Following years of research and development on artificial intelligence, on Wednesday the Air Force announced it had down-selected two companies that would build and test prototype aircraft in a competition for the service’s Collaborative Combat Aircraft (CCA) program due in 2026.
Once a winner is selected in 2026, the service has the ambitious goal of to begin receiving the first of 1,000 lower-cost ‘loyal wingman’ combat drones by 2028 to fly alongside its manned stealth fighters.
The selection has left the U.S.’s three biggest military aerospace giants—Boeing, Lockheed-Martin and Northrup-Grumman—out in the cold. But the two finalists, Costa Mesa, California-based newcomer Anduril Industries and established combat drone builder General Atomics Aeronautical Systems, are also a study in contrasts offering very different takes on what the CCA should resemble.
The Air Force had earlier narrowed the field to five companies, and indicated it might choose as many as three for the competition’s final phase—while also indicating it might lack funding to evaluate a third competitor for the competition’s final phase. Technically, the Air Force says its still open to considering outside offerings for CCA, presumably if self-funded.
Meanwhile, statements from the Air Force have suggested it’s interested in a more capable, and more expensive aircraft than the cheap and cheerful ‘attritable’ approach originally proposed for Loyal Wingman aircraft. The service wants more powerful turbofan engines ranging between 3,000 to 8,000 lbs. thrust, implying interest in supersonic flight. And it’s implied an expected unit cost of $20-30 million per aircraft, rather than the $10 million or less price originally proposed for loyal wingman drones.
That would make each CCA cost roughly as much as an advanced jet trainer—one-third to one-quarter the price of a new F-35 or F-15EX fighter, but still too costly to casually sacrifice.
Veteran aviation writer Bill Sweetman commented on social media that the finalists “…seemed to have a clearer vision of what the customer thought they wanted it. Whether that's actually what's needed is another matter, but I'll be interested to see the target costs.”
It’s worth noting that the CCA competition pertains specifically for the drone aircraft platform, which will then be mated to an ‘autonomy core’ furnished by the government. Elements of that core are being developed in DARPA’s Air Combat Evolution program (focused on within-visual range air-to-air combat) and the Air Force Research Laboratory’s Autonomous Air Combat Operations project (focused on basic flight maneuvers and beyond-visual range combat).
That’s not to say Anduril and General Atomics won’t likely incorporate some of their own autonomous technologies in proposed aircraft, which might include, say, capacity for basic flight maneuvers and collision avoidance, cooperative teaming behavior, takeoff and landing algorithms, or target recognition capabilities.
General Atomics’ Gambit
In a sense, General Atomics' products—the piston-engine MQ-1 Predator and larger, turboprop-powered MQ-9 Reaper—epitomize the popular concept of what armed drones are due to their extensive use by U.S. military and CIA in the 2000s and 2010s for surveillance and targeted strikes. These aircraft boasted long endurance and range, and could be piloted from anywhere on the globe thanks to their satellite datalinks.
Rather than proposing a single generalist aircraft, General Atomics’ pitch for CCA revolves around a family of mission-specialized “air dominance” aircraft called Gambit 1 through 4 it first unveiled in March 2022. These would share a 70% common airframe to maintain economies of scale.
Gambit 1 would serve as sensor platform optimized for long endurance and long-range reconnaissance/surveillance missions, including an ability to relay findings in real time via networking. A description noting use of discrete 'passive sensors' implies it will almost certainly incorporate infrared-search and track (IRST) system, though it might also support longer-range (but less stealthy) active radars to. Its long-endurance capability would be facilitated by use of a high-aspect wings and a “fuel-optimized engine.”
Meanwhile, a slower, lower endurance Gambit 2 would be designed for air-to-air combat carrying both sensors and missiles of its own. The latter would likely include the AIM-9X Sidewinder missile used for within-visual range engagement, but might extend to the heavier AIM-120 and future AIM-260 radar-guided beyond-visual range missiles.
Gambit 3 would “look like” Gambit 2, but be designed for serving as a simulated enemy ‘aggressor’ aircraft in training, including a capability to imitate the characteristics of different types of aircraft—a mission it could do more cheaply than manned aggressor jets.
Finally, Gambit 4 would boast a stealthy tailless, blended-wing body design reminiscent of the B-2 and B-21 stealth bomber. It would be focused performing ‘combat reconnaissance’ missions, or “long-endurance missions of a specialized nature.” This sounds like a platform able to penetrate the enemy air defense zone and loiter there for relatively long time to remote targeting and surveillance missions, though the word 'combat' could imply an attack capability too.
Interestingly, Grumman's Gambit 1 design was seemingly already selected by the Air Force for a mysterious program known as the Offboard Sensing System (OBSS) which resulted in an aircraft known as the XQ-67A which first flew in February this year. New rhetoric about OBSS’s role as fostering development of a ‘common genus’ for a wider range of mission-specialized drones very much seems in line with what we know about Gambit concept.
It's known that General Atomics turbofan-powered MQ-20 Avenger drones—rejected by the Air Force in 2013 as a successor to the Reaper—have been actively used as autonomy testbeds.
Anduril’s Fury
Anduril was founded in 2017 by tech enterpreuneur Palmer Luckey, who prior was known for developing Oculus Rift gaming goggles and has since attracted controversy for his support for Donald Trump and promotion of technological border security systems.
Arguably Anduril’s biggest product is its AI-enhanced Lattice system for networking sensors and classifying detected targets, which it has incorporated into fixed border surveillance towers as well as the Altius and Ghost drones and the Anvil and Roadrunner-M drone interceptors as well as platforms of other companies.
Anduril’s entry into the CCA competition was made possible when in September 2023 it bought out North Carolina-based Blue Force Technologies and its Fury loyal wingman aircraft it had developed. Fury, which has yet to have a confirmed first flight, is expected to be capable of high-subsonic speeds (Mach .95) and tolerating 9G maneuvers, and is touted as being affordable. It was initially intended as a ‘stealth adversary’ aircraft to help U.S. pilots practice combatting stealth fighters.
However, its modular, open-architecture system, which includes Lattice, means Fury could be adapted for actual combat roles. Anduril is also looking into boosting Fury to supersonic speed, initially for short bursts, and later even while cruising by adopting an advanced new engine.
However, Anduril has never before mass-produced a jet aircraft, though the company argues it’s able to carry out agile development and scalable production thanks to open-architecture design approach, its capability to manufacturer lightweight composite materials, and use of digital prototyping.
Teaching AIs to Fly Jet Fighters
Unlike prior armed drones, CCAs will be designed to function with a high degree of autonomy to facilitate their use on the forward edge of air combat, where datalinks to ground stations can’t be assured and where nearby manned aircraft crews may have their hands too full to micromanage what their ‘loyal wingmen’ are doing. Such convenient 'buddying' of manned with unmanned platform is known as Manned-Unmanned Teaming.
Additionally, some air combat tasks may require split-second reaction times that isn't feasible when command link latency and human reflex speeds are factored in.
A watershed in moving towards jet fighter autonomy occurred in DARPA’s 2020 Alpha Dogfight Trials in which an AI pilot trained by Heron Systems through billions of simulated aerial duels comprehensively defeated a human pilot at within-visual range dogfights involving simulated F-16s.
Recently, it was revealed that the Air Force since tested AI dogfighting capability onto an actual flying X-62A VISTA test aircraft (based on the F-16D jet) in 2023. Making that leap is a big deal, because outside of a simulator, AI has a much harder time accurately perceiving the world around it and must learn to deal with sometimes ambiguous sensor data and less-than-perfect knowledge of the ‘world state.’
Beyond CCA Increment 1
The Air Force has said it plans to procure at least 1,000 CCA Increment 1 jets, but might easily procure twice as many. Presently, the idea is for a pair of CCAs (not all of which will necessarily be armed) to accompany each of 300 F-35A stealth fighters and 200 forthcoming sixth-generation NGAD fighters that will eventually replace the F-22 in service
However, the Air Force has already dangled a still very vague ‘Increment 2’ CCA which is remains up for grabs by interested manufacturers, and which will be open to involvement with foreign partners. That seems almost certain to include Japan which has already begun collaborating with the U.S. on loyal wingman AI)
Coupled with the Navy and Marine Corps also looking to eventually procure loyal wingmen jets, that means a wider range of loyal wingmen and air combat autonomy agents will be on offer in the coming of years. That also applies to similar autonomous combats jets separately being developed across the world, particularly by China, the UK, France/Germany, Russia, South Korea and Turkey.
Sébastien Roblin has written on the technical, historical, and political aspects of international security and conflict for publications including 19FortyFive, The National Interest, MSNBC, Forbes.com, Inside Unmanned Systems and War is Boring. He holds a Master’s degree from Georgetown University and served with the Peace Corps in China. You can follow his articles on Twitter.
