The traditional defense acquisition model is officially dead, and the U.S. Air Force just signed its death warrant. By bypassing the usual decade-long bureaucratic slog, the military just awarded rapid production contracts for its first-generation Collaborative Combat Aircraft (CCA) program. They didn't just pick one winner. They picked two.
Anduril Industries and General Atomics just secured engineering, manufacturing, development, and production contracts to build the first fleet of semi-autonomous drone wingmen. The goal? To get more than 150 of these combat-capable uncrewed fighters on the tarmac by the end of the decade, scaling up to an eventual fleet of roughly 1,000 aircraft.
This isn't a slow-rolling test program. It's an aggressive, high-stakes sprint to solve an existential problem: America is running out of mass in the air.
The Dual-Vendor Gamble That Decouples Hardware From Brains
For decades, the Pentagon fell into a predictable trap. They would spend 20 years and hundreds of billions of dollars to design a single, incredibly complex stealth fighter like the F-35, locking themselves into one contractor and a frozen architecture. If a component failed or a supplier went under, the entire program choked.
The CCA program breaks that model completely. By funding both General Atomics’ FQ-42A Dark Merlin and Anduril’s FQ-44A Fury, the Air Force is intentionally building a split initial fleet.
This isn't an indecisive compromise. It is a deliberate strategy to force continuous competition and diversify the physical supply chain. The aircraft are dropping their "Y" prototype prefixes and heading straight toward full-scale serial production four months ahead of schedule.
But the real magic isn't the airframes. It's how the Air Force is treating the software.
The military has explicitly separated the physical drones from the mission autonomy software that flies them. Think of it like Android running on phones built by different companies. The Air Force established a vendor pool of six companies to build the software brains: Anduril, General Atomics, Lockheed Martin, Northrop Grumman, RTX Collins Aerospace, and Shield AI.
To speed things up, three of those firms—Anduril, Shield AI, and Collins Aerospace—received immediate production options to accelerate software delivery. They will compete in intense six-month phases, leading to a primary software provider selection by summer 2027.
Everything runs on the government-owned Autonomy Government Reference Architecture (A-GRA). If a software update is built by Shield AI, it can be instantly ported into a General Atomics FQ-42A airframe. If Anduril updates its software, it can deploy across the entire fleet without rewriting code for a specific engine or wing configuration.
The Brutal Math Behind the $30 Million Ceiling
Why the sudden rush? It comes down to cold, hard attrition math.
A single F-35 costs around $80 million to $100 million, depending on the variant. More importantly, it takes years to build, and losing a pilot is an operational disaster. In a high-end conflict against a near-peer adversary, the U.S. simply cannot replace human-piloted fighters fast enough to sustain a prolonged war.
The Air Force capped the cost of a CCA at roughly one-third of an F-35. That means these drones must clock in at less than $30 million per tail.
At that price point, the aircraft are "attritable." They aren't disposable suicide drones, but they are cheap enough that commanders can risk them in heavily defended airspace without sweating the financial or geopolitical fallout of a loss.
These robotic wingmen will fly alongside crewed fighters like the F-35A, the F-22A Raptor, and the upcoming sixth-generation fighter platforms. A single human pilot will act as a quarterback, directing a small pack of CCAs to fly ahead, jam enemy radars, scout out missile batteries, or even carry and launch air-to-air weapons.
Real Hardware Flying Real Sorties right now
It's easy to dismiss military drone announcements as vaporware or CGI concept videos. This is different. The hardware is already burning jet fuel.
General Atomics developed its FQ-42A under its "Gambit" modular concept. It went from a clean-sheet design to its first flight in just 15 months, taking to the sky in August 2025. It uses a core chassis that can be swapped with different wings and sensor suites depending on whether the mission calls for long-endurance surveillance or raw air-to-air strike capability.
Anduril’s FQ-44A Fury has been moving at an equally frantic pace. Born from its prototype award in April 2024, the aircraft completed its first flight in October 2025. The company has already logged dozens of sorties from multiple airfields. During recent tests, Anduril proved that its uncrewed fighter could switch between two entirely different mission autonomy software suites mid-flight.
More impressively, during an exercise with the Air Force's Experimental Operations Unit, a skeleton crew with only a few days of training successfully launched, recovered, and turned multiple FQ-44 sorties. They did it completely clean, without relying on the massive, specialized infrastructure of a traditional airbase.
The FQ-44 has the ferry range to self-deploy globally, can operate from short fields, and boasts a combat radius that significantly exceeds current crewed fighters. It's a purpose-built system designed to keep up with supersonic fighters while carrying lethal payloads.
The Unresolved Friction Points
Despite the rapid contract awards, this program isn't guaranteed smooth sailing. It's a massive experiment, and it comes with real operational friction.
First, there's the training and maintenance headache. Running two completely different airframes simultaneously means the Air Force has to train technicians on two separate hardware systems, maintain two distinct parts pipelines, and manage differing airframe wear-and-tear profiles. Some critics point out that this looks dangerously similar to past naval acquisition blunders where managing multiple specialized designs ended up ballooning costs instead of shrinking them.
Second, the system isn't crash-proof. Just two months ago, in April, General Atomics’ prototype suffered a high-profile crash at its California facility due to an autopilot software error. It paused flight testing and served as a blunt reminder that relying heavily on complex code introduces entirely new failure modes.
Finally, the Air Force hasn't actually revealed how many drones are included in these initial production lots, or what the exact price per unit is for the first run. They claim the numbers are classified to keep adversaries guessing, but it also shields the program from early budget scrutiny if the initial low-rate production costs spike above that $30 million target.
Your Next Strategic Moves
If you are a defense tech contractor, commercial software developer, or hardware manufacturer, you need to adapt to this shift immediately. The old days of bidding on a monolithic 30-year program are fading.
Stop trying to sell all-in-one proprietary systems. If your product doesn't use open architecture or comply with government-owned frameworks like A-GRA, the Pentagon won't buy it. Focus on building modular components that can be hot-swapped into existing platforms.
Look closely at the software vendor pool. Companies like Shield AI and Collins Aerospace are winning because their software is platform-agnostic. If you build AI pilots, sensor-fusion algorithms, or threat-detection software, target your development to plug directly into the ongoing six-month competitive cycles.
Expect this rapid-procurement model to bleed into other branches of the military. The Navy and Army are watching the CCA program closely. Start auditing your manufacturing capabilities now to see if you can support fast, low-cost serial production rather than low-volume, exquisite hardware.
