License Plate Readers Are Watching You. Now They Want to Read Your Phone.

Not Just License Plates: They’re Going to Track So Much More - YouTube

Consumer Investigative Report  ·  Privacy & Technology

Informed Consumer — Special Investigation

Privacy Alert

A defense contractor's new surveillance system pairs roadside cameras with Bluetooth, Wi-Fi, and RFID scanners to build a dossier on every driver—and every passenger—who passes by. No warrant required.

Investigative Report  |  June 2026  |  Based on company disclosures, court records, and civil liberties research

Bottom Line Up Front

Italian defense giant Leonardo is marketing a system called ELSAG SignalTrace that bolts wireless-signal sensors onto existing license plate reader cameras, silently harvesting the Bluetooth IDs, Wi-Fi signatures, RFID tags, and other electronic identifiers broadcast by every phone, wearable, infotainment system, access badge, and pet microchip in range of the roadside unit. The data is fused with license plate images and stored in a corporate database, queryable by law enforcement at will and without a warrant under current federal law. SignalTrace works in rail stations and shopping centers as well as roadsides—meaning it can track pedestrians, not just drivers. No federal statute explicitly governs this technology. Civil-liberties groups warn it represents the most significant escalation in mass surveillance infrastructure since license plate readers themselves were introduced, and real-world abuses of the existing plate-only system—including documented use against a woman who had an abortion—illustrate exactly what is at stake.

The Technology: From Plate to Profile

For more than a decade, automated license plate readers (ALPRs) have been a fixture of American roads. These high-speed cameras photograph passing vehicles, convert the plate number to machine-readable text, stamp it with a time and location, and upload it to a searchable database. Widely used by local police, border agents, and parking enforcement, ALPRs have become the backbone of a nationwide vehicle-tracking infrastructure. According to the Electronic Frontier Foundation's Atlas of Surveillance, more than 1,800 law enforcement agencies have deployed ALPRs, and at least 4,000 agencies can run searches through the commercial ALPR network maintained by one major vendor alone.

Now that infrastructure is being upgraded in a way that changes the nature of the surveillance entirely. Leonardo US Cyber and Security Solutions—the American arm of Leonardo S.p.A., an Italian defense and aerospace conglomerate with a market capitalization exceeding €29 billion—is actively marketing a product called ELSAG SignalTrace that turns a roadside plate camera into something far more ambitious: a passive signal-intelligence node capable of harvesting the electronic identifiers broadcast by every wireless device within range.

According to Leonardo's own product sheet and website, SignalTrace integrates with the company's existing ELSAG ALPR cameras and its back-end Enterprise Operations Center (EOC) software. In addition to reading license plates, an equipped sensor collects the unique identifiers emitted by mobile phones, Bluetooth-enabled wearables such as smartwatches and wireless earbuds, RFID-enabled items including workplace access badges, and even pet microchips. The system's algorithms then look for patterns: when the same cluster of device identifiers repeatedly appears alongside the same vehicle, SignalTrace links those devices to the vehicle's plate number and builds what the company calls an "electronic fingerprint."

"When multiple devices consistently move together with a vehicle, SignalTrace's algorithms link them to that vehicle's license plate and time-stamped location data. This correlation provides investigators with another layer of actionable intelligence, even if a suspect changes or removes a plate."
— Leonardo ELSAG SignalTrace product sheet, 2025

That last phrase is telling. By correlating a vehicle's occupants with their devices, investigators can identify and track a person even if they switch cars, borrow a friend's vehicle, or deliberately obscure their plate. The car tracking a person; the person is now the tracking anchor.

Leonardo received a patent for the core technology underpinning SignalTrace in 2024. In its public patent announcement, the company described ELSAG EOC Plus as "an electronic detection system designed to help law enforcement identify people of interest by the signatures their electronic devices emit such as fitness trackers, RFID tags, and local signals from mobile phones." Leonardo's customers include police departments, border security agencies, and other government organizations across the United States. The company's U.S. arm also holds contracts with U.S. Special Operations Command and the General Services Administration.

Not Just for Roads: The Shopping-Mall Problem

One detail in Leonardo's public materials deserves particular attention. A bullet point in the company's product literature lists SignalTrace as "effective for use in off-road areas such as rail stations and shopping centers." That single phrase dissolves the assumption that this system is exclusively about tracking cars.

License plate readers, by their nature, require a vehicle. SignalTrace's sensor array does not. A unit placed at a train station entrance, inside a mall corridor, or at a sporting venue can harvest the same electronic fingerprints from pedestrians carrying phones and wearing smartwatches that a roadside unit harvests from drivers. Without a vehicle in the picture, the data would be correlated differently—but the signal collection is the same, and the stored identifiers are just as linkable to an individual's movements over time.

Leonardo's materials state that the system "stores device and correlation data securely in the EOC Enterprise Operations Center for future queries and analysis." In other words, even if no investigation is underway when you walk past a sensor, your device identifiers are logged and retained. "The SignalTrace system simply stores data until a specific request is made of the system by an investigator," the company's website states—framing passive, continuous collection of data on innocent people as a feature rather than a concern.

The Existing Plate-Reader Problem: Already Serious

To understand why SignalTrace represents an escalation, it helps to understand how badly the existing, plate-only ALPR infrastructure has already been abused.

In 2016 and 2017, EFF and the journalism nonprofit MuckRock obtained records from more than 200 law enforcement agencies representing over 2.5 billion license plate scans. The data showed that 99.5 percent of plates scanned were not associated with any suspected crime. On average, agencies shared that data with at least 160 other agencies. The EFF concluded that repeated plate captures at multiple points along a person's daily route were sufficient to establish what intelligence analysts call a "pattern of life"—revealing where someone lives, works, worships, receives medical care, and who their associates are.

That concern has since moved from theoretical to documented. In May 2025, a Johnson County, Texas sheriff's deputy used the nationwide plate-reader network maintained by Flock Safety—then comprising roughly 83,000 cameras across the country—to search for a woman who had self-administered an abortion. The search record stated plainly: "had an abortion, search for female." The officer queried 6,809 different camera networks, including cameras in states where abortion is legal, such as Washington and Illinois.

When the search was first reported by 404 Media, the Johnson County Sheriff and Flock Safety both denied it had anything to do with criminal investigation, claiming the search was a missing-person welfare check. Court records obtained by EFF in October 2025 told a different story: the case had been opened as a "death investigation" of a "non-viable fetus" on the same day the ALPR search was conducted, and deputies had consulted the District Attorney about whether the woman could be charged with a crime. The DA concluded charges were not possible under Texas law, but the investigation had proceeded for weeks, including review of the woman's text messages.

Documented ALPR Abuses: A Partial Record

• Abortion investigation, Texas (2025): Johnson County deputies used Flock Safety's 83,000-camera network to search for a woman who had self-administered an abortion. Court records showed a criminal "death investigation" was open, contradicting public claims it was a welfare check.
• Protest surveillance (2025): EFF obtained audit logs showing hundreds of ALPR searches by police related to political demonstrations, including the 50501, Hands Off, and No Kings protests.
• Racially discriminatory searches: EFF documented more than 80 law enforcement agencies using racial slurs or ethnic stereotypes (targeting Romani people) in ALPR search queries, often without specifying a crime.
• School residency checks, noise complaints, background investigations: An EFF analysis of millions of Flock Safety searches found that, without a warrant requirement, agencies have used ALPR data "for virtually any whim."
• ICE immigration enforcement: Flock Safety data was used by U.S. Immigration and Customs Enforcement as part of the Trump administration's deportation program; the Illinois Secretary of State launched an audit after EFF research showed CBP was accessing Illinois data in violation of state privacy law.
• Officer stalking risk: The EFF's Street Level Surveillance project notes that individual officers have historically abused plate-reader access for personal reasons; a 1998 Washington, D.C. case involved an officer blackmailing patrons of a gay bar after looking up their plates.

The EFF's full-year 2025 investigation found that between June 2024 and October 2025, the San Jose Police Department and partner agencies searched San Jose's Flock database 3,965,519 times—all without warrants. That figure led EFF and the ACLU of Northern California to file a federal lawsuit against San Jose challenging warrantless ALPR searches. The suit remains pending.

Flock Safety itself has been expanding in ways that compound privacy concerns. The company has announced that police departments will be able to obtain not only still photos from plate cameras but also video, with live-feed and 15-second clip capabilities. It has also integrated with data brokers, allowing law enforcement to combine plate-reader sightings with commercial datasets about individuals—an arrangement critics call an "end-run" around the warrant requirements the Supreme Court has imposed on direct government surveillance.

The Legal Landscape: A Framework Built for a Simpler Era

Courts have generally held that ALPRs, as currently deployed, do not trigger Fourth Amendment warrant requirements. The key precedent is Carpenter v. United States (2018), in which the Supreme Court required a warrant for historical cell-site location information (CSLI) because it produced an "all-encompassing" log of a person's movements. Courts since then have consistently found that ALPR snapshots—discrete images at fixed locations—fall short of the Carpenter threshold. In Commonwealth v. Church, the Virginia Court of Appeals reversed a lower-court ruling that had required warrants for Flock data, holding that the technology does not require one. A 2025 federal district court ruling in Kansas reached the same conclusion.

But those rulings were written for plate-only ALPRs. SignalTrace is a qualitatively different system: it combines license plates with Bluetooth harvesting, Wi-Fi collection, RFID scanning, and persistent algorithmic correlation—stored indefinitely in a corporate database. Legal scholars and the courts themselves have recognized that the constitutional calculus may shift as the technology evolves. A concurring judge in the Ninth Circuit's United States v. Yang (2020) explicitly cautioned that "if the technology evolves in the way that amici hypothesize, then perhaps in the future a warrant may be required."

SignalTrace is precisely that evolution. No federal statute currently regulates mass Bluetooth or RFID harvesting from public spaces. State ALPR laws were written to govern plate-image collection; they say nothing about signal interception. As one analysis put it, SignalTrace sits in "a legal gray zone"—permitted not because legislators or courts have approved it, but because they had not yet imagined it when existing rules were written.

The Electronic Communications Privacy Act of 1986 (ECPA), which governs interception of electronic communications, contains an exception for signals "readily accessible to the general public." Leonardo's defense—that SignalTrace merely captures "device frequencies emitted into the air" and "does not decrypt or capture the contents of the devices or their communications"—appears calibrated to fit within that exception. Whether courts will agree that passively harvesting a device's unique identifiers, linking them to a person, and retaining them indefinitely amounts to something more than reading a publicly broadcast signal is an open question that will almost certainly reach the federal courts.

Security Vulnerabilities in the Existing Infrastructure

Whatever the legal status of ALPR data, the security of that data once collected is a distinct and serious concern. In June 2024, the Cybersecurity and Infrastructure Security Agency (CISA) issued an advisory identifying seven vulnerabilities in one major brand of ALPR camera systems, including missing encryption and insufficiently protected credentials. Security researchers have demonstrated the ability to access massive ALPR datasets and live-broadcast camera feeds. A 2025 Virginia State Crime Commission report found systemic non-compliance with the state's own ALPR data rules: 13 percent of agencies admitted granting out-of-state entities continuous access to their data; 6 percent had given the federal government continuous access; and 35 percent failed to implement required public notice measures.

One in ten ALPR readings contain an error, according to a study cited by the ACLU of Iowa in its December 2025 statewide ALPR survey. Those errors have real consequences: people have been stopped by police at gunpoint due to incorrect plate reads. Adding wireless device harvesting to this error-prone foundation does not reduce the risk of mistaken identification; it compounds it.

What You Are (and Are Not) Broadcasting

To understand what SignalTrace can actually capture, it helps to know what your devices broadcast without your active knowledge. Bluetooth Low Energy (BLE) devices—which include most modern smartphones, smartwatches, wireless earbuds, AirTags, and tire pressure monitoring systems—continuously advertise their presence to nearby devices. While Apple introduced MAC address randomization to reduce trackability, research has documented that randomization is imperfect; other identifiers (including certain BLE advertising payloads) can persist across randomization cycles and allow consistent re-identification of a device. Wi-Fi probing, in which a device searches for familiar networks, similarly broadcasts identifiers. RFID chips embedded in workplace access badges, transit passes, and pet microchips emit a fixed identifier when queried by a nearby reader.

Leonardo's disclaimer—that the system does not read the content of communications—is technically accurate but practically misleading. The content of your phone calls is not what creates a surveillance dossier. What creates a dossier is the persistent association of your device's unique identifiers with specific locations at specific times: the oncologist's office on Tuesday, the political meeting on Thursday, the border crossing on Saturday. That pattern of presence, not the content of your conversations, is the intelligence product.

What You Can Do

Individual technical countermeasures offer limited protection. Disabling Bluetooth reduces your exposure to BLE harvesting, but many vehicle infotainment systems re-enable Bluetooth automatically on restart, and tire pressure monitoring systems broadcast continuously without any user-accessible off switch. Leaving your phone at home eliminates that signal, but not the RFID in your access badge or transit card. Turning off Wi-Fi reduces probe-request broadcasts, but does not eliminate them on all devices. MAC address randomization, while helpful, is not a complete solution given the state of current research on BLE fingerprinting.

More durable protections require policy action:

What Consumers Can Do Now

• Attend public meetings. Local governments typically vote on ALPR contracts; municipalities from Austin to Evanston to Eugene have canceled or declined Flock Safety contracts following organized community opposition. Your city council members are elected officials who are legally obligated to hear from constituents.
• Contact state legislators. Ask them to extend state ALPR data-protection laws to cover wireless signal harvesting, to require warrants for database queries, to mandate data minimization and short retention periods, and to prohibit cross-agency and cross-state data sharing without judicial oversight.
• Support federal legislation. The Fourth Amendment Is Not For Sale Act, which would ban law enforcement from purchasing location data from brokers without a warrant, passed the House in 2024 but was blocked in the Senate. Urge your senators to support warrant requirements for commercial surveillance data.
• Reduce your wireless footprint where practical. Disable Bluetooth when not in active use. Use a Faraday sleeve for RFID-enabled badges when not needed. Understand that these steps reduce but do not eliminate your exposure.
• File public records requests. Organizations like EFF and MuckRock have published guides to requesting ALPR audit logs from your local police department. These logs have been the foundation of every major documented abuse case.

The core issue is structural. Mass surveillance infrastructure, once deployed, does not remain confined to its original stated purpose. Systems sold for stolen-car recovery are used to investigate abortions. Systems designed for criminal suspects are queried in noise-complaint investigations. Systems marketed to local police are accessed by federal immigration enforcement. The pattern is consistent and well-documented. Adding a wireless device-harvesting layer to a nationwide network of roadside sensors—and extending that network into train stations and malls—does not change that pattern. It accelerates it.

"When a single search can access more than 83,000 cameras across nearly the entire country," EFF legislative activist Rin Alajaji observed in October 2025, "the potential for abuse is enormous."

SignalTrace is that same network, upgraded to also know who is in every car.

Verified Sources

1. [1] Leonardo US Cyber and Security Solutions. ELSAG SignalTrace — Integrated Signal Intelligence Platform (product page). Leonardo Company US, 2025.
   https://www.leonardocompany-us.com/lpr/elsag-signaltrace
2. [2] Ismail, Adam. "License Plate Cameras Will Soon Track Phones, Wearables, Infotainment, and Even Your Pets." The Drive, June 2026.
   https://www.thedrive.com/news/license-plate-cameras-will-soon-track-phones-wearables-infotainment-and-even-your-pets
3. [3] "Leonardo's SignalTrace Could Let Police Plate Readers Track Your Devices." The Deep Dive, June 2026.
   https://thedeepdive.ca/leonardo-signaltrace-alpr-device-tracking/
4. [4] "License Plate Readers Move Beyond Plates to Track Phones, Wearables, and Pet Microchips." SOFX, June 2026.
   https://www.sofx.com/license-plate-readers-move-beyond-plates-to-track-phones-wearables-and-pet-microchips/
5. [5] Guariglia, Matthew. "New ALPR Vulnerabilities Prove Mass Surveillance Is a Public Safety Threat." Electronic Frontier Foundation Deeplinks Blog, June 19, 2024.
   https://www.eff.org/deeplinks/2024/06/new-alpr-vulnerabilities-prove-mass-surveillance-public-safety-threat
6. [6] Maass, Dave and Beryl Lipton. "Data Driven: Explore How Cops Are Collecting and Sharing Our Travel Patterns Using Automated License Plate Readers." Electronic Frontier Foundation, updated April 2021.
   https://www.eff.org/pages/automated-license-plate-reader-dataset
7. [7] Electronic Frontier Foundation. "Automated License Plate Readers (ALPRs)." Street Level Surveillance, updated October 1, 2023.
   https://sls.eff.org/technologies/automated-license-plate-readers-alprs
8. [8] Guariglia, Matthew, et al. "More License Plate Reader Mission Creep: School Residency Verification, Background Checks, and Noise Complaints." Electronic Frontier Foundation Deeplinks Blog, May 2026.
   https://www.eff.org/deeplinks/2026/05/more-license-plate-reader-mission-creep-school-residency-verification-background
9. [9] Maass, Dave, and Rindala Alajaji. "She Got an Abortion. So a Texas Cop Used 83,000 Cameras to Track Her Down." Electronic Frontier Foundation Deeplinks Blog, May 2025; updated October 7, 2025.
   https://www.eff.org/deeplinks/2025/05/she-got-abortion-so-texas-cop-used-83000-cameras-track-her-down
10. [10] Maass, Dave, and Rindala Alajaji. "Flock Safety and Texas Sheriff Claimed License Plate Search Was for a Missing Person. It Was an Abortion Investigation." Electronic Frontier Foundation Deeplinks Blog, October 2025.
    https://www.eff.org/deeplinks/2025/10/flock-safety-and-texas-sheriff-claimed-license-plate-search-was-missing-person-it
11. [11] Electronic Frontier Foundation. "EFF's Investigations Expose Flock Safety's Surveillance Abuses: 2025 in Review." EFF Deeplinks Blog, January 2, 2026.
    https://www.eff.org/deeplinks/2025/12/effs-investigations-expose-flock-safetys-surveillance-abuses-2025-review
12. [12] American Civil Liberties Union. "Flock's Aggressive Expansions Go Far Beyond Simple Driver Surveillance." ACLU.org, August 18, 2025.
    https://www.aclu.org/news/privacy-technology/flock-roundup
13. [13] Schneier, Bruce. "Enhanced License Plate Tracking." Schneier on Security, June 2026.
    https://www.schneier.com/blog/archives/2026/06/enhanced-license-plate-tracking.html
14. [14] Flock Safety. "Automated License Plate Readers and the Fourth Amendment: A Public-Safety-by-Design Perspective from Flock." Flock Safety White Paper, November 2025.
    https://www.flocksafety.com/blog/automated-license-plate-readers-and-the-fourth-amendment-a-public-safety-by-design-perspective-from-flock
15. [15] ACLU of Iowa. "Automatic License Plate Reader Report Raises Concerns About Expansion of Government Surveillance in Iowa." Press release, December 16, 2025.
    https://www.aclu-ia.org/press-releases/automatic-license-plate-reader-report-raises-concerns-about-expansion-of-government-surveillance-in-iowa/
16. [16] Carpenter v. United States, 585 U.S. 296 (2018). U.S. Supreme Court. Warrant requirement for historical cell-site location information.
    https://www.supremecourt.gov/opinions/17pdf/16-402_h315.pdf
17. [17] United States v. [Defendant], No. 2024cr10010-48, U.S. District Court for the District of Kansas (May 30, 2025). ALPR Fourth Amendment analysis under Carpenter.
    https://ecf.ksd.uscourts.gov/cgi-bin/show_public_doc?2024cr10010-48+=
18. [18] Cybersecurity and Infrastructure Security Agency (CISA). Advisory on ALPR camera system vulnerabilities. U.S. Department of Homeland Security, June 2024. (Summarized in Style Weekly, June 2026.)
19. [19] Samples, Jackson. "Barcoding Bodies: RFID Technology and the Perils of E-Carceration." 23 Duke Law & Technology Review 89–113 (2024).
    https://scholarship.law.duke.edu/dltr/vol23/iss1/4/
20. [20] "It's Time to Turn Off the Flock Safety Cameras." Style Weekly (Richmond, VA), June 2026. Includes Virginia State Crime Commission compliance data and CISA advisory summary.
    https://www.styleweekly.com/its-time-to-turn-off-the-flock-safety-cameras/
21. [21] "Surveillance Cameras Now Track Phones, Smartwatches, and Pet Microchips Alongside License Plates." Georgia Record, June 22, 2026. (Includes summary of Leonardo's 2024 patent announcement and product-sheet language.)
    https://www.georgiarecord.com/business-ga/2026/06/22/surveillance-cameras-now-track-phones-smartwatches-and-pet-microchips-alongside-license-plates/
22. [22] Electronic Frontier Foundation. "California Automated License Plate Reader Policies." (Includes Flock Safety abortion-investigation and SFPD inter-state data-sharing documentation.)
    https://www.eff.org/pages/california-automated-license-plate-reader-policies

© 2026  |  This report is intended for educational and consumer-information purposes. It does not constitute legal advice.  |  Reproduction for non-commercial use permitted with attribution.

 

 

Türkiye's CCA Surge:

Leonardo and Baykar K-SWARM Trials Show M-346 Evolving into Airborne Command Node for KIZILELMA Unmanned Fighter

Story Link 

How Baykar and TAI Are Outpacing the West in Autonomous Air Combat Teaming

With live crewed-uncrewed flight trials, world-first autonomous BVR kills, and a trans-national production architecture anchored to GCAP, Ankara has moved from drone exporter to architect of the next generation of air warfare.

By: Stephen L Pendergast 

Published: June 22, 2026  

Classification: Open Source / Unclassified

Bottom Line Up Front

Türkiye has executed a rapid and verifiable series of collaborative combat aircraft (CCA) milestones that place it ahead of most Western programs in converting autonomous teaming from concept to live operational demonstration. Baykar's KIZILELMA unmanned fighter became the world's first autonomous jet-powered aircraft to destroy an aerial target with a beyond-visual-range missile (November 2025) and the first to conduct fully autonomous close formation flight with a second armed jet (December 2025). The subsequent K-SWARM live trials in May 2026 showed an Italian M-346 acting as an airborne command node for KIZILELMA, validating a deployable crewed-uncrewed architecture now linked to GCAP development. Simultaneously, TAI's ANKA-3 stealth UCAV completed critical design review and is entering serial production for a Turkish Air Force order expected to exceed 50 aircraft, while Aselsan has activated an Indigenous Flight Datalink (IVDL) connecting the TF KAAN fifth-generation fighter to both KIZILELMA and ANKA-3. Together, these programs constitute an integrated autonomous combat air ecosystem — not a collection of demonstrators — backed by a NATO-connected European production infrastructure through the Leonardo-Baykar LBA Systems joint venture. The pace and breadth of this effort materially challenges US, European, and Chinese assumptions about which nations will shape the doctrine and industrial architecture of autonomous air combat in the 2030s.

On June 22, 2026, Leonardo and Baykar announced the successful completion of the first live K-SWARM trials, confirming that an Italian M-346 jet had commanded and coordinated Baykar's KIZILELMA unmanned fighter through autonomous formation changes, separations, and rejoins over Baykar's Çorlu flight center in Türkiye during May. The announcement distilled months of largely quiet but operationally significant progress into a single headline. Yet the K-SWARM result is better understood not as a breakthrough event but as the latest data point in a deliberate, multi-year Turkish program to build a complete autonomous combat air ecosystem — one that now spans two sovereign defense companies, a newly activated NATO-adjacent production venture, and connections to the highest-profile sixth-generation fighter program in Europe.

The K-SWARM Trials: From Simulator to Sky

The K-SWARM program paired Leonardo's M-346 Fighter Attack variant — configured with a newly integrated onboard avionics suite and a dedicated crewed-uncrewed computing system — with Baykar's KIZILELMA unmanned combat air vehicle. An Italian Air Force T-346A flew as a chase aircraft, applying disciplined flight-test methodology to the complex mixed formation. Following KIZILELMA's autonomous taxi and takeoff, the unmanned fighter rejoined the M-346 in flight using Baykar's Smart Fleet Autonomy algorithms, developed and validated through the company's Hardware-in-the-Loop laboratory at Çorlu before the live phase began.

Once the formation stabilized, the M-346 assumed full command authority. Pilots directed formation changes, separations, and rejoins; KIZILELMA executed each maneuver autonomously and with accurate command response. Critically, the architecture employed supervised autonomy rather than remote piloting: the crewed aircraft retained tactical authority, while the unmanned platform handled complex flight tasks without imposing excessive cockpit workload on the pilot. An advanced radio-frequency data exchange system synchronized mission data between the platforms, protected in real time by Leonardo's GCC Tactical Platform cyber-defense architecture.

"The M-346 acted as a human-controlled airborne command node, while KIZILELMA functioned as an autonomous combat asset responding to pilot direction — autonomy as an extension of pilot authority, not a replacement for it."

Leonardo's preparatory engineering chain was substantial. The company's Avionic and Flight Control Innovation Labs and its PC2LAB product and concept laboratory in Turin developed, modeled, and refined the CUC-T (Crewed/UnCrewed Teaming) algorithms and tactics before the live phase began, linked to an M-346 Full Mission Simulator in Venegono. The result was a validated digital thread from software laboratory to live flight — a process compressed to months rather than years by Baykar's mature autonomy infrastructure and KIZILELMA's advanced onboard AI capabilities.

KIZILELMA: A Fighter-Class Platform, Not a Converted Drone

Understanding why K-SWARM matters requires understanding what KIZILELMA is. Baykar's unmanned fighter is not a surveillance drone with weapons tacked on. The aircraft is a single-engine, low-observable, carrier-capable, jet-powered multirole UCAV with an 8.5-ton maximum takeoff weight, a 1.5-ton payload capacity, a cruise speed of Mach 0.6, and a combat radius of approximately 500 nautical miles. Its canard-delta configuration with internal weapons bays provides both maneuverability and low-observable strike capability. The third production prototype, featuring afterburning propulsion, aerodynamic refinements, and an updated avionics architecture, completed its first flight in September 2024. Mass production began in October 2024, with Baykar Chairman Selçuk Bayraktar stating an objective of more than ten aircraft by 2026.

The platform's combat credibility rests on a dense 2025 test record. On November 20, 2025, KIZILELMA electronically engaged and simulated the destruction of a Turkish Air Force F-16 using its MURAD AESA radar in conjunction with a Gökdoğan beyond-visual-range missile digital kill chain — the first time such a sequence was validated on an unmanned platform. Ten days later, on November 30, Baykar announced that KIZILELMA had executed what the company described as the world's first live autonomous BVR air-to-air kill, destroying a high-speed jet-powered target drone over the Black Sea near Sinop with a Gökdoğan missile guided by MURAD radar — a verified engagement, not a simulation. On December 28, 2025, two KIZILELMA aircraft accomplished the world's first fully autonomous close formation flight by a pair of armed, jet-powered unmanned aircraft, relying entirely on artificial intelligence, onboard sensors, and instantaneous data exchange between airframes.

In May 2026, Baykar signed the first KIZILELMA export agreement, with Indonesian company PT Republik Aero Dirgantara committing to an initial batch of 12 aircraft with options extending to 60, plus local production and maintenance facilities — a signal that international confidence in the platform has reached contract-award level.

ANKA-3: The TAI Stealth Wingman

While KIZILELMA has received the most international attention, the second pillar of Türkiye's autonomous combat air architecture is Turkish Aerospace Industries' ANKA-3 stealth UCAV. The flying-wing platform — 7.9 meters long with a 12.5-meter wingspan and a maximum takeoff weight of 6,500 kilograms — is designed for high-subsonic operations at up to 40,000 feet with approximately ten hours of endurance and a 1,600-kilogram payload. Its configuration with two internal weapon bays enables low-observable strike carriage distinct from the older ANKA family's external-stores approach.

ANKA-3's milestones in the fifteen months preceding this article's publication bracket the trajectory of the program. In October 2024, ANKA-3 became the first drone in history to be controlled by another aircraft in a loyal wingman role, a capability demonstration that preceded K-SWARM by more than a year and used a different crewed-uncrewed architecture. In January 2025, ANKA-3 completed the first internal release of a guided glide bomb — a Tolun small diameter weapon dropped from 20,000 feet at 180 knots — proving low-observable strike capability. In March 2025, the aircraft launched from Mürted Air Base, traveled to the Açıkır Test Range with ANKA providing target designation, and released an Aselsan LGK-82 laser-guided weapon from 10 kilometers at 25,000 feet and 200 knots. In December 2025, ANKA-3's 46th sortie validated critical autopilot and autonomous flight envelope tests, closing the basic flight-control validation phase and opening the door to more demanding autonomous mission profiles.

TAI CEO Mehmet Demiroglu confirmed at SAHA Istanbul 2026 that ANKA-3 has completed its critical design review with the production configuration frozen, and that the Turkish Air Force is expected to order more than 50 aircraft in 2026. At the World Defense Show 2026 in February, TAI unveiled a manned-unmanned teaming concept demonstration integrating the TF KAAN fighter with two ANKA-3 drones in a coordinated takeoff, formation maneuvering, and simulated strike sequence — the clearest public statement yet of Türkiye's intent to field a complete CCA ecosystem around its indigenous fifth-generation fighter. TAI stated that the communication, firing, and guidance links between KAAN and ANKA-3 will be operational before KAAN enters Turkish Air Force service.

Aselsan's IVDL: The Digital Spine

Platform capability is a necessary but not sufficient condition for operational CCA. What distinguishes a coherent autonomous air combat architecture from a collection of impressive demonstrators is the command-and-control fabric connecting them. In June 2025, Aselsan CEO Ahmet Akyol confirmed the activation of Türkiye's Indigenous Flight Datalink (IVDL), a wide-bandwidth, high-throughput, electronic-warfare-resistant datalink enabling real-time communication between the TF KAAN fifth-generation fighter and both KIZILELMA and ANKA-3. The IVDL allows KAAN to function not only as a frontline fighter but as an airborne command-and-control node orchestrating multiple unmanned combat assets simultaneously.

In the intended operational architecture, KAAN would direct KIZILELMA and ANKA-3 to conduct suppression of enemy air defenses, electronic attack, and deep-strike missions while remaining at a survivable standoff distance. KIZILELMA — higher speed and more agile — serves as the close-in air combat and strike teammate; ANKA-3 — lower observable and longer-endurance — handles reconnaissance, electronic warfare, and internal-bay precision strike in contested airspace. Complementing the IVDL are low-observable sensor additions to KAAN itself: Aselsan's TOYGUN electro-optical suite and passive KARAT IRST sensor, which reduce KAAN's RF emissions and extend its survivability in advanced integrated air defense environments.

The GCAP Connection: From Ankara to Rome to London

The industrial architecture surrounding Türkiye's CCA programs extends well beyond national borders. At the 2025 Paris Air Show, Leonardo CEO Roberto Cingolani made explicit the link between KIZILELMA and the Global Combat Air Programme (GCAP), the trilateral sixth-generation fighter initiative among Italy, the United Kingdom, and Japan. Cingolani stated that Italy was evaluating KIZILELMA alongside unmanned variants of the M-345 and M-346 as candidate loyal wingman platforms for GCAP, and described the M-346/KIZILELMA combination as a "near-term, exportable teaming model built around existing aircraft rather than waiting for sixth-generation platforms."

The formal industrial vehicle for this integration is LBA Systems, a 50-50 joint venture between Leonardo and Baykar formally launched at the Paris Air Show in June 2025. The venture assigns specific production roles to Italian facilities: Leonardo's Grottaglie plant — currently a composite fuselage manufacturer for the Boeing 787 — will undertake composite manufacturing and final assembly of KIZILELMA. The Ronchi dei Legionari facility will handle TB3 naval-variant final assembly with sensor integration. The former Piaggio Aerospace facility at Villanova d'Albenga, which Baykar acquired by the end of 2024, will assemble TB2 and Akinci platforms. Turin concentrates engineering and European airworthiness certification. Rome hosts a multi-domain innovation center focused on command-and-control, ISR networking, autonomy, and data links.

Under the LBA model — summarized by Leonardo as "Baykar's platforms, Leonardo's systems" — Italian industrial content on KIZILELMA includes LEOSS-T electro-optical payloads, BriteStorm compact EW systems, Osprey AESA radars, Skyward IRST sensors, and European-standard data links. The K-SWARM trials demonstrated a proof of concept for this integration architecture at the system level, validating that Leonardo's avionics, cyber defense, and mission computing can function as the CUC-T command layer above Baykar's Smart Fleet Autonomy substrate. This division of labor — European certification, sensor integration, and command-layer software on top of Turkish autonomous airframe capability — could emerge as a replicable model for other European nations seeking CCA capability below the cost and schedule threshold of a clean-sheet program.

Comparative Context: Where Others Stand

The pace of Turkish CCA development stands in contrast to the timeline pressures facing other programs. The U.S. Air Force's Collaborative Combat Aircraft Increment 1 program, having down-selected to General Atomics' YFQ-42A Gambit and Anduril's YFQ-44A Fury in April 2024, completed first flights of both prototype aircraft in August and October 2025, respectively. Both are now in developmental trials at Nellis Air Force Base. A production downselect is planned for 2026, with Increment 2 awards expected in early fiscal year 2026 targeting additional capability — improved stealth, sensors, and integration with B-21 and E-7 platforms. The USAF plans to spend more than $8.9 billion on CCA programs from fiscal 2025 through 2029, with an initial operational capability goal of approximately 2030. The Netherlands signed a letter of intent for CCA participation at the end of 2025 and ordered its first two aircraft in April 2026, marking the first allied purchase. The U.S. Marine Corps separately selected Northrop Grumman and Kratos in January 2026 to develop its first operational CCA, transitioning the XQ-58 Valkyrie from testbed to operational wingman.

Australia's Boeing MQ-28A Ghost Bat program, funded to AUD 4 billion (approximately USD 2.6 billion) in April 2024, has demonstrated controlled flight and interoperability with the E-7A Wedgetail — a single E-7A operator controlling two in-flight Ghost Bats alongside a third digital Ghost Bat — but has not yet demonstrated live autonomous BVR weapons employment or crewed-uncrewed formation control of the sophistication shown in K-SWARM. China publicly revealed four new CCA prototypes at its September 3, 2025 Victory Day parade, including two fighter-size designs apparently powered by WS-10 or WS-15 class turbofans, but their operational autonomy levels remain unknown to outside observers. Russia's Sukhoi S-70 Okhotnik and Kronshtadt Grom programs continue in some form, but the depth of their autonomy integration has not been publicly demonstrated at the level of the Turkish milestones.

The United Kingdom's Autonomy Contributory Platform program has fielded its first demonstrator, StormShroud, with an initial order of 24 platforms entering service in May 2025 to develop teaming and Combat Cloud integration. A Tranche 2 tender is expected in spring 2026, with contract award between 2027 and 2029. France and Germany continue development under the Future Combat Air System program, while Japan's loyal wingman drone program for the F-X fighter remains at early funding stages. In this global context, Türkiye's combination of live BVR weapons demonstration, autonomous formation flight, crewed-uncrewed command validation, and series production commitment — all achieved within an eighteen-month window — represents a tempo that most programs have not matched at the operational demonstration level.

Risks, Constraints, and Open Questions

Türkiye's CCA advances carry significant industrial and strategic dependencies that independent analysts note as material risks. ANKA-3's current production configuration is powered by a Ukrainian Ivchenko-Progress AI-322 turbofan engine, a supply chain complicated — though not yet broken — by the ongoing conflict in Ukraine. TAI CEO Demiroglu stated at SAHA Istanbul that Ukraine has continued to produce and deliver engines under wartime conditions, and identified the domestically developed TEI TF6000 turbofan as a contingency alternative, with a larger variant possible if needed. The eventual twin-engine, domestically powered ANKA-3 variant designed for supersonic performance alongside KAAN remains at the concept phase; the current priority is completing the single-engine configuration. KIZILELMA faces a parallel propulsion question: Baykar has announced a domestic propulsion development effort in part to reduce dependence on subcontractors unable to match production rate demands, but the timeline for domestically powered KIZILELMA variants has not been publicly confirmed.

TF KAAN itself — the intended apex crewed node for the Turkish CCA architecture — is still maturing. Three flight prototypes are in various stages of ground and airborne testing as of mid-2026, with production aircraft deliveries expected by 2028 to 2029. An indigenous engine program, based on TEI's TF-TEN/TF10000, is in early testing. Until KAAN enters service with the IVDL-linked architecture that Aselsan has described, the full autonomous air combat ecosystem exists in validated segments rather than as an integrated operational force. The ANKA-3 June 2025 prototype crash during the Anatolian Eagle exercise also serves as a reminder that flight-test campaigns for complex autonomy-enabled aircraft carry inherent risk, even as the program's overall trajectory has been strongly positive.

Beyond platform maturity, the doctrinal and regulatory dimensions of autonomous air combat remain contested across all programs. What level of human supervision is required for lethal autonomous engagement decisions — the rules of engagement, certification standards, and liability frameworks — will shape how quickly any of these systems can transition from demonstrations to operational deployment. Turkey's supervised-autonomy approach in K-SWARM, which explicitly preserves human tactical authority over lethal decisions, appears designed in part to pre-empt this challenge.

Strategic Implications

The broader significance of Türkiye's CCA surge extends beyond platform performance. Ankara has used its drone program as an instrument of strategic autonomy — reducing dependence on Western weapons suppliers, generating substantial export revenues, and building industrial leverage with NATO allies — since the TB2's combat debut in Libya and Nagorno-Karabakh. The CCA programs represent an escalation of that strategy into the highest tier of airpower technology, at a moment when Türkiye's relationship with the F-35 program remains severed following its 2019 S-400 purchase from Russia.

The LBA Systems architecture is particularly notable in this context. By embedding KIZILELMA production inside Italy's aerospace industrial base, coupling it to European airworthiness certification, and linking it to GCAP development data, Baykar and Leonardo have created a pathway through which a Turkish-origin UCAV could eventually serve as a loyal wingman for a sixth-generation fighter operated by NATO allies who have no political relationship with Ankara sufficient to justify a bilateral arms purchase. Italy provides the NATO wrapper; Baykar provides the unmanned platform; GCAP provides the operational requirement. Whether KIZILELMA ultimately enters GCAP's CCA selection — a decision not yet made — is secondary to the fact that the industrial and technical foundations for that possibility are now being actively constructed.

For air force planners in Europe, Southeast Asia, and the Gulf, the implication is straightforward: Türkiye's autonomous combat air ecosystem is no longer a future program to be monitored. It is an extant, operationally demonstrable capability being actively exported, industrially embedded in NATO-aligned production networks, and linked to the most significant sixth-generation fighter program outside the United States. The K-SWARM trials are not an endpoint; they are the first public proof of a command architecture that Leonardo has already described as a foundation for future GCAP combat air operations. The next phase of trials, expected to introduce multi-aircraft coordination, sensor tasking, target handoff, and dynamic mission replanning, will determine how rapidly that architecture can evolve from formation control to genuine mission-level autonomy.

Verified Sources and Formal Citations

1. Leonardo S.p.A. Press Release. "Leonardo and Baykar Set Major Milestone for Advanced Crewed/Uncrewed Capability Development with Successful First K-SWARM Live Trials." June 22, 2026.
   https://www.leonardo.com/en/press-release-detail/-/detail/22-06-2026-leonardo-and-baykar-set-major-milestone-for-advanced-crewed-uncrewed-capability-development-with-successful-first-k-swarm-live-trials
2. Army Recognition / Nicanci, Teoman S. "Leonardo and Baykar K-SWARM Trials Show M-346 Evolving into Airborne Command Node for KIZILELMA Unmanned Fighter." June 22, 2026.
   https://www.armyrecognition.com/news/aerospace-news/2026/
3. The Defense News. "Leonardo to Launch Manned-Unmanned Teaming Trials with M-346F and KIZILELMA in 2026." March 23, 2026.
   https://www.thedefensenews.com/news-details/Leonardo-to-Launch-Manned-Unmanned-Teaming-Trials-with-M-346F-and-KIZILELMA-in-2026/
4. Army Recognition. "Leonardo's M-346 Light Attack Fighter to Control Baykar's Two KIZILELMA Combat Drones in Loyal Wingman Trial." March 23, 2026.
   https://www.armyrecognition.com/news/aerospace-news/2026/leonardos-m-346-light-attack-fighter-to-control-baykars-two-kizilelma-combat-drones-in-loyal-wingman-trial
5. Army Recognition. "Türkiye's Kizilelma Unmanned Fighter Executes World-First Beyond Visual Range Air-to-Air Strike." December 1, 2025.
   https://www.armyrecognition.com/news/aerospace-news/2025/tuerkiyes-kizilelma-unmanned-fighter-executes-world-first-beyond-visual-range-air-to-air-strike
6. Interesting Engineering / Young, Chris. "Turkey Stages World's First Autonomous Jet Dogfight in Historic Test." December 30, 2025.
   https://interestingengineering.com/military/worlds-first-autonomous-formation-flight
7. Army Recognition. "Türkiye's Aselsan Links KAAN Stealth Fighter with ANKA-3 and KIZILELMA Drones for Manned-Unmanned Teaming." June 19, 2025.
   https://armyrecognition.com/news/aerospace-news/2025/tuerkiyes-aselsan-links-kaan-stealth-fighter-with-anka-3-and-kizilelma-drones-for-manned-unmanned-teaming
8. Army Recognition. "Türkiye Unveils Autonomous Wingman Concept Linking KAAN Fighter with ANKA III Drones." World Defense Show 2026 coverage, February 2026.
   https://www.armyrecognition.com/archives/archives-defense-exhibitions/2026-archives-news-defense-exhibitions/world-defense-show-2026/
9. Migflug / Defense Analysis. "Turkey Pairs the KAAN With Two Robot Wingmen." World Defense Show 2026.
   https://migflug.com/jetflights/turkey-kaan-anka-iii-manned-unmanned-teaming-world-defense-show-2026/
10. Army Recognition. "Türkiye's ANKA III Flying Wing Stealth Drone Reaches Critical Milestone in Autonomous Capability." December 9, 2025.
    https://www.armyrecognition.com/news/aerospace-news/2025/tuerkiyes-anka-iii-flying-wing-stealth-drone-reaches-critical-milestone-in-autonomous-capability
11. Türkiye Today. "TAI Finalizes ANKA-3 Design; Turkish Air Force to Order Over 50 Aircraft in 2026." January 24, 2026.
    https://www.turkiyetoday.com/nation/tai-finalizes-anka-3-design-turkish-air-force-to-order-over-50-aircraft-in-2026-3213467
12. Air Force Technology. "ANKA III UCAV, Turkey." Updated February 18, 2026.
    https://www.airforce-technology.com/projects/anka-iii-ucav-turkey/
13. Wikipedia. "TAI Anka-3." Accessed June 22, 2026.
    https://en.wikipedia.org/wiki/TAI_Anka-3
14. Wikipedia. "Bayraktar Kızılelma." Accessed June 22, 2026.
    https://en.wikipedia.org/wiki/Bayraktar_K%C4%B1z%C4%B1lelma
15. The Aviationist / Satam, Parth. "LBA Systems to Build TB2, TB3, Akinci and Kizilelma UCAVs in Italy." November 9, 2025.
    https://theaviationist.com/2025/11/09/lba-systems-to-build-ucavs-in-italy/
16. FlightGlobal. "Leonardo-Baykar Joint Venture to Build Kizilelma Fighters and UAVs at Three Italian Plants." November 7, 2025.
    https://www.flightglobal.com/military-uavs/leonardo-to-build-baykar-kizilelma-uncrewed-fighter-at-grottaglie-factory-under-lba-systems-plan/165190.article
17. EDR Magazine. "PAS 2025 — LBA Systems: Leonardo and Baykar Join Forces to Expand Their Footprint on the UAV Market." June 19, 2025.
    https://www.edrmagazine.eu/pas-2025-lba-systems-leonardo-and-baykar-join-forces-to-expand-their-footprint-on-the-uav-market
18. European Security & Defence / Dean, Sidney E. "CCAs, RCs, Loyal Wingmen and Effectors: Developing Unmanned Systems for the Future Air Superiority Team." November 13, 2025.
    https://euro-sd.com/2025/11/articles/technology/47629/ccas-rcs-loyal-wingmen-and-effectors-developing-unmanned-systems-for-the-future-air-superiority-team/
19. Air Force Technology. "Collaborative Combat Aircraft (CCA), US." Updated January 29, 2026.
    https://www.airforce-technology.com/projects/collaborative-combat-aircraft-cca-usa/
20. Wikipedia. "Manned-Unmanned Teaming." Accessed June 22, 2026.
    https://en.wikipedia.org/wiki/Manned-unmanned_teaming
21. Calibre Defence. "The Growing Collaborative Combat Aircraft Marketplace." October 6, 2025.
    https://www.calibredefence.co.uk/the-growing-collaborative-combat-aircraft-marketplace/
22. Army Recognition. "Türkiye Positions ANKA III Stealth Combat Drone for Production as Next-Gen Strike Asset." February 13, 2026.
    https://www.armyrecognition.com/archives/archives-defense-exhibitions/2026-archives-news-defense-exhibitions/world-defense-show-2026/
23. Grey Dynamics. "Anka-3: Demonstrating Turkey's Growing UCAV Expertise." November 29, 2025.
    https://greydynamics.com/anka-3-demonstrating-turkeys-growing-ucav-expertise/
24. TRT World. "Why Türkiye's Homegrown KAAN Fighter Could Reshape Asia's Airpower Calculus." February 26, 2026.
    https://www.trtworld.com/article/8bb2b411da56
25. Defence Security Asia. "Türkiye's KAAN Combat Aircraft Nears 2028 Delivery, Reshaping NATO Air Superiority and Global Defence Force Posture." June 2026.
    https://defencesecurityasia.com/en/kaan-fighter-aircraft-turkiye-fifth-generation-airpower-nato-strategic-deterrence/

KIZILELMA — Key Specifications

Developer	Baykar Technologies
Type	Stealth UCAV / Unmanned Fighter
MTOW	8,500 kg
Payload	1,500 kg
Cruise speed	Mach 0.6
Max speed	Mach 0.9
Combat radius	~500 nm
Ceiling	25,000 ft
Radar	Aselsan MURAD AESA
AAM	TÜBİTAK SAGE Gökdoğan
Maiden flight	14 Dec 2022
Prod. prototype	25 Sep 2024
Production	Begun Oct 2024
Carrier-capable	Yes (STOL; TCG Anadolu)

ANKA-3 — Key Specifications

Developer	TAI (TUSAŞ)
Configuration	Tailless flying wing
MTOW	6,500 kg
Payload	1,600 kg
Max speed	Mach 0.7 / ~425 kt
Ceiling	40,000 ft
Endurance	~10 hr at 30,000 ft
Engine	AI-322 turbofan (current); TEI TF6000/10000 planned
Weapon bays	2 internal + 5 ext. pylons
Maiden flight	28 Dec 2023
CDR	Completed Jan 2026
TuAF order	>50 units expected 2026

Key Milestone Timeline

• Oct 2024ANKA-3 first operated as loyal wingman controlled by another aircraft
• Sep 2024KIZILELMA Prototype 3 (afterburner) first flight
• Oct 2024KIZILELMA mass production begins
• Jan 2025ANKA-3 first internal guided bomb release
• Jun 2025LBA Systems JV launched; Aselsan IVDL activated linking KAAN–KIZILELMA–ANKA-3
• Oct 2025KIZILELMA MURAD AESA radar flight test
• Nov 2025KIZILELMA simulated BVR kill against F-16 (electronic)
• Nov 30, 2025KIZILELMA executes world-first live autonomous BVR air-to-air kill (Gökdoğan/Black Sea)
• Dec 2025ANKA-3 autopilot / autonomous envelope validation (sortie 46)
• Dec 28, 2025World-first fully autonomous close formation flight, two armed jet-powered KIZILELMA
• Jan 2026ANKA-3 CDR complete; production configuration frozen
• Feb 2026TAI unveils KAAN + 2× ANKA-3 teaming at World Defense Show 2026
• May 2026K-SWARM live trials at Çorlu: M-346 commands KIZILELMA in autonomous formation
• May 2026Indonesia signs first KIZILELMA export agreement (12 aircraft, options to 60)
• Jun 22, 2026K-SWARM results publicly announced by Leonardo and Baykar

LBA Systems Production Network

Grottaglie (Taranto): Composite manufacturing & final assembly — KIZILELMA

Ronchi dei Legionari (Trieste): TB3 naval final assembly, sensor/weapon integration

Villanova d'Albenga (Genova): TB2 & Akinci assembly (Piaggio Aerospace facility)

Turin: Engineering & European certification

Rome: Multi-domain C2, ISR networking & autonomy innovation

Structure: 50-50 JV, headquarters Italy. "Baykar's platforms, Leonardo's systems." CEO: Francesco Sabatini (Leonardo). Chairman: Haluk Bayraktar (Baykar).

Global CCA Program Comparisons

US USAF YFQ-42A / YFQ-44A: First flights Aug–Oct 2025; downselect expected 2026; IOC ~2030; $8.9B budget FY25–29.

US USMC (Kratos/Northrop): Selected Jan 2026; XQ-58 transitioning to operational CCA role.

Australia MQ-28A Ghost Bat: AUD 4B funded; E-7A control demonstrated; no live autonomous BVR kill reported.

UK ACP StormShroud: Initial 24 units in service May 2025; Tranche 2 tender expected spring 2026.

China: Four CCA prototypes displayed Sep 2025 parade; autonomy levels unverified externally.

Russia S-70 Okhotnik / Grom: Development ongoing; no confirmed autonomous BVR demonstration.

Aviation Week & Space Technology  ·  Analytical / Staff Report  ·  June 22, 2026  ·  All rights reserved

 

 

AFRL "Flyer" — Deep Dive:

Aerospace & Defense · Supercomputing · 22 June 2026 — Technical Follow-On

Analysis · Supply Chain · Global HPC Competition · Operations

Who Built It, What's Inside, Where It Stands Against China, and How the Machine Gets Shared

Four questions the ribbon-cutting left unanswered: the contractor and chip supply chain behind TI-23; Flyer's true rank in the global supercomputer order—including China's secret exascale machines; what 8.7 petaflops actually buys for hypersonic aerothermal simulation; and the bureaucratic and technical machinery that decides which research teams get core-hours and when.

Aerospace & Defense / HPC Technology 22 June 2026

BLUF — Bottom Line Up Front

Penguin Solutions (a subsidiary of Smart Global Holdings, formerly Penguin Computing) is the most likely integrator of the TI-23 Flyer and Raven systems, having won the $68 million DoD HPCMP contract in September 2021 for exactly two platforms at AFRL and Navy DSRCs. NVIDIA GPU accelerators are confirmed; AMD CPUs are the precedent architecture. Globally, Flyer's 8.7 petaflops places it far outside the top 100 publicly ranked machines — yet China's most capable systems (Tianhe-3 "Xingyi" at ~1.3 exaflops sustained and Sunway OceanLight at comparable scale) are deliberately withheld from TOP500 rankings, complicating direct comparison. For hypersonic CFD, Flyer's GPU-dense architecture targets turbulence-transition physics and aerothermal coupled-field problems that previously required months of wall-clock time on older AFRL clusters; individual simulation campaigns that consumed six months on predecessor hardware have been cut to three weeks. Project access is governed by a five-tier allocation hierarchy — from standard batch through Frontier multi-year dedicated-priority campaigns — scheduled via Slurm under fair-share priority rules that weight urgency, project tier, and historical utilization.

I. Who Built Flyer? The Prime Contractor and Supply Chain

AFRL has not publicly named the integrator of TI-23 at the June 2026 ribbon-cutting — a routine omission for DoD procurement, where vendor identity and system architecture below the top-level headline figures are often withheld on acquisition sensitivity grounds. However, the public record is sufficiently specific to identify the most probable prime with high confidence.

In September 2021, the DoD High Performance Computing Modernization Program awarded Penguin Computing — now operating as Penguin Solutions, a subsidiary of Smart Global Holdings (SGH) — two contracts collectively valued at $68 million. Those contracts called for delivery of Penguin TrueHPC supercomputing platforms, plus managed services and high-performance storage, to exactly two sites: the Air Force Research Laboratory DSRC and the Navy DSRC. The award predates the formal TI-23 program designation, but the number and location of the contracted systems precisely matches the TI-23 Flyer (AFRL, unclassified) and TI-23 Raven (AFRL, classified) pair. Penguin's own president, Sid Mair, confirmed at contract award that the platforms would be used for "the highly complex problems the user community is tasked with solving" and cited AMD and NVIDIA partnerships as the enabling semiconductor relationship.

TI-23 Probable Supply Chain — Based on Public Procurement Record

• System Integrator (Prime)
• Penguin Solutions
• Subsidiary of Smart Global Holdings (SGH); formerly Penguin Computing. Winner of $68M DoD HPCMP contract (Sep 2021) for two TrueHPC platforms at AFRL and Navy DSRCs. Penguin has offered HPC solutions since 1998; acquired by SGH (Cayman Islands-based) for $85M in 2019.
• GPU Accelerator (Confirmed)
• NVIDIA Corporation
• NVIDIA GPU model unspecified at ribbon-cutting. Brig. Gen. Wickert cited "NVIDIA chips and GPUs" explicitly. Penguin's TrueHPC platforms for DoD have used NVIDIA H100 or A100-class datacenter GPUs in recent configurations. Export-controlled; not available to Chinese or sanctioned entities.
• CPU (Precedent Architecture)
• AMD
• Raider (TI-21, the immediately preceding AFRL system) used AMD CPUs + NVIDIA GPUs per 2022 AFRL DSRC disclosure. The TI-23 architecture is unconfirmed at the CPU level; AMD EPYC is standard in Penguin TrueHPC configurations. TI-23 may also follow the full AMD + NVIDIA GPU-accelerated pattern used in El Capitan and Frontier.
• High-Performance Storage
• Likely DDN or equivalent
• 18 PB storage capacity confirmed. Previous AFRL DSRC generations (HPE SGI 8600 era) used DDN Lustre parallel storage; Penguin TrueHPC configurations typically offer DDN or IBM Spectrum Scale. Specific vendor not publicly disclosed for TI-23.
• Interconnect Fabric
• InfiniBand (likely) or HPE Slingshot
• Not publicly disclosed. Raider used Penguin's integrated fabric. At 186,000 processors, a high-bandwidth low-latency interconnect (NDR InfiniBand at 400 Gbit/s per port, or HPE Slingshot) is required to prevent I/O bottlenecks in coupled-field hypersonic simulations.
• Network Backbone (DREN)
• Verizon
• DREN 4, the DoD Research and Engineering Network connecting all five DSRCs, is operated under a commercial services contract awarded to Verizon in 2021. Supports 1–100 Gbit/s data transfer; IPv6-native with IPv4 compatibility. More than 210 DoD sites connected.
• Managed Services
• Penguin Solutions
• Included in 2021 contract. Bryon Foster (AFRL DSRC director) confirmed 24/7/365 operational commitment; Penguin's managed services team provides system administration, user support, and continuous operation assurance for DoD HPCMP centers.

One earlier generation of AFRL DSRC systems — a set of four SGI 8600 clusters commissioned before the Raider era — was supplied by Hewlett Packard Enterprise (HPE) and powered by 24-core Intel Xeon Scalable processors with Intel Omni-Path fabric and DDN Lustre storage at 12 petabytes per system. That HPE/Intel architecture has given way to the AMD CPU + NVIDIA GPU architecture across recent DoD HPCMP insertions, consistent with the broader industry migration toward heterogeneous compute for mixed HPC/AI workloads. Flyer's NVIDIA GPU content is the clearest signal of this architectural shift: the same class of datacenter GPU that trains large language models at scale is now the core computational substrate for aerothermal shock physics at the Air Force Research Laboratory.

II. Where Flyer Stands in the Global Supercomputer Order

Benchmarked at 8.7 petaflops, Flyer does not approach the upper tier of the current global TOP500. As of the June 2025 TOP500 list, the entry threshold (rank 500) stood at 2.57 petaflops — meaning Flyer would rank somewhere in the vicinity of the 200th to 250th position worldwide, a mid-table placement. The three leading U.S. Department of Energy machines occupy a different computational order of magnitude entirely: El Capitan at Lawrence Livermore National Laboratory tops the list at 1,742 petaflops (1.742 exaflops), Frontier at Oak Ridge at 1,353 petaflops, and Aurora at Argonne at 1,012 petaflops. Europe's JUPITER system at Forschungszentrum Jülich in Germany, inaugurated in September 2025, became the continent's first exascale-class system at 1,000 petaflops.

Global Supercomputer Landscape — Selected Systems, June 2026

System	Country / Operator	Peak (Rmax)	Architecture	TOP500 Rank	Primary Mission
El Capitan	USA / LLNL	1,742 PFlop/s	HPE Cray EX, AMD EPYC + MI300A	#1	Nuclear stockpile stewardship
Frontier	USA / ORNL	1,353 PFlop/s	HPE Cray EX235a, AMD EPYC + MI250X	#2	DOE open science
Aurora	USA / Argonne NL	1,012 PFlop/s	HPE Cray EX, Intel Max Series	#3	AI + science
JUPITER Booster	Germany / FZ Jülich	~1,000 PFlop/s	NVIDIA GH200 (Grace Hopper)	#4	European research
Tianhe-3 "Xingyi"	China / NUDT Tianjin	~1,300–1,700 PFlop/s (est.)	Phytium CPUs + Matrix-3000 (domestic)	Not submitted	Defense/AI — classified
Sunway OceanLight	China / NSC Wuxi	~1,000+ PFlop/s (est.)	Sunway Microelectronics (domestic)	Not submitted	Ocean modeling / defense
Sunway TaihuLight	China / NSC Wuxi	93 PFlop/s	Domestic SW26010 manycore	#24 (Nov 2025)	General research
Flyer (TI-23)	USA / AFRL WPAFB	8.7 PFlop/s	Penguin TrueHPC, NVIDIA GPUs, AMD CPUs (prob.)	~200–250 (est.)	Defense RDT&E (unclassified)
Raven (TI-23)	USA / AFRL WPAFB	~5–6 PFlop/s (est.)	Penguin TrueHPC (classified config.)	Not submitted	Defense RDT&E (classified)
Raider (TI-21)	USA / AFRL WPAFB	12 PFlop/s	Penguin, AMD + NVIDIA	~50 (at commission, 2023)	Defense RDT&E (predecessor)

* Flyer's TOP500 rank is an estimate; AFRL DSRC systems are not routinely submitted to the TOP500 list. Raven specifications are estimated from the combined 14 PFlop/s target minus Flyer's 8.7 PFlop/s. Chinese exascale figures reflect unverified intelligence community and open-source assessments; neither system has been formally benchmarked on HPL. Sources: TOP500.org, Data Center Dynamics, Tom's Hardware, Wikipedia, NextPlatform.

The China comparison requires careful handling. On the public TOP500 list, China's highest-ranked submission as of November 2025 is the 2016-vintage Sunway TaihuLight at 93 petaflops — ranking 24th, more than ten times slower than Flyer's own predecessor, Raider. This number is almost universally acknowledged in the HPC community as a profound underrepresentation of China's actual computational capability. Since 2017, Chinese institutions have progressively withdrawn new systems from TOP500 submissions, offering Gordon Bell Prize competition papers as the primary evidence of continued development instead.

By most credible open-source estimates, China operates at least two exascale-class systems. Tianhe-3 (nicknamed "Xingyi"), developed by the National University of Defense Technology and housed at the National Supercomputing Center in Tianjin, has been estimated at approximately 1.3 exaflops sustained (1.7 exaflops peak) on HPL-equivalent benchmarks — performance that would place it between Frontier and El Capitan if submitted to TOP500. The Sunway OceanLight at the National Supercomputing Center in Wuxi is estimated at comparable throughput. A third exascale-class system, allegedly built by China's Sugon (now operating under U.S. entity-list restrictions that blocked its access to AMD Hygon CPUs) and housed at the Shenzhen supercomputing center, has an uncertain status. China's Ministry of Industry and Information Technology stated in mid-2024 that China's aggregate installed computing capacity across all systems had reached approximately 230 exaflops of theoretical peak — a figure that, taken at face value, represents aggregate capacity of the entire government, commercial, and research sectors, not individual machine peak performance.

"China's OceanLight system … at least seems to aspire to exaflop-sized performance — albeit one that remains inscrutable to international standards." NextPlatform, February 2024, citing HPC community analysis

The geopolitical dimension of this opacity is significant: U.S. export controls imposed since 2019 have systematically denied China access to advanced NVIDIA H100 and A100 GPUs — the same accelerators almost certainly powering Flyer — and have restricted Huawei and Sugon from U.S.-origin technology. China's response has been to develop indigenous alternatives, including Huawei's Ascend series and domestic interconnect fabrics, at considerable cost in both money and performance parity. The Tianhe-3 and OceanLight systems are built on entirely domestic silicon from Phytium (CPUs) and Sunway Microelectronics respectively — a deliberate supply-chain independence that trades performance efficiency for invulnerability to U.S. sanctions. From AFRL's operational standpoint, the direct performance comparison is less strategically relevant than the access-and-security envelope: Flyer operates within a classified-adjacent, export-controlled, security-vetted infrastructure that no Chinese entity can access, while Chinese defense supercomputing operates in a comparable walled garden that U.S. researchers cannot reach.

III. What 8.7 Petaflops Actually Buys for Hypersonic CFD

The engineering claim that matters most for AFRL's mission is not where Flyer ranks on TOP500 but what it enables in hypersonic research that predecessor systems could not accomplish on acceptable timescales. The physics problem is instructive.

A hypersonic vehicle traveling at Mach 5 or above generates a shock-layer environment in which the flow chemistry, heat transfer, and boundary-layer dynamics are tightly coupled across multiple physical phenomena: high-temperature gas dissociation and ionization, turbulent boundary-layer transition (which determines transition from laminar to turbulent heat flux — a factor of three to seven in heating rate), ablation and oxidation of thermal protection materials, and plasma sheath formation that attenuates radio-frequency communications. Each of these phenomena operates on different length and time scales, and their accurate coupled representation requires fine spatial resolution that scales computational cost as roughly the cube of the linear resolution factor. Halving the grid spacing in all three dimensions multiplies computational cost by approximately eight. Running to higher Mach numbers, longer bodies, or more complex geometries compounds this further.

AFRL's own documentation for its predecessor Raider system (TI-21) illustrated the practical consequence: a Navy simulation project that previously required six months of continuous compute time on the available DSRC hardware was cut to three weeks on Raider's 12-petaflop architecture — a factor of roughly eight in elapsed time. The Flyer-Raven TI-23 combination offers approximately 1.17 times Raider's total performance in direct petaflop terms (14 combined versus 12), but GPU-accelerated HPC does not scale linearly with headline petaflop count for all workload types. The key enabler is the NVIDIA GPU's performance on the specific linear algebra kernels that dominate CFD solvers: sparse matrix-vector products, fast Fourier transforms for spectral methods, and the iterative linear solvers (GMRES, multigrid) that consume most clock cycles in Reynolds-Averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES) codes.

AFRL's computational aeroscience community uses several DoD HPCMP-maintained and community-developed codes for hypersonic work: Kestrel (the HPCMP CREATE Air Vehicles code), OVERFLOW (NASA/AFRL), and US3D (University of Minnesota, developed under AFRL-funded research). These codes have been progressively GPU-ported over the past five years, and all three benefit directly from Flyer's NVIDIA GPU density. Purdue University's AFRL-sponsored hypersonic turbulence research — which uses HPCMP computing allocations to design passive surface treatments for boundary-layer transition delay — provides a concrete example: computational simulation campaigns design the surface geometry, physical tunnel tests at AFRL's Mach 6 Ludwieg tube validate the results, and the agreement between prediction and experiment determines whether the next simulation campaign is warranted. The HPCMP Frontier ceramics program extended this paradigm to materials discovery, predicting approximately 900 candidate thermal protection compounds — a screening campaign that would have been physically infeasible through bench synthesis.

At 8.7 petaflops with GPU acceleration, Flyer enables LES campaigns on full-vehicle hypersonic geometries that were previously limited to reduced-geometry or 2D-symmetric configurations on RANS solvers at lower fidelity. The practical implication: aerothermal predictions at the nose cap, control surface junction, and inlet throat of a hypersonic glide vehicle can now be run in high-fidelity LES in days rather than months, enabling design-space exploration that was previously incompatible with acquisition program timelines.

IV. Operations: How Projects Are Scheduled and Prioritized

An Engineers Guide to Scheduling a CFD job on Flyer. Flyer does not operate as a dedicated single-purpose machine for any one research program. Like all HPCMP DSRC systems, it is a shared national resource serving hundreds of concurrent projects across the services, agencies, and sponsored civilian and contractor researchers. The allocation and scheduling architecture that governs who gets core-hours, when, and at what priority is a two-layer system: a resource allocation governance layer that determines how many core-hours each project receives per fiscal year, and a job scheduling layer that determines when allocated compute jobs actually run.

At the governance layer, the HPCMP's Resource Management team collects and manages computational requirements through the Portal to the Information Environment (pIE), coordinated with Service/Agency Approval Authorities (S/AAA) — the branch-level gatekeepers who validate that a proposed project aligns with DoD mission priorities and that the requesting organization has appropriate clearances and DoD sponsorship. All users must have a DoD government scientist or engineer as a sponsor; principal investigators may be from government, industry, or academia. The S/AAA approval is required before any core-hour allocation is granted.

Projects are allocated resources in five general tiers, each with distinct proposal requirements, priority treatment in the scheduler, and compute quantum available:

● URGENT

Urgent Projects

DoD HPCMP-designated. Bypass normal queue ordering; immediate resource claim. Reserved for time-critical operational or acquisition crises where delay constitutes unacceptable mission risk. Requires HPCMP Director-level approval.

● HIGH PRIORITY

High Priority Projects

Pre-authorized elevated queue access. Jobs in the "high" Slurm queue jump ahead of standard allocations. Typically reserved for acquisition program milestones, test and evaluation support, or Congressional-directed priorities.

● FRONTIER

Frontier Projects

Multi-year commitments of exceptional compute allocations — resources that would not be achievable through standard channels. Competitive proposal process, reviewed by the High Performance Computing Advisory Panel (HPCAP). Eligible PIs include government, industry, and academia with DoD sponsor. Dedicated Slurm queue partition.

● AE PROGRAM

Acquisition Engineering

Dedicated track for acquisition engineering, mission engineering, and T&E workflows tied to DoD programs of record. Access to compute resources, DREN network access, and specialized modeling/simulation software tools (CREATE suite). Managed separately from S&T allocations.

● STANDARD

Standard Allocations

The baseline tier. Annual core-hour allocations via S/AAA approval. Jobs submitted to standard, large, background, or debug Slurm partitions with fair-share scheduling. Background queue allows low-priority opportunistic jobs (up to 10 concurrent per user); debug queue is limited to short verification runs.

At the job scheduling layer, Flyer (and Raider before it) uses the Slurm Workload Manager — the same scheduler that powers more than 60 percent of TOP500 systems globally, including the Tianhe-2 in China. Slurm manages the assignment of specific compute nodes to queued jobs based on resource availability, time limits, and priority scores. Priority is computed from a combination of factors: project allocation tier (Urgent/High/Frontier jobs receive favorable treatment), fair-share weighting (teams that have recently consumed large fractions of their allocation receive lower instantaneous priority to equalize system access over time), job size and wall-clock time limits, and queue-specific policies set by AFRL DSRC system administrators.

The HPCMP additionally offers Dedicated Support Partitions (DSPs) — reserved node sets held for specific programs that require guaranteed turnaround, analogous to a dedicated laboratory instrument that cannot be preempted by other users. DSPs are a premium resource available only to programs that have justified the need for guaranteed access windows, such as large-scale time-sensitive test support campaigns where computational results must be available before a physical test event proceeds.

For Frontier-class projects — the largest and most computationally demanding campaigns — the proposal process is formally competitive. Proposals are reviewed by the High Performance Computing Advisory Panel (HPCAP), a board of senior DoD computational science and engineering experts, and must demonstrate both scientific or engineering rigor and a clear connection to DoD acquisition or S&T mission outcomes. Multi-year commitments are evaluated at annual Intermediate Program Reviews (IPRs), and continued funding is contingent on demonstrated progress. The HPCMP Institute program, a parallel track, funds software development efforts that produce deployable tools for the wider HPCMP user community — CREATE-class codes and similar deliverables — and imposes the same IPR discipline with weekly activity reports and monthly financial reporting.

"Sometimes even before the system is running, we're ordering the next one. These systems are so large, and it takes so much to build them, get them in the building and up and running that there is a continual process." Bryon Foster, Director, AFRL DoD Supercomputing Resource Center (2023, discussing the TI-23 procurement)

The operational rhythm that results is one of perpetual queuing under a priority framework that mirrors the broader DoD funding hierarchy: programs of record with acquisition milestones get AE or High queue access; large multi-year S&T campaigns compete for Frontier status; smaller research efforts and academic partners operate in the standard tier under fair-share rules that prevent any single group from monopolizing the system. The 24/7/365 operational commitment means Flyer's 186,000 processors are never intentionally idle — unused cycles are consumed by lower-priority background queue jobs rather than allowed to go to waste. Over a projected five-year service life, this continuous-utilization model underpins the $800 million lifetime savings figure: every hour of simulation delivered is an hour of physical testing avoided.

V. The Supply Chain Vulnerability Underneath Flyer's Capability

A final dimension of Flyer's significance that received no attention at the ribbon-cutting but is implicit in its architecture: the entire system depends on NVIDIA GPU silicon that is manufactured, at leading edge, at TSMC in Taiwan — and is now subject to the same export control architecture that restricts Chinese access to that same silicon. The U.S. maintains its lead in deployable GPU-accelerated supercomputing precisely because NVIDIA H100 and successor chips cannot be legally exported to China, and because China's domestic accelerator industry (Huawei Ascend, Biren, Cambricon) has not yet closed the performance gap at scale. AFRL's Flyer is therefore simultaneously a national defense capability and a proof point in an ongoing industrial policy contest: the U.S. bet that maintaining NVIDIA export restrictions and investing in domestic GPU-dense HPC infrastructure compounds American advantage in simulation-based weapons development faster than China can build comparable capability from domestically produced silicon.

Whether that bet holds depends on China's pace of domestic semiconductor development — a question that has as much bearing on AFRL's next supercomputer procurement as any doctrinal priority or budget line. From Wright-Patterson's secure facility on Area B, Flyer's NVIDIA GPU nodes run a race that began not in Dayton but in the foundry halls of Hsinchu and the national supercomputing centers of Tianjin and Wuxi.

Verified Sources & Citations

1. "Penguin Computing Awarded $68M to Provide HPC Capabilities to DoD." HPCwire, 14 September 2021.
   https://www.hpcwire.com/off-the-wire/penguin-computing-awarded-68m-to-provide-hpc-capabilities-to-dod/
2. "DoD Awards $68M Contract to Penguin Computing for Two Supercomputers." Data Center Dynamics, 15 September 2021.
   https://www.datacenterdynamics.com/en/news/dod-awards-68m-contract-to-penguin-computing-for-two-supercomputers/
3. "Air Force Research Lab Adds 12PFLOPS HPC System." Inside HPC & AI News, 25 September 2023.
   https://insidehpc.com/2023/09/air-force-research-lab-adds-12pflops-penguin-hpc-system/
4. "HPE to Deliver Seven Supercomputers to Department of Defense." TOP500.org.
   https://www.top500.org/news/hpe-to-deliver-seven-supercomputers-to-department-of-defense/
5. "TOP500 List — June 2025." TOP500.org, June 2025.
   https://top500.org/lists/top500/list/2025/06/
6. "TOP500." Wikipedia, updated May 2026. (Aggregate country performance figures; El Capitan, Frontier, Aurora rankings.)
   https://en.wikipedia.org/wiki/TOP500
7. Balci, Mete. "State of Supercomputers Around the World in 2026." metebalci.com, May 2026. (Four exascale systems; JUPITER; Chinese non-submissions.)
   https://metebalci.com/blog/state-of-supercomputers-around-the-world-in-2026/
8. Sanchez, Christopher. "State of Global AI Compute (2025 Edition)." sanchez.vc GeoCoded Special Report, August 2025. (US 6.696 EFlop/s aggregate; Chinese capacity claims; Epoch AI/Georgetown data.)
   https://www.sanchez.vc/geocoded-special-reports/state-of-global-ai-compute-2025-edition
9. "China Publishes List of Its Most Powerful Supercomputers, with No Exascale Systems to Be Found." Data Center Dynamics, May 2026. (487.94 PFlop/s top Chinese public system; 63 TOP500 submissions.)
   https://www.datacenterdynamics.com/en/news/china-publishes-list-of-its-most-powerful-supercomputers-with-no-exascale-systems-to-be-found/
10. Hruska, Joel. "China May Have Unmatched Supercomputer Abilities, Third Exascale Machine Apparently Online." Tom's Hardware, September 2023. (OceanLight, Tianhe-3 "Xingyi," Sugon/Shenzhen status.)
    https://www.tomshardware.com/news/industry-expert-chinas-supercomputer-might-may-be-unmatched
11. "Two Chinese Supercomputers Break Exascale Barrier." Tom's Hardware, October 2021. (Tianhe-3/Phytium + Matrix-3000; OceanLight/Sunway Microelectronics architecture.)
    https://www.tomshardware.com/news/two-chinese-exascale-supercomputers
12. "The Mystery of Tianhe-3, The World's Fastest Supercomputer, Solved?" NextPlatform, February 2024. (Architectural analysis of Xingyi; comparison to El Capitan.)
    https://www.nextplatform.com/2024/02/09/the-mystery-of-tianhe-3-the-worlds-fastest-supercomputer-solved/
13. "More Details Slip Out About China's Two Secret Exascale Supercomputers." Data Center Dynamics, November 2021. (Tianhe-3: 1.3 exaflops HPL sustained, 1.7 peak.)
    https://www.datacenterdynamics.com/en/news/more-details-slip-out-about-chinas-two-secret-exascale-supercomputers-third-may-be-delayed/
14. "Supercomputing in China." Wikipedia, updated June 2026. (Exascale systems; TOP500 withdrawal; domestic silicon.)
    https://en.wikipedia.org/wiki/Supercomputing_in_China
15. "Exascale Computing." Wikipedia, updated June 2026. (Tianhe-3 and OceanLight operational as of 2023; JUPITER 2025.)
    https://en.wikipedia.org/wiki/Exascale_computing
16. Choi, Charles Q. "Top500: Frontier Still No. 1. Where's China?" IEEE Spectrum. (OceanLight citation; Chinese HPC opacity analysis.)
    https://spectrum.ieee.org/frontier-exascale-top500-export-controls
17. DoD HPCMP Resource Management. "Resource Management — Allocation Processes." hpc.mil. (pIE portal; S/AAA approval; Frontier and DSP programs.)
    https://www.hpc.mil/solution-areas/resource-management
18. DoD HPCMP. "Call for FY 2025 DoD Frontier Project Proposals." hpc.mil. (Frontier eligibility, purpose, multi-year commitment structure.)
    https://www.hpc.mil/calls/call-for-fy-2025-dod-frontier-project-proposals
19. DoD HPCMP. "Call for DoD HPCMP Acquisition Engineering Project Requests." hpc.mil. (AE program structure, lifecycle phase coverage.)
    https://hpc.mil/calls/call-for-dod-hpcmp-acquisition-engineering-project-requests
20. "Raider Slurm Guide." centers.hpc.mil / AFRL DSRC, updated April 2026. (Slurm queues: urgent, high, frontier, standard, large, background, debug; queue limits and priority rules.)
    https://centers.hpc.mil/users/docs/afrl/raiderSlurmGuide.html
21. AFRL Regional Network / Air Force Tech Connect. "Supercomputer Allocation Enables Purdue Hypersonic Research." Air Force Tech Connect, February 2024. (Mach 6 Ludwieg tube; turbulence transition; HPCMP allocation enabling validation experiments.)
    https://airforcetechconnect.org/news/supercomputer-allocation-enables-purdue-hypersonic-research
22. DoD HPCMP. "Call for FY23 DoD HPCMP Institute Proposals." hpc.mil. (Institute structure; weekly/monthly/annual reporting; IPR milestones; software deliverable requirements.)
    https://www.hpc.mil/calls/call-for-fy23-dod-hpcmp-institute-proposals

Prepared in the style of IEEE Spectrum as a technical follow-on companion to the 22 June 2026 AFRL Flyer commissioning report. Penguin Solutions prime contractor identification is based on the 2021 public contract record and is inferred, not confirmed, for TI-23. Chinese exascale performance figures are open-source assessments; neither Tianhe-3 nor OceanLight has been publicly benchmarked on standard HPL metrics. All Slurm queue parameters reflect the documented Raider (TI-21) configuration at centers.hpc.mil and are presumed to be substantially replicated on Flyer (TI-23), but Flyer-specific queue documentation had not been published as of this writing.