Pairing a top-tier fighter with a “loyal” drone wingman sounds clean and modern. But the ugly truth is this: the more drones you add to the sky, the more you create a detection and identification mess that can get people killed fast. From my seat—building drone detection radar systems and AI fusion across sensors—I don’t see a future where this works safely without treating radar drone detection and sensor fusion as the main event, not a supporting detail.
Based on what’s been shared publicly, the U.S. Air Force has received the first YFQ-44A Fury and YFQ-42A Vengeance drones from Anduril and General Atomics. The reporting says these drones are meant to operate as a strike pair with the F-35A and a future F-47. There’s also reporting that the Air Force plans to have 150 similar drones by 2030, and that additional funding was requested for 2027: $1 billion for procurement plus more for continued development within the concept (the second figure was shown, but the way it’s written is unclear).
If you’re reading that as “cool, more capability,” you’re not wrong. If you’re reading it as “we’re scaling complexity into a battlefield that’s already hard to read,” you’re also not wrong. The second reading worries me more.
Because the hard part isn’t building a drone that can fly near an F-35. The hard part is making sure everyone in the area—friendly pilots, ground air defense, ships, allied systems—can correctly answer three questions in real time: What is that? Whose is it? What is it about to do?
That’s not a nice-to-have. That’s the difference between “teamwork” and “friendly fire,” between “deterrence” and “accidental escalation.”
Imagine a real scenario: an F-35 is pushing into contested airspace with two autonomous wingmen. A ground battery sees fast-moving tracks. A ship’s radar sees something else. An allied fighter sees a third picture. Somebody’s system tags one of the drones as “unknown” for 20 seconds because the track is messy, the angle is bad, or jamming is thick. In those 20 seconds, do you hold fire and risk being wrong, or shoot and risk hitting your own? People love to talk about autonomy like it reduces human error. In practice, it often just moves the error to a different moment—when the pressure is higher and the time is shorter.
This is why radar drone detection isn’t a single feature. It’s a survival requirement. And radar alone isn’t enough either, because the real world doesn’t politely separate targets by category. You need AI fusion from different sensors—radar, passive detection, optical, whatever else is available—so the system can keep a stable picture when any one sensor is degraded. If you can’t fuse and cross-check, you’re basically asking operators to gamble under stress.
What’s really going on here, in my view, is that the Air Force is trying to buy time. Drones are a way to spread risk, extend reach, and complicate an opponent’s planning without putting a pilot in every cockpit. That’s a smart instinct. But there’s a temptation to treat “150 by 2030” like a scorecard. Fielding numbers is easy to celebrate. Fielding clarity is not.
And clarity is where this concept can break.
More drones in formation with high-end fighters doesn’t just change tactics; it changes the airspace as a system. Every extra aircraft is another emitter, another reflection, another track, another chance for misclassification. If the opponent adds decoys, spoofs identities, or floods the area with cheap drones, the sky becomes a crowded room where everyone is shouting. In that room, the side with better fused sensing and faster, more reliable identification has the advantage. The side with the most airframes might still lose control of the situation.
There’s also a quiet incentive problem. When procurement ramps, people start optimizing for what gets counted: deliveries, flight hours, demo videos. Meanwhile, the unglamorous parts—interoperability, deconfliction, identification reliability, the “what happens when two allied systems disagree” drills—get pushed to later. Later is where programs go to die, or to create operational debt that shows up in a crisis.
I’m not saying these drones are a bad idea. I’m saying the balance of attention is often wrong. If we’re going to pair drones with the F-35A and a future F-47, then the sensing and fusion layer needs to be treated like part of the aircraft, not a separate box someone bolts on after the fact. Otherwise you end up with incredible platforms that can’t be trusted in a crowded fight.
The strongest argument on the other side is simple: you can’t wait until everything is perfect. You need to field, learn, iterate. That’s fair. But “iterate” only works if the feedback loops are real—if misidentifications, near-misses, and tracking failures are captured honestly and fixed fast, not buried because they’re inconvenient or classified away from the people who can solve them.
So yes, getting the first Fury and Vengeance drones into Air Force hands is a milestone. But the milestone that matters more is whether the surrounding detection, identification, and fused awareness can keep up as the fleet scales from a handful of test aircraft to something like 150 operational drones sharing airspace with manned fighters, allies, and air defenses.
If you were responsible for the safety of the pilot, the drone, and the air defense teams on the ground, what would you prioritize first: getting more drones delivered quickly, or proving that radar drone detection and multi-sensor fusion can prevent misidentification when the sky gets noisy?