The USAF’s Collaborative Combat Aircraft (CCA) aims to field at least 1,000 unmanned platforms to complement the F-35 and the Next Generation Air Dominance (NGAD) platform. Between two to eight CCAs would accompany each fighter, improving their survivability and situational awareness by acting as decoys as well as sensor and weapon platforms. Increment 1 is expected to carry Raytheon AIM-120s in the air-to-air role. More broadly, the CCA program has been marketed as a means to achieve “affordable mass”, cultivate the industrial base, and lower service sustainment costs. The Air Force has ambitiously outlined a flyaway unit cost goal of between 1/4 and 1/3 that of the F-35 (equivalent to $25M-$33M in FY25). The service had expected to field least three increments of CCAs two-year cycles starting in FY28. Source selection for Increment 1 is expected in FY26. The program is expected to be reviewed by the Trump Administration.

Program History

Key Highlights:

• The U.S. has funded multiple generations of UCAV demonstrators but has struggled to translate these concepts into operational platforms over three decades
  • The U.S. Air Force in particular has vacillated across the cost-capability spectrum from exquisite to attritable designs
• Autonomy has remained the greatest technological barrier to fielding UCAVs in contested environments
  • Major advancements in AI control agents began around 2015
• The contemporary U.S. market for CCAs has shifted towards attritable platforms. The Air Force effort is being shaped by three main drivers: (1) resourcing constraints from the $63+ billion (in FY20 dollars) cost overrun from the Sentinel ICBM, (2) a desire to broaden the industrial base to encompass non-traditional suppliers, and (3) Kendall’s atypical role as a service secretary
  • The USAF has stated its priority for Increment 1 is acquisition speed, followed by cost and “minimum viable capability”

The CCA program is arguably the culmination of several truncated unmanned aerial vehicle (UAV) programs over the course of three decades. Broadly these programs helped to mature new composite aerostructures and low observable (LO) design techniques. Many of these programs sought to field at least semi-autonomous control systems. As these technologies matured, the USAF has struggled to identify its optimal cost-capability mix between attritable and exquisite systems. The service initially favored exquisite systems from the 1990s into the early 2010s, then preferred attritable systems from the mid-2010s to the early 2020s before appearing to pivot back toward higher-end systems during the early part of Air Force Secretary Kendall’s tenure. As of the time of this writing, the Air Force now favors attritable systems with the development of CCA Increment 1.   

1998-2014, UCAV Pre-History & Evolution 

Unmanned Combat Aerial Vehicle (UCAV) demonstrators & CCAs. Credit: Aviation Week.

The Unmanned Combat Aerial Vehicle (UCAV) and equivalent Navy program (UCAV-N) described below were envisioned to have dynamic mission replanning and coordinated multi-vehicle control systems. For example, formations of UCAVs could detect radar sites and re-route in-flight to minimize the risk of detection. The USAF conceptualized Command and Control (C2) of its UCAVs from manned aircraft such as the proposed multi-mission E-10 C4ISR platform or the back-seater of a fighter aircraft. For the subsequent X-47B, Northrop planned to have one ship-based pilot control up to five aircraft. In contrast, contemporary CCA concepts would be directed from forward deployed single-seat fighter aircraft. 

The DoD’s UAV Roadmap 2002 theorized, “increased onboard processing will be the key enabler of more responsive flight control systems, onboard sensor data processing, and autonomous operations (AO) for future UAVs… Scalable levels of AO will probably be necessary to accommodate varying rules of engagement (ROEs) for contingencies ranging from peacekeeping to force on force.”

More broadly, processing power has been the foundational enabler of image recognition algorithms within the past ten years and language learning models within the past five years. As of the time of this writing, autonomy arguably remains the most significant technical barrier to fielding CCAs.

Credit: UAV Roadmap 2002. The Unmanned Combat Armed Rotorcraft (UCAR) was a short-lived U.S. Army program to explore an unmanned replacement for the RAH-66.

USAF-DARPA Unmanned Combat Aerial Vehicle (UCAV)

The UCAV program sought to explore unmanned Suppression of Enemy Air Defense (SEAD) concepts at 1/3 the cost of the F-35 with a notional initial operational capability (IOC) around 2010. System requirements included 8,000 lbs. empty weight and a payload of 2,000 lbs. or less. Pilots would control up to six UCAVs. The system was expected to have a combat radius of 650 nautical miles (nm) with a loiter of 30 minutes on target.

The program began in 1998. Lockheed Martin withdrew but Boeing, Raytheon, and Northrop Grumman competed for the project. In March 1999, Boeing was awarded a $110 million ($214 million in FY25) development contract for two flying demonstrators. The company also contributed $21 million ($41 million in FY25) of its own funds.

The first X-45A demonstrator first flew on May 22, 2002. The pair of aircraft flew a total of 64 flights, amassing 63.4 flight hrs. through to August 10, 2005. The goal of the flight program was to validate multi-ship operations for eventual SEAD/DEAD operations before the development of the larger and more capable X-45B.

USN-DARPA Unmanned Combat Aerial Vehicle, Navy (UCAV-N/N-UCAV)

The Navy anticipated a need for an unmanned intelligence, surveillance, and reconnaissance (ISR), airborne surveillance, strike, and SEAD/DEAD platform for its future carrier air wings (CVW). In June 2000, Boeing and Northrop Grumman were each awarded $2 million to refine their concepts over a period of 15 months. Goals for the program included a 6,000-12,000 lbs. empty weight, payload greater than 2,000 lbs. (up to 4,000 lbs.), strike mission endurance of 6 hrs., ISR endurance of 9-12 hrs., load factor of 6gs, and Very Low Observable (VLO) signature. The final aircraft was expected to have an unfolded wingspan potentially greater than 60 ft., but smaller than 67 ft. for carrier suitability. Proposed sensor packages included a synthetic aperture radar and conformal UHF-band antennas at the same frequency as the E-2C.

In July 2000, Northrop Grumman decided to internally fund its X-47A demonstrator. Scaled Composites completed the aircraft a year later and the aircraft first flew on Feb. 23, 2003. The aircraft’s distinct kite planform was used for its inherent LO properties and its long fuselage length provided ample room for weapons and fuel at the cost of aerodynamic efficiency. As early as 2001, Northrop Grumman was already exploring future cranked-kite concepts to meet the anticipated 12-hour endurance ISR requirement.     

USAF-USN Joint Unmanned Combat Air Systems (J-UCAS)

In December 2002, the DoD consolidated the Air Force’s separate UCAV program with the UCAV-N to form the Joint Unmanned Combat Air Systems (J-UCAS) program. The J-UCAS would perform strike, ISR, EW, and SEAD missions. Expanding mission requirements drove the initial low-cost concept to a 90,000 lbs. class maximum takeoff weight (MTOW) UCAV.

DARPA awarded Northrop Grumman a $1.04 billion ($1.67 billion in FY25) contract on August 19, 2004, to build three full-scale X-47B demonstrators over five years. DARPA awarded Boeing a similar $767 million ($1.23 billion in FY25) for three X-45Cs on Oct. 12, 2004.

The 2006 Quadrennial Defense Review (QDR) canceled J-UCAS and instructed the Air Force to develop a new bomber and the Navy to acquire a long-range UAV to expand the reach of U.S. carrier groups.

X-45B, X-45C & Phantom Ray

The X-45 was intended to be developed in “spirals”, with the larger and more capable X-45B (spiral 1) representing a configuration closer to a production-worthy design. Boeing had begun construction of the X-45B as early as 2002 but halted work to rescope the program to meet the emerging requirements from J-UCAS.

Boeing X-45C program manager Darryl Davis remarked, “recent conflicts have indicated a need for greater range and persistence over the battlefield than originally planned.” The X-45C concept had a combat radius of 1,300 nm, a payload of 4,500 lbs., and an empty weight equal to an F-16 at 18,000 lbs.

Boeing continued to internally develop the X-45C under “Project Reblue” starting in mid-2007. The company had hoped to secure future follow-on defense work following the loss of UCAS to Northrop Grumman and Joint Strike Fighter to Lockheed Martin. Boeing also claimed the aircraft’s software was highly applicable to the USAF’s Next Generation Bomber and the USN’s Unmanned Carrier Launched Airborne Surveillance and Strike (UCLASS) programs.

Boeing unveiled its Phantom Ray demonstrator in May 2009 and the aircraft first flew on May 4, 2011. According to a Boeing press release, a least ten flights over a period of six months were planned to test ISR, SEAD/DEAD, electronic attack, “hunter-killer”, and autonomous aerial refueling capabilities.

MQ-20

GA-ASI first explored jet-powered concepts for its MQ-9 in the early 2000s. This concept later evolved into the Predator C which first flew in 2007. GA-ASI sold a single MQ-20A for evaluation to Air Force Materiel Command in 2011 indicating that it may have sold up to seven additional aircraft to an undisclosed U.S. government operator since then. However, the Air Force has never considered the Avenger as a ready replacement for the Reaper in the same way the MQ-9 was seen as a clear replacement for the MQ-1. The MQ-20A has since been used to test autonomous control agents and reinforcement learning architectures.

MQ-X

In a freedom of information request filed by the War Zone, the Air Force described its short-lived MQ-9 follow-on program: “The next generation RPA, or MQ-X as it was more commonly called, promised a more viable solution to meeting the air forces future unmanned ISR needs. During a meeting of the Air Force requirements oversight council in July 2010, ACC outlined the baseline requirements for MQ-X. Unlike the Predator C which offered only minor improvements over the MQ-9, the MQ-X had to be ‘survivable in a contested environment, weather tolerant, and have robust and agile communications.’” The program was ultimately terminated as part of budgetary sequestration in 2012.

Navy Unmanned Combat Air System (N-UCAS) & X-47B

The resulting N-UCAS initially shared many of the basic desired capabilities of the J-UCAS concept. As a pathfinder program, the Navy began the Unmanned Combat Air System Demonstration (UCAS-D) program to validate carrier UAV technologies and operations. On Aug. 3, 2007, the USN awarded Northrop Grumman a $635 million ($945 million in FY25) contract to continue X-47B development and to produce two prototypes under a six-year UCAS-D contract. Total UCAS-D spending would reach $1.4 billion ($1.8 billion in FY25) by 2015.

On February 4, 2011, the X-47B completed its first flight, and carrier tests took place in 2013. The Unmanned Carrier Launched Airborne Surveillance and Strike (UCLASS) was launched in 2011 with the aim of translating lessons learned from UCAS-D into an operational platform. In June 2011, the Joint Requirements Oversight Council (JROC) issued the UCLASS initial operational capabilities document (JROC 087-11) which detailed a survivable, long endurance, ISR, and precision strike platform. However, the JROC rescoped the program in December 2012 with memoranda 086-12 and 196-12. The new requirements downgraded the survivability specification and made affordability the top priority.

The new platform would instead provide ISR and strike capabilities in permissive environments. Payload requirements dropped from 6,000 lbs. to 1,000 lbs. Other notable Key Performance Parameters (KPP) included a unit cost of $150 million ($205 million in FY25) or less and an ability to conduct one 600 nm orbit from the carrier. In April 2013, the USN approved of the aforementioned UCLASS requirements but added aerial refueling as a growth capability. The UCLASS Request for Proposals (RFP) was released in April 2014.

A key requirement for the 2014 UCLASS concept was an unrefueled endurance of 14 hours to provide the carrier group with continuous ISR orbits. The USN argued that a more affordable initial design geared toward the ISR mission could be modified over time to perform more demanding missions. Several members of Congress were deeply critical of the USN’s amended UCLASS concept and sought a long-range, LO, strike platform. The shift toward a tanker from an ISR program began in 2015. In August 2015, the JROC examined Next Generation Air Dominance requirements, which included carrier-based aerial refueling tankers to replace the Super Hornet. As early as March 2000, the USN had identified its need for a carrier-based refueling asset to support its Super Hornets.

In early 2016, then Deputy Secretary of Defense Robert Work led an Office of the Secretary of Defense review on all ongoing DoD UAV ISR programs. The review led to the Carrier Based Aerial Refueling System (CBARS) program which effectively superseded UCLASS. In a mission tanking role, the new UAV would be able to deliver at least 15,000 lbs. of fuel to carrier aircraft operating 500 nm from the strike group. Refer to the Boeing MQ-25 program profile for additional details.

2015-2021: Post-UCAS-D & Roper Programs  

The expendable-exquisite spectrum as unveiled by Roper was a means to frame the discussion of future unmanned systems. Credit: USAF.

The failure to transition the X-45C and X-47B into acquisition programs after a combined investment of over $2 billion (in adjusted FY25 dollars) left the DoD with little demand for future exquisite UCAVs – particularly in the context of budget sequestration. Critically, DARPA would shift its focus from maturing UAV hardware to AI technologies in the mid-2010s. The concept of a government-owned, stand-alone autonomy core that could be transferred between platforms also arose in this period. Notable autonomy-enabling programs included:

• Collaborative Operations and Denied Environments (CODE): government-owned autonomy core applicable to existing systems to improve the viability and lethality of unmanned systems in A2/AD environments. See below for additional details.
  • Follow-on Air Combat Evolution (ACE) program.
• Target Recognition and Adaptation and Contested Environments (TRACE): an algorithm to interpret synthetic aperture radar (SAR) imagery, automatically identifying and locating a target within three seconds. Trained via machine learning (ML) techniques. $6 million, 40-month contract awarded to Deep Learning Analytics on July 20, 2015, and prototype delivered by March 2016. General Dynamics acquired the company on March 1, 2019.
• Distributed Battle Management (DBM): algorithms designed to maximize situational awareness between manned and unmanned platforms with the least bandwidth in contested environments.
• Systems of Systems Integration Technology and Experimentation (SoSite): an effort to integrate mission systems across platforms to promote distributed sensing and greater countermeasures resilience. Live testing in 2018 by Lockheed Martin at China Lake, CA, featuring a ground station, flying test bed, C-12, and flight test aircraft which transmitted data through the company’s STITCHES integration system. Lockheed also used DARPA’s automatic target recognition software on the APG-81.

Dr. Will Roper served as Assistant Secretary of the Air Force for Acquisitions, Technology, and Logistics from Feb. 1, 2018, to Jan. 20, 2021, following his term as head of the Strategic Capabilities Office (SCO) from August 2012 to February 2018. Roper was broadly critical of prior exquisite UCAVs. During his tenure, he sought to prioritize the experimentation of expendable and attritable systems including the XQ-58 Valkyrie, Skyborg, MQ-Next, Gray Wolf cruise missile, low-cost engines, and “Golden Horde” demonstrations.

Roper’s concept of an attritable “loyal wingman” paired with developments in AI control agents and testbeds would eventually form the foundations of the subsequent CCA program.

The DARPA AI Programs

Collaborative Operations and Denied Environments

In January 2015, DARPA announced the CODE program ahead of the FY16 budget release. The initiative sought to develop advanced autonomous behavior algorithms and software for UAVs and cruise missiles in denied environments. Previously, such systems required active control by a human operator or continued communications to provide targeting updates. Near-peer adversaries responded by fielding systems to deny and degrade these C2 capabilities. Furthermore, legacy UAVs such as the MQ-1 and MQ-9 were manpower intensive and required operators to continuously review sensor feeds to garner intelligence insights. DARPA foresaw collaborative autonomy as being a key mechanism to reduce C2 vulnerabilities and manpower requirements. The new modular software architecture would be compatible with the bandwidth and communication systems of existing platforms.

USAF Major Paul Calhoun described some of the new CONOPS enabled by future algorithms in 2015,   

Collaborative autonomy is a somewhat vague term, but perhaps some specific examples will clarify its meaning. Imagine a dozen cruise missiles deep in enemy territory looking for a mobile surface-to-air missile (SAM) site. One could assign each missile independently a search/kill box and hope to find and destroy the SAM in this way. Using collaboration instead, the cruise missile pack could set up a coordinated search grid, notify other missiles of targets of interest, and bring multiple sensors and azimuths to bear to increase the probability of accurate target identification. Adding to this scenario, assume GPS is not available—removing a trusted outside navigational source. This makes accurate positioning and targeting difficult. Within a collaborative network, relative position can be determined. Using known landmarks or a single navigational beacon, the entire pack of missiles can update their position. In this example, absolute position is not that important. Known relative position to the target is sufficient to close the kill chain. Once a target is identified, the cruise missiles could encircle the target and strike simultaneously, overwhelming any missile defense systems in place. Collaboration allows for greatly increased effectiveness and efficiency, allowing the salvo size to be reduced. This effects-based thinking preserves resources while optimizing mission success.

Calhoun further described the benefits of collaborative behaviors in combining RF waveforms constructively (which scales at the number of platforms squared) to increase detection range and to overcome adversary jamming. Autonomous behaviors could also reduce communications bandwidth needs by a factor of 20. Increasing automation would also enable one operator to control as many as six UAVs, according to then-DARPA Deputy Director Stephen Walker in 2015.

In June 2016, DARPA selected Raytheon and Lockheed Martin for CODE Phase I for system definition and preliminary design activities. The first phase cost $7.4 million. Phase II sought to mature the algorithm Suite 2 which ultimately culminated in a series of modified RQ-23 Tigershark UAV tests at Naval Air Station China Lake announced in January 2018. Demonstrated behaviors included dynamic re-routing to avoid pop-up threats, operating in GPS-denied environments, and multi-vehicle collaborative autonomy. The second phase was expected to cost $11.7 million.

DARPA transitioned the program to the USN around 2019 with the Research & Autonomy Innovation Development Environment & Repository (RAIDER) program. Raytheon announced in August 2021 that it had completed a series of flights on Navy-owned BQM-34 target drones operated by its autonomous software suite. The demonstrations were funded by SCO, but the U.S. Navy declined to comment on the activity to Aviation Week.

CODE was also tested onboard the MQ-20 in September and November 2022. GA-ASI announced it had demonstrated a reinforcement learning (RL) architecture software to define the boundaries of the operating envelope available to the MQ-20, then trained the CODE AI agent to calculate the best way to maneuver to an objective. Previous MQ-20 flights with AI agents used less sophisticated training techniques such as supervised and unsupervised learning techniques. Lockheed Martin also contributed a TacIRST pod for the demonstration. The pod enabled GA-ASI to operate the payload using an AI agent RL-based machine learning system.

For the November test, GA-ASI used, “a government furnished CODE autonomy engine and the government-standard Open Mission Systems (OMS) messaging protocol to enable communication between the autonomy core and TacIRST. In addition, GA-ASI used General Dynamics EMC2 [Einstein Box], and open architecture MFP [multi-function processor] with multilevel security infrastructure to run the autonomy architecture, demonstrating the ability to bring high-performance computing (HPC) resources to ACPs [autonomous collaborative platforms] to perform quickly tailorable mission sets depending upon the operational environment.”

A total of four aircraft participated including an MQ-20A as well as a pair of F-5s from Tactical Air Support and Sabreliner. The MQ-20s and IRST-equipped F-5s were connected over a tactical targeting network technology (TTNT) mesh network to share sensor observations. Five virtual MQ-20s also participated in the test.   

Alpha Dogfight Trials

As a precursor to the later Air Combat Evolution (ACE) program, DARPA launched the Alpha Dogfight Trials (ADT) – a three-stage virtual competition between AI agents in September 2019.

DARPA took inspiration from Google’s televised AlphaGo matches which culminated in the 4-1 victory over 9-dan Go professional Lee Sedol in March 2016. The program was developed by its subsidiary DeepMind and supported by the company’s Tensor Processing Unit (TPU) hardware which was specifically designed to support AI applications.

Entrants to the DARPA challenge would utilize the JBSim open-source flight dynamics model for the simulation environment and the FlightGear Flight Simulator for the visuals. DARPA selected the Johns Hopkins Applied Physics Laboratory (APL) to help run the competition.

An APL White Paper explained the four levels of AI agents envisioned:

• rudimentary opponent: flies straight and level at a fixed speed to simulate a cruise missile
• basic opponent: executes time-based minor altitude and speed deviations such as executing a one-circle flow maneuver
• scripted opponent: recognizes the state of any engagement and employs a scripted response akin to a novice pilot
• RL opponent: state-of-the-art agent developed without human input intended to be analogous to an expert pilot

Phase 1 exhibition match entrants included: Boeing subsidiary Aurora Flight Sciences, defense AI company EpiSci, Georgia Tech Research Center, simulation specialist Heron Systems, Lockheed Martin, research company Perspecta Labs, deep learning experts PhysicsAI, and reasoning software developer SoarTech.

Phase 2 marked the first competitive assessment between teams. The top four teams were Aurora Flight Sciences, PhysicsAI, Lockheed Martin, and Heron Systems. The latter two emerged as finalists. In the third and final round, Heron’s Falco AI agent defeated Lockheed. Both teams utilized aggressive agents and appeared evenly matched, but Heron System’s Ben Bell said Falco’s refresh rate of 10 Hz compared to most team’s 50 Hz gave it a decisive edge. Bell also credits the company’s training regime. Over five weeks, Falco trained against 102 unique AI agents in billions of simulations.

Heron System’s agent emerged victorious 5-0 against F-16 fighter weapons school instructor “Banger” who had approximately 2,000 flight hrs. The result attracted some criticism as the competition did not replicate the following real-world limitations:

1. Falco had perfect information on its aircraft and its opponent courtesy of the open source JSBSim, enabling it to exploit its fine-precision control
2. The exercise did not feature a safety bubble to avoid collisions, as would a real training engagement
3. The human pilot did not have to endure g-forces
4. Simple 1 vs. 1 basic fighter maneuvering (BFM) engagements featuring a gun. Damage was dealt by maneuvering within a 1° cone 3,000 ft. long
5. The AI agents were not permitted to learn from each match

DARPA ACE program manager Lt. Col. Dan Javorsek underscored the purpose of ADT was to build operator confidence in AI programs, ““If we convinced even a couple of pilots that what they were seeing out of this Heron autonomous agent looked like something that was intelligent and creative and making smart decisions in this dynamic BFM engagement, then I’m considering it a success because those are the first steps I need to create trust in these sorts of agents,” says Javorsek.”

Banger noted the potential for AI wingmen in a four-ship formation would not replace the human pilot but would enhance the lethality of the “human weapon system.”

Shield AI subsequently announced its acquisition of Heron Systems in July 2021. The San Diego-based start-up is also known for its V-bat UAV which competedto replace the Group 3 RQ-7 Shadow in U.S. Army service.

Javorsek would retire from the Air Force in April 2023 and join EpiSci as its Chief Technology Officer in August 2023.   

Air Combat Evolution (ACE)

DARPA’s ACE program sought to transition AI agents to simulated WVR engagements with real aircraft. In 2020, DARPA reported the program had four Technical Areas (TA):

• TA1: increase air combat autonomy performance in local behaviors (individual aircraft and team tactical) with the use of simulators
• TA2: build and calibrate trust in air combat local behaviors by using small UAVs
• TA3: scale performance in trust to global behaviors (heterogeneous multiple aircraft) with the L-39 and F-16
• TA4: build infrastructure for full-scale air combat experimentation

By 2021, DARPA had prioritized development of the X-62A testbed which had previously served as the NF-16D Variable In-flight Simulation Test Aircraft (VISTA) for thrust vectoring tests in 1992. Javorsek explained, “…when it became clear that the X-62A would be available earlier than expected, the resources were appropriately refocused to that aircraft…I was happy to see the shift to VISTA but I’ll admit I was a bit sad to see the original drone BFM and all L-39 work drop from the program.”

The aircraft was originally built as an F-16D Block 30 Peace Marbel II (Israeli) aircraft and was upgraded with Block 40 avionics. Lockheed Martin Skunk Works made significant modifications to the aircraft between 2021-2022 including the installation of the company’s System for Autonomous Control of the Simulation (SASC) and Model Following Algorithm (MFA) as well as updated VISTA simulation system (VSS) developed by Calspan. The VSS enables the X-62 to mimic the flight characteristics of other aircraft by re-tuning the flight control system to reflect a digital model of the simulated aircraft’s flight envelope. The X-62 can load AI agents that command the VSS through the SACS, the core of which is GD’s OMS EMC2. The human pilot provides an additional layer of flight safety and can override the AI agent if need be. Like the ADT, the AI agent is given perfect awareness of its environment with the use of Secure Live Virtual Constructive Advanced Training Environment (SLATE) pods fitted to both the X-62A and the aircraft of its manned opponent.   

In early December 2022, the X-62A tested four AI agents developed by PhysicsAI EpiSci, Shield AI, and Johns Hopkins APL. Later flight testing included BFM within 2,000 ft. and speeds up to 1,200 mph.

Artificial Intelligence Reinforcements (AIR)

In November 2022, DARPA announced its AIR program to develop BVR autonomy behaviors. The first phase was expected to feature an 18-month evaluation for up to six companies. Phase 2 would support four companies during a 30-month evaluation. DARPA program manager Air Force Lt. Col. Ryan “Hal” Hefron remarked, “AIR is a follow-on program which seeks to develop dominant tactical autonomy from multi-ship, beyond visual range air combat missions.” The FY25 budget specifies testing for both Defensive Counter Air (DCA) and Offensive Counter Air (OCA) behaviors. Phase 1 participants include:

• Joint Epi Sci- PhysicsAI team $6 million, June 2024 (now owned by Applied Intuition as of February 2025)
• BAE Systems with a $4 million award, September 2024
• Northrop Grumman Misson Systems
• Lockheed Martin Mission Systems
• Systems & Technology Research (STR)
• Strategy Robot

Autonomy agents created under AIR will be tested using a fleet of six modified F-16s based at Eglin AFB, FL, under the Viper Experimentation and Next-gen Operations Model-Autonomy Flying Testbed (Venom-AFT) program. The FY25 budget requests $76 million for VENOM prototype modifications as well as $62.6 million for flight testing between FY23-29, totaling $139 million. Work will continue through at least FY29.

XQ-58 Valkyrie

On July 11, 2016, Kratos was selected by the Air Force Research Laboratory (AFRL) to develop a “high-speed, long-range, low-cost, limited life-strike UAS” as part of its Low-Cost Attritable Strike Demonstration (LCASD) program within the broader Low-Cost Attritable Aircraft Technology (LCAAT) portfolio of systems. The new UAS would improve the survivability of manned platforms in contested environments by providing forward sensing, communications gateway, and limited SEAD/DEAD capabilities.

Kratos reported the initial award of $7.3 million could be supplemented by up to $33.5 million in internal funds for a total of $40.8 million. The company outlined the following goals:

• acquisition cost of $3 million ($3.9 million in FY25) or less for the first unit up to 99 units, and $2 million ($2.6 million in FY25) or less for 100 or more units
• 1,500 nm mission radius with 500-lb. payload
• internal weapons capability for at least two GBU-39 small-diameter bombs
• capable of Mach 0.9 dash
• runway independent takeoff and landing
• commercial off-the-shelf materials, subsystems, manufacturing processes, and open Mission systems concepts
• LO design for vehicle shaping, elimination of gaps, and survivable inlet integration

The first XQ-58A flew in March 2019. The Air Force would continue to test the Valkyrie including F-22 to F-35 communications gateway capability in December 2020, release of the ALTIUS-600 loitering munition in March 2021, and surrogate testing for autonomy efforts between 2022-2023 at Eglin Air Force Base (AFB), FL.

As of the time of this writing, Kratos is continuing to develop new derivatives of the XQ-58, including electronic warfare and conventional take-off and landing variants while looking for prospective customers.

ADAIR-UX/Bandit/Project Red 5

In the late 2010s, Blue Force Technologies (BFT) met with then Gen. James Holmes who was serving as Deputy Chief of Staff for strategic plans and requirements. Speaking with The Warzone in September 2023, Anduril’s Van Timmeren recalled BFT briefed its “Grackle” concept for an unmanned nap-of-the-earth ISR platform. While the USAF did not have an open tender specific to Grackle, “somebody in that briefing actually pulled them off…to the side of the meeting, or after the meeting in the hallway, and said have you looked at high-performance aircraft for adversary air purposes?”   

BFT evolved Grackle to its REDmedium concept between 2019 and 2020, which was eventually re-branded to Fury. In September 2019, then Air Combat Command (ACC) Commander Gen. James Holmes commissioned then-Major Andrew Gray to study future adversary air (ADAIR) concepts. The resulting April 2020 paper, “Creating A Balance Adversary Air Enterprise to Bolster USAF Combat Readiness”, called for the acqustion of  two new red air training aircraft: 200 ADAIR-X (manned) and 108 ADAIR-UX (unmanned). The manned option would be a modified T-7B featuring 4,800 lb. more fuel, wingtip pylons, IRST, datalink, and an electronic warfare pod.

ADAIR-UX would be needed to replicate advanced threats from China such as the J-15 and J-20. Proposed requirements include a $4 million flyaway cost, cost-per-flight hour (CPFH) of $4,000, variable RCS from 0.1m^2 (-10 dBSM) to 15 m^2, electronic jamming capabilities, and autonomy software. Following the internal paper, Air Combat Command (ACC) announced it would seek a new fleet of low-cost unmanned red air training assets in August

In March 2022, AFWERX selected BFT under its Bandit program to mature the Fury air vehicle over 12 months to the “critical design level, perform engine ground testing and validate the design of the engine installation,” according to the Air Force. The contract includes options to build up to four Fury aircraft. The initial award of $9 million was funded through the Small Business Innovation Research (SBIR) program. In January 2023, BFT announced it had conducted ground testing which validated Fury’s carbon fiber composite propulsion flow path system.

Subsequent awards amount to $44 million but only $24 million had been obligated as of August 2024, according to usaspending.gov. The Air Force’s FY23 budget requested a total of $67 million for future ADAIR development. However, the program was canceled in 2023 and was superseded by the CCA program with the FY24 budget.

On Sept. 7, 2023, Anduril Industries announced it had acquired BFT. Anduril proceeded to offer a Fury derivative aircraft for the USAF’s CCA program.

In July 2024, the DoD’s Test Resource Management Center (TRMC) awarded GA-ASI a $98 million contract to modify a pair of MQ-20As with new sensors, datalinks, and advanced mission systems to enable them to act as surrogates for air-to-air threats during training missions.

MQ-Next/Next-Gen Multi-Role UAS

Following Congressional scrutiny of its proposal to terminate MQ-9 production in FY21, Roper announced the service’s MQ-Next program in June 2020. MQ-Next originally sought to field a survivable, medium-altitude Reaper replacement which would enter service in 2031. In July 2020, Roper said the USAF must balance MQ-9-level functionality with the need for a more survivable aircraft, which will feature more automation so that fewer personnel are required per system. The new, affordable aircraft must fulfill more than just the MQ-9 ISR mission because of declining budgets. MQ-Next must operate successfully in a high-end fight that can provide a “picket line” in the defensive counter air role.

One possibility is outfitting the new aircraft with an air-to-air weapon, Roper said. The service was still deliberating if MQ-Next would be a standalone aircraft or be integrated with its Skyborg autonomous UCAV program. The MQ-Next rollout received a lukewarm response from the Hill, with multiple members of Congress expressing concern over the program’s long-time horizon and lack of concrete details.

The Air Force greatly refined its MQ-Next concept of operations by the March 2021 RFI release by the Intelligence, Surveillance, and Reconnaissance (ISR/Strike) program executive office. The RFI discusses three focus areas: airborne moving target indication (AMTI), high-value airborne asset protection (HVAAP) and attritable technologies. As part of the broader Advanced Battle Management System (ABMS) architecture, the Air Force is attempting to disperse the functions of large, maned C4ISR assets like the E-8 J-STARS and E-3 Sentry into disaggregated platforms through artificial intelligence and fused sensor inputs. The RFI also states the Air Force “will accelerate capability development within the pipeline stressing a ‘Speed to Ramp’ initiative focused on fielding capability prior to the 2026/2027 timeframe. Solutions proposed for integration on future Next-Gen Multi-Role UASs should target the 2030 timeframe.”

Around this time, U.S. contractors unveiled the following Next-Gen Multi-Role UAS concepts:

• Northrop Grumman SG-2 (SG-102) – high-end attritable platform at $20 million (without mission systems), cranked-kite platform reminiscent of X-47B, 1,000 nm range, MTOW of 20,000 lb., Distributed autonomy/responsive control (DA/RC) battle management system evolved from UCLASS control system proposed 15 years ago. Northrop also proposed the larger SG-1 (SG-101). Additional LO features could be added but at the penalty of an expected 50-100% increase in the unit fly-away cost. First discussed in Sept. 2020.  
• Northrop Grumman Model 437 – attritable $5-6 million platform featuring a P&W 535 3,400 lbf. engine (initially advertised a with $2.4 million FJ44-4A as part of the total flyway cost), 3,000 nm range at Mach 0.8, MTOW of 7,500 lb., payload of either a pair of AIM-120s or a side-looking radar. The airframe features a top-mounted inlet, narrow curved fuselage bisected by a chine line, V-tail, and a lambda wing planform. Developed from Scaled Composites Model 401 demonstrators which first flew in 2017. The Model 437 was made public in September 2021.
• Lockheed Martin – flying wing concept seemingly descended from the RQ-170/ Sea Ghost lineage. Lockheed offered a similar concept to Poland’s short-lived Harpi Szpon (Harpy Claw) program to field an unmanned complement to the F-35. The aircraft features a compressed dorsal mounted inlet with a mesh grill (for RCS reduction), edge treatments, swept tips to minimize tip diffraction, single-W trailing edge, ventral mounted internal mission bay shown in promotional video for EO/IR – other payloads likely ESM, EW, etc.
• GA-ASI – promotional images show a flying wing concept but the company provided few details. The concept would later show significant similarities to Gambit-4. 
• Boeing – Boeing similarly provided few details as to its MQ-Next offering.

MQ-Next was ultimately canceled. For additional details, refer to the MQ-9 profile.

Skyborg

ARFL Skyborg concept. Image Credit: USAF via YouTube

Skyborg was the follow-on program to LCASD within LCATT and would eventually consist of three main elements: (1) autonomy core systems (ACS), (2) low-cost air vehicles, and (3) operational experimentation. Skyborg would operate alongside F-22s as part of the manned aircraft’s loadout. Later roles included defensive counter air (DCA), Advanced Battle Management System (ABMS) as a communications gateway node, strike, ISR, and close-air-support. The service had hoped Skyborg could begin replacing its pre-Block F-16s after 2025 and its MQ-9s after 2030. Cumulative outlays on Skyborg RDT&E between FY21-23 totaled $251 million and prior FY18-23 SCO spending on Avatar reached $240 million, according to analysis by CSIS.

The Skyborg concept was originally proposed when Roper led the SCO between 2012-2017. Flying demonstrations for Avatar which hosted autonomous software developed under DARPA’s CODE program occurred in 2015 and 2016, but many of the program’s details remain classified. Roper proceeded to transfer the project over to AFRL in 2017 at which time it was known as “Avatar”. A year after taking over Air Force acquisitions, the program was re-named back to Skyborg in October 2018 along with the creation of a new program office. 

In March 2019, Roper revealed the project to a group of reporters a week before AFRL issued the first RFI to industry. Skyborg was initially proposed to follow a traditional acquisitions model involving a single prime contractor responsible for both the system’s hardware and software. The USAF designated Skyborg as one of its four “Vanguard” pathfinder prototyping and experimentation programs in 2019. By early 2020, Skyborg was reorganized to feature distinct air vehicle and control agent components with the requirement for government ownership of the ACS. Skyborg received its own program element with the FY21 budget. At that time, AFRL had hoped to achieve an early operational capability for the Skyborg family of systems by FY23. The initial “Attritable-One” Skyborg capability would be among the first fielded of 30 proposed product lines for ABMS.

Leidos won the ACS in a May 19, 2020, contract worth $28 million. Leidos had previously developed the control software for the Navy’s Sea Hunter unmanned surface vessel which transited between San Diego and Honolulu in 2018. The company’s ACS made its first flight onboard a Kratos UTAP-22 target drone on April 29, 2020. The ACS underwent additional testing on a pair of MQ-20s during the Orange Flag 21-3 exercise at Edwards Air Force Base (AFB). The Air Force Test Center reported,

“The flight demonstrated matured capabilities of the ACS that enabled two MQ-20s to fly autonomously while communicating with each other to ensure coordinated flight. Additionally, the aircraft responded to navigational commands, stated within specified geo-fences, and maintained flight envelopes. Both aircraft were monitored from a ground command and control station.”

In July 2020, Kratos, Boeing, and Northrop Grumman were awarded contracts to, “produce missionized unmanned aerial vehicle prototypes with the ability to fly in experimentation events while teaming with manned aircraft.” The Air Force Life Cycle Management Center (AFLCMC) announced contracts with ceiling values to the following: Boeing at $25.7 million, Kratos at $37.8 million, and GA-ASI at $14.3 million.

The Air Force established the Skyborg Prototyping, Experimentation, and Autonomy Development (SPREAD) indefinite delivery/indefinite quantity (IDIQ) contracting vehicle in September 2020 with a ceiling value of $400 million. Actual awards under SPREAD total approximately $59 million as of August 2024, according to the contracting tool usaspending.gov. For example, despite the July 2024 AFMLC release of Boeing being awarded $25.7 million, actual awards totaled $5 million.

Roper’s ambitious schedule to field Skyborg ultimately failed and the program did not progress past initial RDT&E efforts.

Off-Board Sensing Station (OBSS)

The de facto successor to the Skyborg airframe effort was theOBSS program. AFRL describes OBSS as the second increment of LCAAT, building upon the lessons learned from the XQ-58 program. The program sought to develop a highly modular, conventional takeoff and landing, jet-powered UAV at an affordable price.

AFRL held a competition between GA-ASI, Kratos, and five undisclosed competitors in 2021 before down-selecting the first pair of companies for further development. On Oct. 25, 2021, Kratos was awarded a $17 million contract to complete a critical design review (CDR) of its OBSS proposal over the next 12 months. GA-ASI was awarded a similar $17.8 million contract the next day. The contracts included an option for a follow-on 15-month manufacturing and demonstration phase. 

The aircraft’s primary mission was to act as a “sensor extension” for manned aircraft, according to Kratos unmanned systems division president Steve Fendley. The aircraft would also be expected to carry “significant offensive weapons volume” allowing the aircraft to serve as a “weapons bay extension for manned aircraft,” he adds.

AFRL had planned source selection for the end of Q3 of 2022 followed by experimentation flights in 2024. OBSS would be followed by the armed Off-Board Weapon System (OBWS) in 2026 and an EW variant by 2028. According to AFRL, OBWS is expected to be faster, more maneuverable, and have a longer range than the OBSS, at the cost of endurance.

GA-ASI Gambit family featuring a common core. Image Credit: GA-ASI

Kratos offered its Demigorgon, featuring a chinned “platypus” nose, top-mounted inlet and target flyaway cost of $8 million (assuming annual production quantities of 50 aircraft per year). GA-ASI offered its Gambit family of UAVs featuring a common core set of components accounting for 70% of the overall cost of each variant. GA-ASI has subsequently noted that its OBSS offering shares 78% commonality with its CCA offering. Proposed variants include ISR, air-to-air, training, and combat reconnaissance (a flying wing derivative with greater LO than the ISR variant). Both concepts would be 10,000 lb. MTOW class aircraft.

XQ-67A. Image Credit: GA-ASI

On Feb. 2, 2023, GA-ASI announced its Gambit had been chosen for OBSS and would proceed to flight testing the following year. Total OBSS awards to GA-ASI as of August 2024 total $58.8 million, per USA spending.gov.

The XQ-67A made its first flight on Feb. 28, 2024. As of the FY25 budget, total FY23-29 OBSS & OBWS spending totals $219 million.

CCA Program: History & Shaping Factors

The USAF’s CCA program is intended to solve long-term structural, operational, budgetary, and industrial challenges. Upon his July 2021 appointment as Air Force Secretary, Frank Kendall began studying options for unmanned wingmen. The service secretary first publicly mentioned the possibility of an unmanned complement to the B-21 in December of that year. Kendall was broadly critical of open-ended Vanguard experimentation efforts carried out by the prior USAF Assistant Secretary for Acquisition Will Roper (Feb. 2018-Jan. 2021) and insisted on launching a program of record ending in procurement.  The resulting CCA program was launched in FY24 and was intended to act as a force multiplier for both NGAD and a subset of the F-35 fleet.

Kendall outlined an initial target to field at least 1,000 CCAs, including at least three increments to be fielded in two-year cycles starting in FY28. Originally, 200 NGAD platforms and 300 F-35 Block 4s would receive two CCAs each. 

Since its inception, the U.S. market for CCAs has progressively shifted toward more attritable systems with the USAF’s CCA Increment 1 down-select of Anduril’s Fury and GA-ASI’s Gambit in April 2024. Five undisclosed companies are competing for the government owned autonomy reference architecture. Final source selection is expected in FY26. The elimination of major industry primes follows Kendall’s progression of flyaway cost targets from half that of the F-35A to between ⅓ and ¼ ($25-33M). Several factors are likely shaping the service’s preferences: (1) resourcing constraints from the $63+ billion (in FY20 dollars) cost overrun from the Sentinel ICBM, (2) a desire to broaden the industrial base to encompass non-traditional suppliers, and (3) Kendall’s atypical role as a service secretary.

In January 2024, the Air Force notified Congress that the Northrop Gruman LGM-35A Sentinel ICBM triggered a Nunn-McCurdy breach from excessive cost growth. By July 2024, CAPE assessed the program’s cost had grown by 81% from $77.9 billion in FY20 to $140.9 billion. Challenges related to the Sentinel’s ground-based infrastructure is driving growth as opposed to the missile itself. The $63+ billion deficit effectively adds zero additional capability beyond prior plans while simultaneously necessitating diverting resources from existing programs. Service officials expect these costs to materialize in the 2030s – or in tandem with peak modernization efforts including NGAD, Next Generation Aerial Refueling System (NGAS), CCAs, and continued B-21 FRP. Cost growth from the Sentinel program appears to have contributed to Kendall’s decision to pause NGAD in June 2024.

The Defense Industrial Base (DIB) pressures appear to be a key shaping factor for the CCA program – arguably above any operational considerations. The post-Cold War era DIB management strategy has largely been to hold a “franchise system” in which a mission market is competed once every 30 years. The losing primes effectively have to sustain themselves off of international sales with a legacy platform until another domestic opportunity arises. The U.S. strategy has arguably resulted in an ever-diminishing DIB that is less competitive and retains little excess capacity to meet surge requirements. The service aims to cultivate nontraditional suppliers and foster additional competition both by holding more frequent competitions via the CCA increment system and by dividing each award into multiple sub-segments.

The Air Force’s overriding objective appears to be to field a capability quickly by the end of the decade. Kendall has also framed CCAs as a key part of his legacy. His prior positions in DoD have afforded him outsized influence relative to his uniformed counterparts. Kendall has been crucial toward not only the genesis of the CCA program, but also continuing to shape its future in relation to other programs such as pausing the NGAD contract award to re-evaluate its requirements and calling for a unified CONOPS for NGAD, NGAS, and CCA Increment 2. It’s unclear if Trump’s pick as the next Air Force Secretary, Troy Meink, will share the same vision and influence going forward. A review of historical UCAV programs shows the USAF’s frequent vacillations between exquisite and more attritable platforms as well as a tendency to cancel programs before they can mature into operational platforms.

A pivot from Kendall’s strategy already appears underway as of the early Trump Administration. NGAD requirements have firmed up, NGAS has been deprioritized, and CCA Increment 2 remains undefined. As of March 2025, senior Air Force leadership were lobbying to continue NGAD as originally planned. Speaking at the Association of the Air Force in Aurora, CO, Maj. Gen. Joseph Kunkel – the service’s Director of Force Design, Integration and Wargaming on the Air Staff, explained service studies validated the exquisite NGAD concept, “We tried a whole bunch of different options and there is no more viable option than NGAD to achieve air superiority in this highly contested environment.” Later that month, Rep. Rob Whittman (R-VA), confirmed both the USAF and Navy had briefed the President on their NGAD concepts. Aviation Week confirmed USAF CSAF Gen. Alvin personally led the briefing on March 12.

Kunkel explained to Aviation Week that Increment 2 is still much less defined than NGAD. The following is a consolidation of his remarks, “We’re still learning how on how to use them. … We know a significant amount about CCA. We know in an analysis we have it’s a game changer on the battlefield. It enables the joint force, it keeps our blue forces much more survivable. It has an impact on this mass that we’re going up against in these highly contested environments. But we’re still working on what this whole suite of CCAs look like. What that looks like, we’re still figuring out and we are working with industry on that. And industry, frankly, has got a lot of operational analysis that’s telling us where this thing needs to go. There’s some folks that are like, CCA Inc. 2 has got to be more exquisite, it’s got to be stealthy, it’s got to have more capability. … Others say actually, not it’s got to be cheaper, it’s got to be less expensive, it’s got to be expendable. We’re kind of allover in this space. What I think we’re going to find out in the end is that CCA Inc. 1 is kind of like the Goldilocks, right in the middle. We may find we need some that are higher end, and I’m pretty sure we’re going to need a lot that are lower end.” [emphasis added]

Production & Variants

Total FY23-29 CCA RDT&E spending is projected at approximately $9.3 billion according to the FY25 budget.

CCA Increment 1

Increment 1 is optimized for the air superiority mission utilizing existing technologies including limited autonomy and LO. As of November 2024, the service’s top Increment 1 priority is acquisition speed, followed by cost and “minimum viable capability”, according to Air Force Materiel Command’s (AFMC) Col. Timothy M. Helfrich – Senior Materiel Leader for the Advanced Aircraft Division, speaking at a Mitchelle Institute event.   Source selection is expected in FY26 ahead of FY28 fielding. A total of 100 Increment 1 CCAs would be “on order or delivered” by FY29, according to Kendall.

Confirmed features include carriage of Raytheon AIM-120 Advanced Medium-Range Air-to-Air Missile (AMRAAM) and air-moving target indication (AMTI) sensors. In November 2023, Air Force Future’s Thomas Lawhead said CCA increment 1 will have a range “relatively the same as the current fighter fleet [combat radius of 500-760 nm], potentially a little bit longer.” Options for aerial refueling would be dependent upon the contractor’s offering. However, the subsequent House FY25 NDAA notes the service’s CCA CONOPS called for a range of 3,000 nm which would imply a combat radius of less than half that figure. The NDAA also specifies an ability to operate from runways one-quarter the length of traditional aircraft. Fighter aircraft such as the F-22 and F-35 typically require a runway of 7,000-8,000 ft. subject to configuration and environmental factors. Of note, the House bill’s range requirements arguably do not align with publicly known features of either Fury or Gambit.

Airframe features vary between Anduril’s Fury and GA-ASI’s Gambit-derived CCA. Both feature varying degrees of survivability, though GA-ASI’s preceding XQ-67 suggests a more disciplined LO approach when compared to Fury in terms of planform alignment, the use of V-tails (as opposed to a single parallel vertical tail), and the use of an internal weapons bay as opposed to wing-mounted hardpoints. Fury does feature a serpentine inlet which is expected to reduce its frontal signature. However, the ADAIR-UX requirement that Fury competed for specified a lower-bound RCS target of -10 dBSM (0.1m^2).

Speaking on the Merge podcast, General Atomics Aeronautical Systems (GA-ASI) VP for Advanced Programs Mike Atwood remarked that the USAF determined the Boeing X-45 (ranging an empty weight of 8,000-18,000 lbs. for the A and C variants respectively) was too large and the 2,500 lbs. Kratos XQ-58 was too small. Accordingly, Gambit is a 10,000 lbs.-class MTOW aircraft sized between the XQ-58 and X-45.

GA-ASI’s Dave Alexander told Breaking Defense in an October 2024 interview that the company’s peak production rate was one aircraft every three days or about 100 per year for its historical programs. The company has the capacity to double its production capacity for CCAs with its existing 5 million square-foot Poway, CA, production facility. Anduril has not confirmed which facility it would produce its CCAs from. However, Anduril has outlined plans to construct its 5+ million square-foot “Arsenal-1” production plant near Rickenbacker international Airport, OH. The facility is expected to begin operations by July 2026.

At the March 2025 Association of the Air Force (AFA) conference, the service disclosed Gambit and Fury would be designated as the YFQ-42A and YFQ-44A respectively.            

CCA Increment 2

Increment 2 was expected to be fielded in FY30, but the program’s requirements have remained  fluid. While the service and industry had an initial expectation that the second increment would feature improved LO and autonomy, the Air Force has since reevaluated earlier assumptions. In October 2024, Lockheed’s CEO Jim Taiclet similarly described Increment 1 as being a proof of concept for experimentation, leaving combat roles for Increment 2. These comments were later rebutted by CSAF Gen. Alvin. On Dec. 10, 2024, Air Force Acquisition Secretary Andrew Hunter confirmed Increment 2’s final requirements would be set by the Trump Administration.

Expanding on the Air Force’s reasoning in a November 2024 event, Helfrich confirmed, “no one should think Increment 2 means Increment 1 plus…that doesn’t mean Increment 2 is more capability. We’re still looking to figure out whether the right balance, and we’re doing analysis, is to further bring down the capability to maximize a low cost or is it that I need to maybe change even what the focus is from a missile truck to something else – maybe the electronic magnetic spectrum or whatnot.”

In the latter half of 2024, Kendall’s team sought to reframe Increment 2 within a broader unified CONOPs for NGAD and NGAS as well as the stand-up of the new Integrated Capabilities Command. In an exit interview with Air & Space Forces Magazine, Kendall pivoted again to suggest Increment 2 should not be exquisite but ought to be more capable than the first, “I think, personally, something that has some increase in cost over Increment 1 would not be outrageous… 20 or 30 percent [$30-$39.6 million flyaway cost], something like that. But, again, it depends upon the mix right? What capabilities do you put on every aircraft, every CCA? What do you distribute?” Kendall explained that beyond that price point, the service would be unable to achieve affordable mass.

CCA Increment 3

The USAF has disclosed few details about Increment 3 other than its FY32 fielding date. In December 2024, USAF Acquisition Secretary Andrew Hunter suggested it was “a little bit out there in the future” and that he had “nothing to say” about Increment 3 in detail.  

Exquisite vs. Attritable Debates

The Air Force’s fluid requirements for Increment 2 has solicited advocacy from industry with traditional primes calling for more survivability and non-traditional OEMs calling for even greater expendability. Traditional primes have argued the service is fixated on achieving low unit flyaway costs while ignoring the overall cost-effectiveness of CCAs. For example, former Lockheed Martin Skunk Works’ chief John Clark told Aviation Week that internal wargaming showed 80% of attritable CCAs would be lost in a conflict, so continuing to buy them would be a “losing proposition.”

Clark described the company’s Increment 1 offering as “gold plated” – seeking to ensure a viable solution by making survivability assumptions. Company promotional materials appear to show a derivative of its RQ-170 family which was offered during the brief MQ-Next program. Clark indicated Skunk Works would not make the same mistake, but there would be a “reckoning” without adequate LO. In response, Lockheed has revived its Project Carrera program to develop a CCA data-link pod for the F-35. Carrera is anticipated to work in concert with its company-funded Speed Racer expendable CCA concept which may inform its Increment 2 offering.

After the April 2024 GA-ASI & Anduril down-select, Northrop Grumman CEO Kathy Warden remarked, “We are really positioned to provide the best solutions that our customer needs against a high-end threat, however, we are not looking to compete in a more commoditized part of the market at very low cost and not survivable systems. That’s just not our business model, and we know it.” In October, Northrop Grumman Aeronautics President Tom Jones has similarly called on the service to reevaluate its survivability requirements, “I’m not saying everything has to look like a B-21. There’s probably a fine line between getting the survivability of a platform right so it can deliver effects and where it would be much less effective at doing that.” 

GA-ASI has said the future of CCAs is more expendable and lower cost – referencing its Long-shot concept funded by DARPA. For further details as to the exquisite vs. attritable debates, refer to the “Analysis: Roles & Missions” section.

Features

Each increment of CCAs is expected to feature a different design. However, there are expected to be common features that could be shared across platforms.

Autonomy Core

Autonomy technologies largely define CCA from historical unmanned combat aerial vehicle (UCAV) programs. CCAs will also be employed within the broader sensor fusion concept. Aircraft systems generate enormous volumes of sensor data, but pilots need actionable information to attain decision-making superiority. Sensor fusion effectively enables the pilot to spend more time deciding how to employ the aircraft rather than flying it. As a capability floor, the level of autonomy in CCAs must be sufficient to reduce rather than increase the pilot workload.

Service analysis and improvements in automation technologies have since convinced the service that higher numbers of CCAs per manned aircraft increase mission effectiveness. In Nov. 2024, Kendall mentioned the service is looking beyond an updated requirement to control three to five CCAs per fighter.

However, AI arguably still represents the highest technical risk element of the program. The first increment is expected to feature a limited degree of autonomy which will be improved with later versions. Contractors also foresee different approaches to autonomy such as being overlaid on top of existing subsystems or being an integrated, all-encompassing software suite.

One of the defining features of the CCA program will be a government-owned AI reference architecture. Contractors foresee different approaches to implementing autonomy. Lockheed Martin Skunk Works views the AI pilot as mimicking a human. The AI is responsible for basic flight behaviors and receives directives and supervision from a human operator. The autonomy core is overlaid on top of the CCA’s onboard processing systems. Skunk Work’s John Clark told Aviation Week’s Steve Trimble,

“We’re introducing just the behaviors that are necessary to allow the pilot to interact with the [CCA’s] in a collaborative system and get the expected behavior out. We’re sing the traditional process of the flight control system does what the flight control system does, the sensor management does what the sensor management process does. Those pieces of software already exist so let me just put the AI behaviors on top. Those are pretty simple behaviors: ‘your objective is to go that way, engage at this rate that’s weapon from this direction.’”

AI will be used to train autonomy behaviors in simulations and exercises to form a menu of behaviors before combat. The CCA’s onboard processor will be expected to:

1. understand the aircraft’s condition state
2. refer its condition relative to the lessons of its simulations
3. chose a set of algorithms in a drop-down menu of behaviors and execute

Clark described a counter-air scenario in which one CCA faces two enemy aircraft. Multiple decisions must be made: should the CCA engage? If “yes,” then should the CCA first aim at the enemy on the left or the right? If the answer is “no,” then a different algorithmic behavior takes over, but it must decide on the best approach to evade the threat. If all of these behaviors are contained within the same algorithm, Clark said, there is a potential for “behavioral cross-talk” that confuses the AI software.

In essence, the CCA will operate not with a single AI pilot onboard, but with multiple sets of algorithms that are optimized for discrete behaviors. Clark told Aviation Week that this approach is more effective than trying to develop a single algorithm capable of performing each of these missions interchangeably. That’s where the monolithic or the God AI are going to run into a problem, he said.

“You’re going to [get] unexpected behaviors because of all these competing variables that aren’t … coupling toward the same decision,” Clark said.

Clark believes this simplified method of interaction between the AI agent and onboard systems will enable each human pilot to direct up to eight CCAs. Other nontraditional defense suppliers foresee the development of a single all-encompassing algorithm to dictate behaviors. In both AI models, the CCA is not expected to write new code for itself as it progresses its mission.

The level of autonomy the industry can achieve is likely to dictate future mission sets. The most basic level of autonomy which will be fielded as part of Increment 1 envisages an armed decoy role. Lockheed Martin Skunk Works President John Clark told Aviation Week in a June 2024 interview, “CCAs are really just turning into expensive targets…You’re going to want the system to just be smart enough to live long enough to do its core mission, and then end its life in a decoy sort of role. And so that’s where you try to focus AI behaviors.” Greater levels of autonomy will also likely be needed in situations where human decision-making is too slow. Clark told Aviation Week that the company’s Enhanced Collaborative High-Frequency Orientation System (ECHOS) AI agent will test AI vs AI engagements featuring two live and two simulated L-39s.

Avionics

CCAs are expected to be able to accommodate multiple sensor payloads with minimal power, weight, and cost (SWaP-C) requirements. For example, BFT announced Fury was able to accommodate 400 lb. payloads in a detachable nose section including radar, ESM and EO/IR systems. At a November 2024 Mitchelle Institute panel, Sheild AI’s Mike Benitez noted that computing requirements for CCAs were largely driven by the need to process large volumes of sensor data vs. execution and decision-making. Benitez noted that off-board sensor processing by platforms such as the F-35 was a possibility in order to keep costs affordable.  

Radar

RTX promotional video showcasing a Fury inspired CCA concept with a Phantomstrike 600 gallium nitride (GaN) active electronically scanned array (AESA) radar. Image Credit: RTX.

Primes have generally been incentivized to offer affordable, air cooled, multi-function, and modular sensors for CCAs such as the Raytheon APG-83. The PhantomStrike is a gallium nitride (GaN) active electronically scanned array (AESA) offered in the 900 series for tactical aircraft and 600 series for CCAs – this appears to  correspond to the size of the array as Raytheon previously reported 906 transmit receiver modules (TRMs) in 2021.

The company reports the system weighs 150 lb., has minimal SWaP-C requirements (air cooling), and costs half as much as an APG-79 or APG-82 ($1.7-$2.25 million). According to Raytheon Vice President of Secure Sensor Solutions Eric Ditmars, GaN “roughly doubles the detection range with the same size, same aperture and the same amount of power”.

CCAs could conceivably be fitted with multi-static radars to help U.S. fighters retain “first look, first shot, and first kill” capability against rapidly evolving threats from the People’s Liberation Army Air Force (PLAAF). The emergence of bi-static systems to counter emerging LO threats has been discussed since at least the 1990s. Historically, bistatic and multi-static systems have been hampered by limitations in available processing power. Very little has been disclosed about the recent progression of such systems but the Franco-German-Spanish Future Air Combat System (FCAS) next-generation fighter plans to employ bi-static and multi-static sensing with unmanned remote carriers. The Air National Guard 2023 weapon system modernization priorities book similarly calls for developing “multi-static and passive long-range kill chains” for the F-35.

Bi-static systems decouple the radar’s transmitter and receiver – enabling the receiver to see a different viewing aspect than would otherwise be possible given the host platform’s field of view. Multi-static systems also decouple the transmitter and receiver but can feature multiple transmitters all of which have viewing angles against prospective targets. Thus, a networked system of multi-static radars could more easily detect LO aircraft by constraining its ability to present only its most favorable viewing aspect i.e. more easily detecting adversary radar spikes. Such a system would have the added benefit of markedly improving the receiver aircraft’s survivability as it would not transmit any signals that could be detected by passive RF systems.

EO/IR

CCAs could also field networked IRSTs and electronic support measures (ESM) systems to facilitate passive long-range kill chains. Long-wavelength IR (LWIR) sensors like Lockheed Martin’s IRST operate at 6-14 µm are optimized to detect airframe emissions. Airframe heating from the front and rear is typically in the range of 30°C-230°C but the total IR signature level varies with earthshine and skyshine as well as aircraft speed and altitude. The angle and angle rate accuracy of multiple systems is typically used to triangulate a target. GA-ASI appears to have demonstrated this capability in 2022 using an MQ-20 equipped with Lockheed Martin’s IRST and a pair of F-5s.

Electronic Support Measures

Passive RF systems can similarly triangulate the location of airborne emitters by angle and time difference of arrival techniques. The overriding objective to lower CCA costs could make multi-function antennas (combining radar, ESM, communication, etc.) compelling.

In September 2024, SRC Inc. spoke to Aviation Week’s Steve Trimble about its Ghost Mantis multi-function array as a potential CCA solution. The system combines radar warning receivers, decoy emulation, and self-protection jamming. SRC’s guiding principle was affordability. SRC Assistant VP for Business Development Dave Toomey remarked,

“How much money is going to be available for that sensor payload on that sensor? We’ve heard Secretary Frank Kendall talk $20 million, we’ve heard $30 million, and somewhere in between those for the entire platform. We said, all right, let’s go down to the low single-digit millions and see what we can build…when you put mission systems on there, you put sensors on there, and when you put payloads on there, and you put [command-and-control] systems on there, you’re going to get up to that $25 to 30 million level really quick. And that’s how we settled on the low single-digit million-dollar range.

Datalink

CCAs are expected to feature Low Probability of Intercept (LPI) datalinks to enable Line of Sight (LoS) communication with manned fighters. Higher degrees of autonomation are expected to limit the need for continuous commands. However, multi-ship formations of CCAs may regularly transmit sensor data to manned assets for off-board processing.

LoS is necessary both in terms of emission control principles and assured communications. Legacy Beyond Line of Sight (BLoS) satellite communications (SATCOM) control systems in UAVs such as the MQ-9 and the RQ-4 are largely seen as unreliable in a near-peer conflict.

Modern LPI datalinks such as the F-35’s Multifunction Advanced Datalink (MADL) employ narrow beam transmissions between aircraft using a daisy chain system. The first aircraft sends the narrow beam signal to the second aircraft which in turn sends the signal throughout the rest of the formation. MADL is a key enabler of multi-ship sensor fusion – enabling sensor feeds across a formation of F-35s to be managed, collated, processed, and exploited.

CCAs could also act as a gateway to translate different datalink waveforms. In Dember 2020, a XQ-58 enabled communications between an F-35 and F-22.   

Weapons

Increment 1 is confirmed to be armed with AIM-120s. BFT released concept art for a “YFQ-XX” derivative of Fury featuring a pair of AIM-120s. As of the time of this writing, GA-ASI has not confirmed the weapons bay capacity of Gambit but a pair of AIM-120s is plausible. Kratos’ Demigorgon rival to the OBSS program had “significant offensive weapons volume” and the Gambit is most similar in class and configuration to Scaled Composite’s Model 437 which can carry a pair of AIM-120s internally.

In May 2023, then-CSAF Gen. Brown testified before the Senate Armed Servies Committee and confirmed the service planned to integrate the AIM-260 Joint Advanced Tactical Missile (JATM) with CCAs.

Other future weapons could include miniaturized air-to-air missile (AAM) concepts. The AFRL proposed Small Advanced Capabilities Missile (SACM) sought to field a missile half the size of the AIM-120, thereby doubling platform missile capacity, while achieving approximately the same range. In September 2018, Aviation Week reported the Air Force would fund flight testing for Lockheed Martin’s “half-raam” Cuda concept utilizing hit-to-kill and divert attitude control system (DACS) technologies developed from the Patriot’s PAC-3 interceptor. Raytheon separately unveiled its similarly sized Peregrine concept in 2019 which features a multi-mode seeker and blast fragmentation warhead. 

Engine 

Both Increment 1 entrants are powered by Williams International FJ44-4 engines which produce up to 3,600 lbf. The FJ44-4A used on the XQ-67A OBSS demonstrator was first selected to power the Cessna CJ4 in 2006. Anduril has confirmed its Fury is powered by the FJ44-4M which currently powers business jets as well as the Leonardo M-345 and L-39M Advanced Jet Trainers (AJTs).

The flyaway cost of a FJ44-4 is approximately $2.4 million. Engines have historically represented between 10-15% of an aircraft’s flyaway cost – implying a CCA cost range of $16-$24 million which closely aligns with the service’s lower-bound cost goal of $25-$33 million (1/4-1/3 the flyway cost of an F-35A).

An October 2023 Air Force RFI anticipates CCAs will field engines in the 3,000-8,000 lbf. thrust range available in 2028, 2030, and 2032. The RFI also specifies interest in options to, “increase range, reduce runway takeoff distance, increased market capability, increase power and thermal capacity, and increased payload.” 

However, the Air Force Life Cycle Management Center (AFLCMC) Propulsion Director John Sneden told Aviation Week in August 2024 that the Air Force was not necessarily locked into that thrust range. Future engine decisions would be informed by the overall requirements for the aircraft. 

Analysis: Roles & Missions

Key Highlights:

• A lack of real-world testing and evaluation, at least within the unclassified realm, has contributed toward a largely theoretical discussion of CCA Concept of Operations (CONOPS)
• Attritable CCA CONOPS are likely to be shaped by their cost, low to moderate levels of survivability as well as the range of their datalinks, sensors, and weapons
• The service’s desire for “minimum viable capability” appears to prefer platforms operating in “contested” airspace rather than the highest threat adversary sanctuary areas
  • The role of a “missile truck” or airborne magazine appears to be the most viable given the aforementioned limitations, subject to the range competition between USAF and PLAAF missiles 

Skyborg CONCOPS 2020 which effectively shows highly survivable assets as penetrating into adversary sanctuary areas, attritable UCAVs operating at the Forward Area Edge of Battle (FEBA), and non-stealthy assets operating in friendly sanctuary areas. Image Credit: AFRL via YouTube

Among the criticisms of the CCA program has been a lack of real-world experimentation and large-force exercises to validate key assumptions and requirements. This is largely a result of the service’s imperative to delivery capability quickly. The Air Force has tested individual capabilities with demonstrators and prototypes but has largely not been able to evaluate CCAs operationally in order to generate new tactics, techniques, and procedures (TTPs). The USAF plans to stand-up a CCA Experimental Operations Unit (EOU) in FY25 to redress these shortfalls. In the interim, the void in testing has largely led to theoretical and vague discussions of capabilities and CONOPS in the public realm.

The Air Force has provided few details as to its CCA CONOPs other than describing it as a “missile truck”, armed decoy, and sensor platform to assist manned fighters. The number of CCAs assigned per fighter has also remained fluid. Initially, each fighter would receive a pair of CCAs, now a range of three to eight CCAs is being discussed. Kendall has further added that CCAs are expected to operate within line-of-sight (LoS) to ensure C2 by manned platforms. Additional details can be inferred from the service’s desire for “minimum viable capability” for Increment 1, inherent design features of the CCA offerings, and prior discussions of both UCAV and “wingman” concepts.

U.S. thinktanks and industry have sought to frame service thinking by advocating for differing CCA CONOPs. In a study conducted in partnership with Kratos and GA-ASI, the Mitchell Institute conducted a U.S.-China wargame. Participants favored attritable $2-15 CCAs with basic LO and minimal sensors and weapons. Day 1 operations included the mass launch of CCAs with similar characteristics as GA-ASI’s longshot from B-52s and F-15EXs as well as from remote airfields. These CCAs would provide passive multi-static targeting to support F-22s and F-35s, carry a pair of AIM-120s and would create “aerial minefields” by acting as decoys for PLAAF KJ-500 AEW&Cs – to stimulate air defenses and expend adversary weapons. Day 2 operations would involve $15-40 million CCAs featuring improved LO, six AIM-120s, AESA radar, IRST and 3,000 mi range. These CCAs would “perform as pickets and air patrols in advanced of stealthy fighters during attacks on KJ-500s, SAGs [surface action groups], and opposing fighters”, with the broader aim to act as decoys.

Core to framing future CCA roles is the concept of stand-in and stand-off forces in relation to an adversary’s integrated air defense systems (IADS). Establishing air superiority at the initial stages of a conflict requires penetrating offensive counter-air (OCA) assets. These platforms would operate within the IADS, targeting critical adversary-enabling capabilities such as AEW&C, ISR, tankers, advanced fighters, and other high-value targets – denying a “sanctuary” area behind the forward area edge of battle (FEBA). Arguably, two effective stand-in or penetrating solutions exist at opposite sides of the cost-capability spectrum. The first features exquisite survivability such as LO platforms like the F-22 and future NGAD platform. The second emphasizes aggressive expendability to saturate adversary defenses such as the Raytheon ADM-160 miniature air-launched decoy (MALD).   

At the FEBA, where the threat is pronounced but not as severe as adversary rear-sanctuary areas, semi-survivable supporting assets can provide penetrating assets with long-range fires. For example, the F-15EX is equipped with either 12 AIM-120Ds or Hypersonic Attack Cruise Missiles (HACM). Behind the FEBA in friendly sanctuary areas, non-survivable enablers would contribute with stand-off capabilities including either extended-range munitions (such as AGM-158 JASSM-ERs launched from B-52s) or supporting capabilities such as BMC2, AEW&C, tanking, etc. As threat systems progress such as VHF radars, assets previously used for stand-in roles may shift to operating at the FEBA or from the FEBA to sanctuary areas. 

The “minimum viable” features of CCA Increment 1 appear to preclude any penetrating OCA roles. For example, ADAIR-UX called for a lower-bound RCS target of -10 dBSM or equivalent to that of a clean configuration Super Hornet. The ADAIR-UX concept was developed at least in parallel with Fury and may have informed its later development from Grackle by BFT.

Using figures provided by Piotr Budowski’s Russian Warplanes and Dan Katz’s State of Stealth as well as the following equation:

R₂/R₁ = (σ₂/σ₁)^1/4

R₁ = the maximum range from which a target of RCS σ₁ can be detected given the noise background.

R₂ = the new maximum range at which the target can be detected if the RCS is reduced from σ₁ to σ₂

The A-100E AEW&C’s S-band radar, S-400’s 92N62E fire control radar, and Su-35’s Irbis-E radar have a detection range of 64 nm (120 km), 54 nm (97 km), and 46 nm (85 km) respectively against a hypothetical attritable class CCA with a -10 dBSM signature in frontal search mode. Any external carriage of weapons would degrade RCS performance. In contrast, the F-22 has an RCS three orders of magnitude lower which corresponds to an almost 85% reduction in detection range or just eight nautical miles for the Su-35’s Irbis-E. This greatly simplified benchmarking of detectability appears to support the Skyborg graphic above denoting CCAs as operating at the FEBA. The low to moderate survivability of attritable CCAs coupled with their cost is also likely to shape their CONOPS.

CCAs are expected to be attritable, but service officials routinely frame sacrificing CCAs for a desired effect. In 2020, Roper described the Skyborg expenditure concept,

“Even though we call Skyborg an attritable aircraft, I think we’ll think of them more like reusable weapons… We’ll do whatever number of take-offs and landings they’re ‘spec’d’ for, and then will attrit them out of the force as targets and just buy them at a steady rate… I expect that the pilots, depending upon the mission decide: does the Skyborg return and land with them and then go to fight another day, or is it the end of its life and is going to go on a one-way mission?

The overlapping and combined arms nature of adversary threat systems behind the FEBA is such that attritable CCAs are arguably both not survivable enough to operate effectively and not expendable enough to justify high losses. $25-$33 million CCAs are considerably more costly than the $4 million Kratos XQ-58 or Skyborg concepts. Therefore, attritable CCAs likely have the greatest utility either at the FEBA in support roles or in sanctuary areas performing defensive counter air roles such as high value target protection.    

At the FEBA, CCAs could provide deep magazine capacity for stealthier assets operating farther forward. Employing CCAs in a “missile truck” role has several advantages. Existing 5^th generation fighters typically carry four to six beyond visual range (BVR) missiles owing to the inherent limits of their internal weapons bays. The CCAs wouldn’t necessarily require their own advanced sensors and could rely upon external targeting and mid-point guidance of penetrating assets. In many ways, this paradigm would be an evolution of current 5^th and 4^th generation fighter teaming. For example, F-22s have remained on station during large force exercises after expending their weapons to designate targets and provide situational awareness to 4^th generation platforms.

The effectiveness of the airborne magazine concept would likely be contingent upon (1) that the attritable CCAs would survive long enough to employ their missiles, (2) the missiles would retain a high probability of kill – despite the degraded launch parameters of CCAs (i.e. from moderate altitude at subsonic speed), and (3) that enough CCAs would survive these sorties to make the overall concept cost effective. The AIM-120D reportedly has a maximum range of approximately 161 km (87 nm) under ideal launch conditions. The AMRAAM has since been outclassed by the Leihua Electronic Technology Research Institute (LETRI) PL-15 which has an expected range of approximately 200 km (108 nm). The J-20 can currently carry four PL-15s internally, but the DoD’s annual China report notes potential upgrades to increase this capacity. The addition of the longer-range Lockheed Martin AIM-260 JATM to CCAs could markedly improve all three aforementioned aspects of the airborne magazine concept.

CCAs could also be used to simulate adversary defenses to draw or redirect adversary forces. Critically, because CCAs are armed, adversary forces are incentivized to engage them as they have the potential to cause harm – unlike current generation decoys like the ADM-160.

The ability of CCAs to contribute to long-range AMTI would be contingent upon their sensors and networking. Affordability pressures and the limited survivability of attritable CCAs is likely to incentivize modular, multi-function sensors with limited space, power, weight, and cooling requirements. Miniaturized air-cooled AESAs such as the Raytheon PhantomStrike 600 are one possibility. However, the utility of such sensors operating from lower threat areas to penetrating assets is unclear. 

Development of more exquisite bi-static or multi-static systems may exist though few details are available. Refer to the radar portion of the Features section for additional details.