The USAF is accelerating its autonomous combat drones to create affordable mass against China, with the F-22 serving as the first pack leader.
In Summary
The US Air Force is accelerating the integration of Collaborative Combat Aircraft, autonomous combat drones designed to accompany manned aircraft. The program is moving fast. General Atomics’ YFQ-42A has flown using Collins Aerospace’s Sidekick autonomy software. Anduril’s YFQ-44A Fury integrates Shield AI’s Hivemind. An F-22 Raptor has already controlled an MQ-20 Avenger in flight, utilizing data links and software-defined radios. The message is clear: the USAF wants to provide its fighter jets with a swarm of wingmen capable of detecting, jamming, striking, or absorbing risk. This objective is not driven by technological indulgence; it is strategic. Facing China, the United States suffers from a shortage of mass, pilots, industrial time, and available aircraft. Autonomous drones promise a less expensive, faster, and more flexible response. However, their operational employment still raises major questions.
The CCA Program Becomes the Central Laboratory for American Air Combat
The US Air Force is no longer just talking about loyal wingmen. It is now discussing an entirely new category of combat aircraft. In March 2025, it assigned two official designations to its first Collaborative Combat Aircraft, or CCA, prototypes: YFQ-42A for General Atomics and YFQ-44A for Anduril. The details matter. The letter F indicates a fighter mission. The letter Q designates an unmanned aircraft. The prefix Y marks its status as a prototype.
This choice is political as well as military. The USAF wants to show that these platforms are not traditional ISR drones, nor are they simple flying targets. They are autonomous combat drones, designed to operate alongside manned fighters in a contested environment. They must be capable of conducting air-to-air, air-to-ground, electronic warfare, intelligence, targeting, and reconnaissance missions.
The program fits within the logic of Next Generation Air Dominance. The future F-47 fighter, the F-35, the F-22, and CCAs are meant to form a family of systems. The idea is not to replace the pilot. It is to multiply their effectiveness. A single manned aircraft could command, supervise, or coordinate multiple unmanned platforms, each carrying sensors, jammers, or munitions.
The earlier working hypothesis mentioned by the USAF involved approximately 1,000 CCAs. This figure came from a simple calculation: two drones for each of the 500 advanced fighters planned for the future force. Even if final volumes depend on the budget, technical maturity, and congressional decisions, the order of magnitude says it all. Washington is not looking for a mere demonstrator. Washington is looking for affordable mass.
The F-22 Could Become the First Operational Pack Leader
The F-22 Raptor is not the newest aircraft in the American arsenal. However, it remains the USAF’s most specialized air superiority fighter. Its stealth, speed, operational altitude, and ability to penetrate dangerous airspace make it a logical candidate to experiment with controlling autonomous drones.
On October 21, 2025, General Atomics, Lockheed Martin, and L3Harris conducted a notable test at the Nevada Test and Training Range. An F-22 controlled an MQ-20 Avenger in flight. The test utilized L3Harris’s BANSHEE data links, Pantera software-defined radios, and an open radio architecture integrated into the F-22. The pilot transmitted commands to the MQ-20 via a tablet interface and the Raptor’s GRACE module.
This test did not transform the F-22 into a flying aircraft carrier. It did, however, demonstrate an essential function: a pilot can give instructions to an unmanned aircraft from a stealth fighter without routing through a traditional ground station. This is a practical breakthrough. In a war against an equipped power, ground stations will be jammed, targeted, or saturated. Command must be distributed.
The F-22 has another advantage. Its fleet is limited, but its pilots are highly specialized in complex air combat. Integrating them early into the development of CCAs allows for the creation of credible tactics. Engineers can write software, but only aircrews know where to inject autonomy into a real mission: at what point to delegate, when to retake control, what information to display, and what level of trust to accept.
The Raptor could therefore become the first operational “quarterback” for CCAs. It will likely not be alone. The F-35, and eventually the F-47, will also be destined to pilot autonomous wingmen. But the F-22 offers an interesting bridge between the current generation and the next air combat architecture.
The Sidekick Software Shows That Autonomy Is Becoming Modular
General Atomics’ YFQ-42A crossed an important milestone in February 2026. The aircraft completed a semi-autonomous mission using Sidekick, the collaborative autonomy software from Collins Aerospace, a subsidiary of RTX. The test lasted over four hours. Sidekick was integrated into the YFQ-42A’s flight control system thanks to the Autonomy Government Reference Architecture, or A-GRA.
This point is essential. The USAF does not want to lock its future drones into a proprietary architecture. It wants software capable of being integrated quickly, swapping platforms, and evolving without rebuilding the entire aircraft. A-GRA serves precisely this purpose. It provides a common structure to connect autonomy, mission systems, and flight controls.
Sidekick does not just follow a pre-programmed route. The software must interpret mission orders, adapt the vehicle’s behavior, and execute tactical tasks. During the test, a human operator on the ground sent various commands to the YFQ-42A. According to General Atomics, the aircraft executed them with precision.
The difference between autopilot and mission autonomy is fundamental. An autopilot maintains an altitude, speed, or trajectory. Mission autonomy chooses how to achieve an objective under constraints. It must manage airspace, fuel status, threats, rules of engagement, friendly positioning, and communication quality.
In real combat, the human pilot will not be able to micromanage each drone. They will not have the time to fly three secondary aircraft while trying to survive surface-to-air missiles, opposing fighters, and jamming. Autonomy must therefore execute an intent, rather than wait for every single gesture.
The Hivemind Software Illustrates the Battle of Digital Brains
Anduril’s YFQ-44A Fury is following a different industrial path. It integrates Hivemind, the autonomy software from Shield AI, while also relying on Anduril’s Lattice ecosystem. In February 2026, Shield AI announced that Hivemind had been selected as a mission autonomy provider to support technology maturation and risk reduction efforts for the CCA program.
Hivemind is presented as an artificial intelligence pilot capable of perceiving, deciding, and acting. Shield AI emphasizes one point: its software is not a simple autopilot. It can modify a trajectory, avoid forbidden zones, react to unforeseen situations, and pursue a mission with little to no human intervention.
This promise meets a very concrete need. In the Pacific, a war against China would unfold over immense distances. American bases in Guam, Japan, or the Philippines would be threatened. Satellite communications could be jammed. Data links could be degraded. A drone that permanently depends on a human operator would quickly become fragile.
Autonomy must therefore function even when the network degrades. This does not mean letting a machine decide everything on its own. It means giving the system enough intelligence to pursue a bounded mission whenever contact with the pilot becomes intermittent.
The real debate will center on the level of authority granted to this software. Detecting, classifying, proposing, maneuvering, or jamming do not raise the same questions as firing a weapon. The USAF will have to establish clear safeguards. The closer autonomy gets to the employment of force, the more politically and legally sensitive human validation becomes.
Munitions Make CCAs Something More Than Simple Sensors
A CCA only holds military value if it delivers a useful effect. A drone accompanying an F-22 or an F-35 can serve as a communications relay, an advanced sensor, or a jammer. However, the toughest challenge remains armament.
The USAF is studying the integration of air-to-air missiles and offensive payloads onto its CCAs. The goal is simple: increase the number of munitions available in air combat. An F-22 can carry a limited number of missiles in its internal bays to preserve its stealth. An F-35 faces the same constraints. In a massive engagement against China, this magazine depth becomes insufficient very quickly.
An armed CCA can play several roles. It can fly ahead to push sensors further forward. It can force the adversary to reveal their radars. It can carry additional missiles. It can draw enemy fire. It can also launch a munition upon human command while the manned aircraft remains further back or more hidden.
This logic changes the economics of combat. A missile fired by a less expensive drone does not become any less dangerous. But the loss of the carrier platform becomes far more acceptable than losing an F-22 and its pilot. It is brutal, but it is the logic of modern attrition. In a high-intensity war, no serious military command can assume all its aircraft will return.
The question of munitions will also have an industrial impact. Producing drones without producing enough missiles would be useless. CCAs must be designed in tandem with stockpiles of AMRAAMs, future missiles, decoys, jammers, and modular payloads. The drone is not an isolated solution; it is a link in a chain of production, maintenance, data, and munitions.
China Imposes an Urgency That the USAF Can No Longer Bypass
The strategic engine of the program is China. Beijing has built a dense A2/AD bubble in the Western Pacific. It combines ballistic missiles, cruise missiles, radars, modern combat aircraft, surface-to-air systems, electronic warfare, cyberattacks, and space capabilities. The objective is to prevent American forces from freely approaching Taiwan or the first island chain.
In this context, the United States has a problem of distance and mass. Bases are far away. Tankers are vulnerable. Runways can be struck. Manned fighters are expensive and take a long time to produce. Pilots require years of training. China, by contrast, can concentrate its forces near its own territory.
CCAs respond to this constraint through numbers. They allow for the addition of platforms without recruiting a pilot for every aircraft. They complicate enemy targeting. They increase the number of sensors and weapon carriers. They can be used for highly risky missions without directly exposing a crew.
This is the meaning behind the phrase affordable mass. It does not mean “cheap” in a civilian sense. A CCA will remain an advanced system. Estimates often place the cost around one-third of the price of a manned fighter, with goals aimed below the cost of a recent F-35. But it must be affordable enough to be bought in volume, used without political paralysis, and replaced faster than a stealth fighter.
To be frank: the USAF is not just looking for an innovation. It is looking to correct a structural weakness. Its fleet is too small for a prolonged conflict against a peer adversary. CCAs are an attempt to rebuild depth without waiting twenty years.
The Pilot Shortage Makes Autonomy Even More Strategic
The lack of pilots is not a secondary issue. The USAF has suffered from a chronic deficit for more than a decade. In 2024, the shortage was estimated at around 1,850 pilots, including more than 1,100 fighter pilots, according to several specialized analyses. In April 2026, the Vice Chief of Staff of the US Air Force also acknowledged to Congress an acute problem with retaining fighter pilots.
Training alone is not enough to solve the problem. The USAF aims for 1,500 trained pilots per year in its modernized pipeline. However, it must also retain experienced pilots. Yet, operational demands, repeated deployments, administrative tasks, and civilian competition take a heavy toll. A qualified fighter pilot cannot be replaced quickly.
CCAs offer a partial answer. An autonomous drone does not replace an experienced flight lead. But it can allow a single pilot to command more effects. Instead of sending four manned fighters, the USAF could deploy two manned aircraft and several CCAs. The number of exposed pilots decreases; the tactical mass increases.
This equation is attractive, but it has a limit. It shifts the cognitive load onto the remaining pilot. Commanding multiple platforms, monitoring threats, understanding AI recommendations, and making firing decisions can become overwhelming. The quality of the human-machine interface will therefore be decisive. If the system adds complexity to the cockpit, it will fail. If it reduces mental workload, it will genuinely transform air combat.
Maintenance Costs Explain the Program as Much as the Technology
CCAs are also a response to maintenance costs. American aircraft are aging. Fighter fleets demand significant maintenance hours. Delays in depots reduce aircraft availability. The Government Accountability Office reported again in 2026 a significant increase in maintenance backlogs at US Air Force depots since 2019.
A manned fighter is expensive to fly, modernize, and maintain. The airframe, engines, sensors, ejection seats, life support systems, safety requirements, and pilot environment must all be preserved. A CCA can be designed with different trade-offs. It does not need a cockpit, a canopy, onboard oxygen, or human protection systems. It can fly less frequently in live training if a significant portion of preparation is done via simulation.
This does not make CCAs free. Engines, sensors, software, data links, and munitions will be expensive. The maintenance of an armed autonomous system will remain demanding. However, the economic model can be different. The USAF hopes for lower ownership costs, faster production, and lighter maintenance requirements compared to manned fighters.
The industrial question remains open. Producing a few prototypes is one thing; producing hundreds of reliable, armed, cyber-secured drones capable of flying through jamming and being maintained at dispersed bases is another. The real test will not just be the first flight. It will be the production ramp-up.
Integration Will Decide Success Far More Than Pure Performance
The trap would be to judge CCAs solely on their speed, range, or payload. These metrics matter, but the core of the issue is integration. A CCA must talk to an F-22, an F-35, a command center, a tanker, an electronic warfare system, and perhaps another drone. It must receive simple orders. It must transmit useful data. It must remain controllable under jamming. It must respect the rules of engagement.
This is why open architectures are so important. The USAF does not want to depend on a single contractor for the aircraft, software, radio, weapons, and updates. It is seeking a modular logic. A General Atomics aircraft must be able to receive Collins software. An Anduril aircraft must be able to carry Hivemind. Links must be non-proprietary as much as possible.
This approach is good on paper, but difficult in practice. Defense contractors protect their technologies. Military software must be certified. Updates must be secure. Cybersecurity becomes critical. A fleet of connected autonomous drones also expands the digital attack surface.
Success will therefore depend less on a single engineering breakthrough than on engineering discipline. CCAs will need to be tested, broken, corrected, and retested. Pilots will have to learn to trust them without blindly obeying them. Air staffs will have to write precise employment concepts. Logisticians will have to follow through. It is a revolution, but a revolution that will be won in the details.
Future Air Combat Will Turn on Trust Between Man and Machine
CCAs pose a simple question: how far can a pilot trust an armed machine? The answer will not be binary. It will depend on the mission. For reconnaissance, autonomy can be broad. For jamming, it can be bounded. For weapons release, it will likely remain more strictly controlled, especially in the early years.
The USAF is moving carefully, but it is moving fast. The Sidekick, Hivemind, and F-22/MQ-20 tests show that the building blocks already exist: autonomy software, open architectures, software-defined radios, pilot interfaces, jet-powered drones, and flight testing. Moving to real combat will require more than just demonstrations. It will require proving reliability, safety, tactical effectiveness, and economic viability.
The hard truth is clear. The United States cannot confront China with a fleet that is too rare, too expensive, and too dependent on pilots who are difficult to train. CCAs are a logical response to this impasse. They will not guarantee air superiority. They may even create new risks, particularly cyber, doctrinal, and political. But they offer a path that the USAF can no longer ignore.
The F-22, designed to dominate the skies alone, could thus inaugurate the era where a fighter no longer fights alone. The symbol is powerful. The USAF’s most exclusive stealth aircraft is becoming the testbed for a more collective, distributed, and automated form of warfare. Tomorrow, air superiority will not depend solely on the best pilot or the best aircraft. It will depend on the ability to make man, machine, software, and munition act together in a shorter timeframe than the adversary.
Live a unique fighter jet experience