Air Force Drones Fly Combat Formation for First Time Using Hivemind

The U.S. Air Force's partly autonomous combat drones flew in formation and executed air combat maneuvers beyond visual range for the first time on Oct. 26, a milestone for the Collaborative Combat Aircraft program and for Shield AI, whose Hivemind autonomy software coordinated the aircraft.

The drones flew in what the company called a "multi-ship" formation and carried out counterair tactics — air-to-air operations intended to gain control of the sky — beyond visual range, according to a recent company release. The exercise did not identify how many aircraft participated, and Shield did not specify which CCA (Collaborative Combat Aircraft, the Air Force's term for drones that team with crewed fighters) took part.

The company framed the demonstration as evidence that three elements — the aircraft, their mission systems, and the Hivemind software that "runs" CCA flights — are reaching a new level of integration.

"We are excited that the aircraft and mission systems are reaching a level of integration to start showing what Hivemind can really do," Todd Wesley, vice president of Hivemind engineering at Shield, said in the announcement.

That claim rests on an undisclosed number of aircraft of an undisclosed type. What the release does establish is that the autonomy software coordinated multiple aircraft simultaneously through both formation flight and counterair tactics — the core capability the Air Force needs if it intends to build a fleet of drone wingmen around it.

Why the exercise matters for the Air Force's drone fleet

The stakes are large. The Air Force has stated its intention to field 500 combat-ready CCA systems by 2032, within an overall planned fleet of 1,000 aircraft. The Oct. 26 exercise is an early, concrete step toward that goal: coordinated multi-aircraft autonomy is a prerequisite for any of those systems to fight as a unit rather than as individual platforms.

The program also carries substantial budget weight. In the fiscal 2027 budget request, the Air Force asked for $996.5 million in CCA procurement and $150 million in advanced procurement for the following fiscal year. Research and development accounted for another $1.37 billion. Those lines total $2.37 billion for the program — a figure that signals how central partly autonomous aircraft have become to the service's force planning.

Congressional appetite for that spending will likely hinge on demonstrations like this one. Formation flight and beyond-visual-range tactics are the sorts of results that convert a software project into a funded weapons program.

What the software cycle tells us about testing speed

According to the statement, each flight produces data that feeds a software release. That process informs decisions on continued testing and the eventual fielding of the systems.

Shield said the process is approaching a faster rhythm. The Air Force is nearing the ability to complete a sortie, revise the software overnight, and return to flight the following day.

"The combined Hivemind and aircraft software release process is nearing the point where the Air Force can fly a sortie, update the software overnight and fly again the next day," Wesley said. "I'm grateful to the Air Force operators, engineers and test teams who fly with us and help us build trust on every flight."

A one-day test cycle would put CCA development on a cadence closer to commercial software iteration than traditional military aviation programs, which often measure gaps between flights in weeks. Wesley's word "nearing" is doing real work in that sentence; the company did not say the overnight cycle has been achieved, only that it is close.

His reference to building trust "on every flight" also points to the program's central challenge beyond engineering: convincing Air Force operators to accept autonomous aircraft in combat roles. Each flight's data serves that purpose as much as it serves the code.

How this connects to the F-22 and YFQ-44A demonstration

The formation exercise followed another CCA program milestone. Last month, an F-22 Raptor completed airborne tactical command and control of a YFQ-44A — an experimental autonomous fighter prototype built by Anduril — while both aircraft were in flight.

In that demonstration, the F-22 acted as a "quarterback," issuing broad directives to the CCA rather than controlling it movement by movement. The event showed a crewed aircraft exercising tactical command over a partly autonomous platform from the air, the command-and-control model the Air Force envisions for future fighter formations.

The two demonstrations cover complementary halves of the CCA concept. The F-22 event showed a human directing one autonomous aircraft; the Oct. 26 exercise showed autonomous software coordinating multiple aircraft with each other. A operational system requires both, and neither has yet been shown working together in a single exercise.

Shield AI's $5.3 billion radar acquisition

The milestone arrived as Shield AI expands its position in unmanned systems through a major corporate deal. The company announced last week that it will acquire Echodyne, a producer of next-generation solid-state radar systems, in a transaction valued at $5.3 billion.

The acquisition was described as supporting Shield AI's expansion into different areas of unmanned systems. Shield AI's chief executive said the purchase serves that broader aim.

The deal links autonomy software with radar sensing — the component an autonomous aircraft uses to perceive threats — under one roof. That vertical integration could matter as the CCA program moves from demonstrations to fielded systems, though neither the release nor the acquisition announcement specified how Echodyne's radar would feed into Hivemind-equipped aircraft.

What comes next

The formation flight, combat maneuvers, and accelerating software process together represent the latest reported progress for Hivemind and the CCA program. Each new flight provides more data for software releases, while Air Force operators, engineers, and test teams continue working alongside Shield.

The open questions are the ones the release leaves unanswered: how many aircraft flew, which CCA airframe was involved, and whether the overnight software cycle holds up at scale. Until those details emerge, the Oct. 26 exercise stands as a credible but unquantified step toward the 2032 target of 500 combat-ready systems.

The trajectory, however, is clear — if Shield's software process keeps compressing, the pace of demonstrated capability will likely accelerate faster than traditional defense program timelines suggest.