CCA Explained: The AI Drone Wingmen About to Transform Air Combat

Imagine a formation of fighters entering contested airspace. The lead aircraft is an F-35, flown by a human pilot. Flanking it on either side, holding perfect station at Mach 0.9, are two aircraft with no one inside them. They carry sensors, weapons, and electronic warfare suites. They are controlled by artificial intelligence. They cost a fraction of the manned jet they fly beside, and if one is shot down, no pilot is lost. This is the Collaborative Combat Aircraft program, and it is no longer a concept. It is in flight testing now.

The United States Air Force is building a new class of AI-controlled drone wingmen designed to fly alongside crewed fighters like the F-35 Lightning II and the forthcoming F-47 sixth-generation fighter. These Collaborative Combat Aircraft, or CCAs, are intended to absorb risk, extend sensor coverage, carry additional weapons, and conduct electronic warfare, all while being cheap enough that losing one in combat is an acceptable cost of doing business.

The CCA program represents the most significant shift in American air combat doctrine since the introduction of stealth. It changes how the Air Force thinks about mass, cost, attrition, and the relationship between humans and machines in warfare. This article explains what CCAs are, who is building them, how they will fight, and why the program matters.

What Is a Collaborative Combat Aircraft?

A Collaborative Combat Aircraft is an autonomous or semi-autonomous unmanned combat air vehicle designed to operate in close coordination with a manned fighter. The "collaborative" distinction is important: unlike traditional drones such as the MQ-9 Reaper, which operate independently on their own missions with remote human pilots, CCAs are designed to fly as part of a mixed formation, taking tactical direction from a human pilot in a nearby crewed aircraft.

The concept is sometimes called "loyal wingman," a term that captures the relationship. The CCA does not replace the human pilot. It extends what the human pilot can do, functioning as an additional set of sensors, an additional weapons magazine, or an expendable asset that can absorb enemy fire or fly into the most dangerous threat zones so the crewed jet does not have to.

A defining feature of CCAs is that they are designed to be "attritable." This is Pentagon jargon for a platform cheap enough that commanders can accept losing it in combat without the loss being strategically significant. The Air Force has targeted a unit cost of roughly $3 million to $10 million per CCA, depending on configuration and capability level. For comparison, an F-35A costs approximately $80 million, and the true cost of modern fighters is even higher when you factor in lifecycle expenses. Losing an $8 million CCA is a budget line item. Losing an $80 million F-35 and its trained pilot is a strategic event.

This affordability is what makes the math work. The Air Force does not plan to buy a handful of CCAs. It plans to buy over a thousand, enough to pair two or more with every F-35A and F-47 in the fleet. At those numbers, even with attrition, the total fleet grows dramatically without a proportional increase in cost or, critically, in the number of human pilots needed.

CCAs sit on a different part of the autonomy spectrum than either traditional remotely piloted aircraft or fully autonomous weapon systems. An MQ-9 Reaper is controlled by a human pilot and sensor operator via satellite link, with the humans making every flight and targeting decision. A hypothetical fully autonomous weapon would operate entirely without human input. CCAs occupy the middle ground: they fly themselves, navigate themselves, and execute tactical maneuvers autonomously, but they operate under the direction of a human mission commander in the lead aircraft. The pilot sets objectives and rules of engagement; the CCA figures out how to accomplish them.

The Two CCA Vendors

In 2024, the Air Force selected two companies to build the first CCAs under what it calls "Increment 1," the initial production variant focused on core autonomous flight, sensor integration, and weapons delivery. The two vendors are Anduril Industries and General Atomics Aeronautical Systems.

Anduril Industries is building the CCA based on its Fury drone, a jet-powered unmanned combat air vehicle the company originally developed with internal funding. The Anduril design has been designated the YFQ-44A under the military's new naming convention for unmanned combat aircraft. According to Air & Space Forces Magazine, the YFQ-44A completed its first flight on October 31, 2025, flying semi-autonomously from a test range. The Fury is designed around Anduril's Lattice software platform, which provides the AI backbone for autonomous decision-making, sensor fusion, and multi-vehicle coordination.

General Atomics, the company behind the MQ-9 Reaper and MQ-1C Gray Eagle, is building a separate CCA design designated the YFQ-42A. General Atomics brings decades of experience building and sustaining large unmanned aircraft fleets for the Department of Defense. According to Air & Space Forces Magazine, the YFQ-42A completed its first flight in August 2025, making it the first CCA to fly.

The FY2026 defense budget request allocates over $804 million specifically for the CCA program, according to Air Force budget documents. Total planned spending between FY2025 and FY2029 exceeds $8.9 billion. This funding covers development, flight testing, initial production, and the AI software that makes autonomous teaming possible.

The Air Force has structured the program in increments. Increment 1, now in flight testing, delivers the foundational CCA platform: an autonomous jet-powered aircraft with internal weapons bays, sensor suites, and the ability to fly in coordinated formations with crewed fighters. Increment 2, still in the design phase, is expected to deliver more specialized variants optimized for specific mission sets such as electronic warfare, extended-range sensing, or heavier weapons loads. The increment approach allows the Air Force to field an initial capability quickly while continuing to develop more advanced versions.

How CCA Will Fight

The tactical employment of CCAs centers on what the Air Force calls "manned-unmanned teaming," or MUM-T. In practice, this means a human pilot flying an F-35 or F-47 will command a small group of CCAs, typically two to four, as part of an integrated combat formation. The pilot does not fly each CCA individually. Instead, the pilot assigns objectives, and the CCAs autonomously determine how to accomplish them.

Communications between the crewed aircraft and its CCAs will rely on a combination of existing and new data links. The Air Force has referenced Link 16, the standard NATO tactical data link, as a baseline, but CCAs will also use more advanced and harder-to-jam networks. The F-35's Multifunction Advanced Data Link (MADL), a low-probability-of-intercept link designed for stealth operations, is expected to play a central role. New mesh networking protocols are also in development that would allow CCAs to communicate with each other and with the lead aircraft even in heavily jammed environments, redistributing data across the network if individual links are disrupted.

The Air Force envisions CCAs performing several distinct mission types within a single sortie, adapting roles as the tactical situation evolves.

The flexibility to shift between these roles within a single mission is a key advantage. A CCA might begin a sortie screening ahead with its sensors, switch to an electronic warfare role as the formation approaches a defended area, and then serve as a weapons carrier during the engagement itself. This adaptability, governed by AI and directed by the human pilot, is what distinguishes CCAs from previous unmanned systems that were designed for single, fixed missions.

The AI Brain

The technology that makes CCAs possible is not primarily the airframe or the engine. It is the artificial intelligence that allows an unmanned aircraft to fly in combat formation, make tactical decisions in real time, and coordinate with other aircraft without a human controlling every input. Building that AI has been a multi-year effort spanning several Defense Department programs.

The most significant precursor is DARPA's Air Combat Evolution program, known as ACE. Launched in 2019, ACE set out to develop AI systems capable of performing within-visual-range air combat, essentially dogfighting, against human pilots. The program achieved a series of milestones that would have seemed implausible a decade earlier.

In 2020, an AI agent developed by Heron Systems defeated a human F-16 pilot five to zero in simulated dogfights during the AlphaDogfight competition, a DARPA-organized event designed to benchmark autonomous air combat performance. The AI demonstrated superior reaction time and precision, though it operated in a simplified simulation environment. Later phases of ACE moved from simulation to subscale aircraft, flying AI-controlled jets against each other and against human-piloted aircraft in real-world flight tests at Edwards Air Force Base.

By 2024, DARPA confirmed that an AI-controlled F-16, designated the X-62A VISTA (Variable In-flight Simulator Test Aircraft), had flown autonomous air combat maneuvers against a human-piloted F-16 at Edwards. The AI managed the aircraft's flight controls, executed tactical maneuvers, and responded to the opposing aircraft's actions without human input on stick and throttle. A safety pilot remained in the X-62A cockpit with override capability, but according to DARPA press releases, the AI handled the flying.

The CCA's onboard AI builds on this foundation but extends well beyond dogfighting. The autonomous system must manage navigation, formation keeping, threat detection, sensor employment, weapons delivery, communications, and self-preservation, all simultaneously, and all while coordinating with the human mission commander and other CCAs in the formation.

The command relationship is described as "human-on-the-loop" rather than "human-in-the-loop." The distinction matters. In a human-in-the-loop system, no action occurs without explicit human approval for each step. That model works for an MQ-9 Reaper flying a deliberate surveillance mission, but it is too slow for air combat where decisions must be made in fractions of a second. In the human-on-the-loop model used for CCAs, the human pilot sets the mission objectives, rules of engagement, and boundaries. The AI executes within those boundaries autonomously. The pilot can intervene, redirect, or override at any time, but the AI does not wait for permission before maneuvering, deploying sensors, or taking defensive action.

Whether the AI will have authority to release weapons autonomously remains one of the most sensitive policy questions surrounding the program. Current Department of Defense policy, articulated in DoD Directive 3000.09, requires "appropriate levels of human judgment" in the use of lethal force. The Air Force has not publicly stated whether CCAs will be authorized to fire weapons without real-time human approval, and this decision may vary by mission type, threat environment, and rules of engagement. What is clear is that the technology to do so exists; the question is one of policy, not capability.

Boeing's XQ-58 Valkyrie: The Proof of Concept

The CCA program did not emerge from nothing. Its conceptual and technological foundations were laid by a series of earlier programs, the most important of which was the Kratos XQ-58A Valkyrie.

The Valkyrie first flew in March 2019 at Yuma Proving Ground, Arizona. Built by Kratos Defense under contract to the Air Force Research Laboratory, the XQ-58A was designed as a low-cost, jet-powered unmanned combat air vehicle that could operate autonomously or in coordination with manned fighters. It was intended to demonstrate the core concept that would become CCA: an affordable, attritable drone wingman.

The XQ-58A demonstrated several capabilities critical to the CCA vision. It flew autonomously using pre-programmed mission plans. It launched from a ground-based catapult system, eliminating the need for a runway. It carried internal weapons bays capable of housing small-diameter bombs or air-to-air missiles. And at an estimated unit cost of roughly $2 million to $4 million in production quantities, it proved that a jet-powered combat aircraft could be built at a fraction of traditional fighter costs.

The Air Force Research Laboratory integrated the Valkyrie into its broader Skyborg program, which developed the autonomy software layer for unmanned wingman concepts. Skyborg focused on creating a common AI "brain" that could be installed in different airframes, providing autonomous flight, basic tactical behaviors, and the ability to coordinate with other aircraft. Flight tests under Skyborg demonstrated that an AI-controlled aircraft could fly formation, respond to dynamic re-tasking, and operate safely in shared airspace with manned jets.

These Skyborg and Valkyrie demonstrations directly informed the CCA requirements. The Air Force learned what worked, what did not, and what level of autonomy was achievable with current technology. When the CCA program was formally launched, it built on thousands of flight hours and years of software development from these precursor efforts. The Valkyrie proved the concept was viable. CCA is the operational realization.

Why CCA Changes the Math

The strategic logic of CCA is best understood through force multiplication. Consider a traditional fighter sortie: one F-35, one pilot, carrying six to eight weapons internally, with the sensor suite of a single aircraft. Now consider the same F-35 accompanied by three CCAs. The formation now has four aircraft sharing sensor data across a much wider area. It carries perhaps twenty weapons across the four platforms. It can distribute its assets so that CCAs absorb the greatest risk while the irreplaceable crewed fighter operates from relative safety.

The cost comparison reinforces the logic. Three CCAs at $8 million each cost $24 million. One additional F-35 costs $80 million plus the multi-year, multi-million-dollar investment in training its pilot. The three-CCA option provides more total combat capability at less than one-third the cost, with zero additional pilot risk. Over a fleet of hundreds of fighters, this arithmetic transforms the Air Force's force structure.

Attrition tolerance is perhaps the most profound change. In every modern air campaign, the Air Force has designed its operations to minimize aircraft losses because every lost jet represents an enormous financial investment and, potentially, a lost pilot. This drives conservative tactics: standoff weapons, extensive suppression of enemy air defenses before penetrating contested airspace, and avoidance of missions where losses are likely.

CCAs fundamentally alter this calculus. If a formation can send attritable drones into the most dangerous environments, absorbing surface-to-air missile shots that would otherwise target crewed jets, the formation can accept a level of attrition that was previously unthinkable. An adversary that shoots down three CCAs has expended three expensive SAMs, revealed the positions of three launchers, and destroyed $24 million worth of hardware. The crewed fighter behind them is unscathed, its pilot alive, and its own weapons now targeted at the air defense sites that just exposed themselves.

The industrial base advantages are also significant. Building a thousand CCAs creates production volume that drives down per-unit costs and sustains manufacturing lines. It provides work for a broader defense industrial base, including newer entrants like Anduril alongside established primes like General Atomics. And it creates a production infrastructure that can surge in wartime, building replacement CCAs far faster than any adversary could replace the expensive systems used to shoot them down.

This is the logic that has driven the Air Force to invest nearly $9 billion in CCA development through the end of the decade. It is not about replacing human pilots. It is about making each human pilot dramatically more effective by surrounding them with affordable, autonomous, expendable teammates. As military drones continue to reshape warfare, CCAs represent the most ambitious application of that transformation to high-end air combat.

Challenges and Risks

The CCA program faces substantial challenges that could delay fielding, increase costs, or limit operational effectiveness. Acknowledging these risks is essential for understanding the program realistically rather than through the lens of promotional materials.

Trust in AI. The most fundamental challenge is convincing fighter pilots and commanders to trust AI-controlled aircraft flying in close formation during combat. A CCA that makes an unexpected maneuver could collide with the crewed fighter it is supposed to protect. An AI that misidentifies a friendly aircraft as a threat could create a fratricide incident. Building trust requires thousands of hours of testing that demonstrate the AI behaves predictably and safely across the full range of combat conditions, including conditions the AI was not explicitly trained for. This verification and validation process is time-consuming and may reveal problems that require significant software rework.

Communications in contested environments. CCAs depend on data links to receive direction from the human mission commander and to share sensor data with the formation. Near-peer adversaries like China and Russia invest heavily in electronic warfare specifically designed to jam, disrupt, or spoof these links. If communications are degraded, CCAs must fall back on pre-programmed behaviors or fully autonomous decision-making, which increases the risk of unintended actions. Ensuring that CCAs behave appropriately across the full spectrum of communications conditions, from perfect connectivity to total jamming, is a defining engineering challenge. The broader lessons from drone swarm development suggest that communications resilience remains one of the hardest problems in autonomous teaming.

Logistics and sustainment. Adding over a thousand new aircraft to the Air Force fleet, even inexpensive ones, creates significant logistics demands. CCAs need maintenance facilities, spare parts supply chains, fuel, weapons loading equipment, and trained ground crews. Basing infrastructure must accommodate additional aircraft. The Air Force's existing maintenance and logistics system is already strained supporting current platforms. Whether the service can absorb a thousand-plus new aircraft without degrading readiness across the fleet is an open question.

Timeline risk. The Air Force has set an aggressive timeline for CCA, targeting initial operational capability in the 2028-2029 timeframe. Defense acquisition programs routinely slip, and new technologies are particularly susceptible to schedule delays. The AI software, in particular, may require more development time than currently planned if testing reveals shortcomings in autonomous decision-making or safety. Budget pressures, changing political priorities, or technical setbacks could push fielding to the right.

Testing requirements. The CCA program must demonstrate that autonomous aircraft can operate safely in shared airspace, coordinate effectively with manned fighters across diverse mission types, employ weapons accurately, and behave predictably in degraded conditions. Each of these areas requires extensive flight testing. The test infrastructure, including ranges, instrumentation, and adversary simulation systems, must be available and adequate. Compressing the test program to meet the timeline introduces risk that problems will be discovered after fielding rather than before.

Timeline: When Will CCAs Fly?

Both CCA Increment 1 designs are already flying. The General Atomics YFQ-42A completed its first flight in August 2025, and the Anduril YFQ-44A followed in October 2025, according to Air & Space Forces Magazine. These initial flights demonstrated basic autonomous flight capabilities and marked the transition from design to physical testing.

The Air Force's stated timeline targets initial operational capability for CCA around 2028 to 2029. This means the service expects to have at least one CCA variant certified for operational use, integrated with F-35 data links and mission systems, and assigned to an operational fighter squadron by that date.

Full integration of CCAs across the fighter fleet, including pairing with the F-47 as that aircraft enters service, is expected in the early 2030s. By that point, the Air Force envisions routine operations where mixed formations of crewed and uncrewed aircraft are the standard rather than the exception.

Between now and IOC, the program must complete an intensive flight test campaign. This includes expanding the flight envelope, testing autonomous formation flight with crewed fighters, demonstrating weapons delivery, validating communications in jammed environments, and proving the AI can handle degraded and unexpected conditions. The Air Force plans to conduct these tests at Edwards Air Force Base and other test ranges through 2027 and 2028.

Increment 2 CCAs, with more specialized mission capabilities, are expected to begin flight testing in the late 2020s and reach operational status in the early to mid-2030s. The staggered approach allows the Air Force to field a basic capability quickly while continuing to develop more advanced variants.

It is worth noting that defense acquisition timelines are aspirational until they are achieved. The CCA program benefits from years of precursor work under Skyborg and Valkyrie, which reduces some technology risk. But integrating autonomous aircraft into operational fighter formations at scale is unprecedented, and the history of defense programs suggests that unforeseen problems will arise. The timeline should be treated as a best-case projection, not a guarantee.

Key Takeaways