defense Update) and that untrained policies conspicuously fail (Liu et al.).
2. Treat effectors as data, not as integrations. When a $1,000 interceptor drone becomes available in April and needs to be in the engagement chain in May, the system that requires a firmware release has already lost. Abu Dhabi's integration bottleneck is the cautionary case (Intelligence Online). MOSA is statutory in the US for a reason (10 U.S.C. § 4401).
3. Model soft kill as conditional, not default. The system must estimate whether jamming will defeat a given track, and fiber-optic control and autonomous terminal navigation mean the answer is increasingly no (HCSS). It must fail over to a kinetic or directed-energy option without human re-planning.
4. Degrade gracefully, because the datalink will fail. Gettysburg lost Link 16 for roughly three hours, on a network already prone to regular outages (Military Times). A C2 layer that assumes a clean common operating picture is a C2 layer that will produce a fratricide report.
5. Keep the human on the loop, and give them one decision. DoD Directive 3000.09 requires "appropriate levels of human judgment over the use of force" (DoDD 3000.09), and the UN CCW process is converging on similar constraints, with support for opening negotiations on the rolling text growing from over 40 states to over 70 during the March 2026 session (WILPF). Meanwhile the operational window is 90 seconds (CSIS). Both facts are true at once. The only resolution is the Iron Dome pattern: the machine computes the full solution set, the human confirms or overrides one recommendation. Anything that asks an operator to compare options in that window is a design failure.

---

The layer between seeing and striking

The Gulf spent a decade buying the ability to see incoming threats and a decade buying the ability to destroy them. Both efforts succeeded. February and March 2026 demonstrated that neither is sufficient, because the campaign is won or lost in the allocation decisions made between them: thousands of them, in seconds, against mixed waves of decoys and real weapons, with a finite and slowly replenished magazine.

That layer is software. In this region, it needs to be sovereign software, for reasons that are commercial as much as strategic: it is where localisation targets can realistically be met, where export-control ceilings do not apply, and where a defence built from many nations' sensors and many nations' effectors can still answer to one national command.

Adamant builds OWL for exactly that layer: detect, track, decide, assign, and manage what is left in the magazine. Built in the Gulf, for the Gulf.