Today, many question the relevance of the aircraft carrier in great power competition. Wargamers fear China’s “carrier killer” missiles, while others lament the diminished range of U.S. carrier-based strike aircraft. Members of Congress question the supercarrier’s value to the taxpayer and demand the Navy seek lower-cost solutions. Yet, advances in artificial intelligence and drone technology are opening the door to a solution for this strategic quagmire: make every ship a carrier.
A squadron (approximately ten aircraft) of Group 3 or 4 unmanned aerial systems (UAS)—those large enough to fly relatively high, far, and for a long time but are still small enough to fit on board a ship of nearly any size (Group 1 are the smallest, Group 5 are the largest)—operated by artificial intelligence (AI) algorithms should be placed on every ship in the Navy, Coast Guard, and Military Sealift Command (MSC) inventory. Deploying them across the surface fleet would enable new capabilities across multiple domains, optimizing combat effectiveness and logistics.
Combat Enablers
The Office of Naval Research issued the Science and Technology Strategy for Intelligent Autonomous Systems (IAS) in 2021describing the “envisioned futures” of swarming, distributed and persistent sensors, and sea control and denial. The Chief of Naval Operations’ Navigation Plan 2022 likewise anticipates a future Navy comprising thousands of autonomous platforms augmenting the force for lethality and survivability. AI-enabled UAS—or intelligent autonomous airborne systems (IAAS) for the purposes of this article to distinguish them from comparable surface and subsurface efforts as well as remote-piloted UAS—can act as force multipliers for each ship in the fleet. As Vice Admiral Jim Kilby clarified, “Unmanned systems in themselves aren’t a goal, they’re an enabler for a capability based on a threat that is rapidly accelerating.”
The Department of Defense (DoD) also has embraced the AI future, recognizing it as critical to affordable combat and deterrence power. In September 2023, Deputy Secretary of Defense Kathleen Hicks announced an ambitious plan to “field attritable [otherwise known as “expendable”], autonomous systems at a scale of multiple thousands in multiple domains within the next 18–24 months.” The so-called Replicator initiative is pushing the Pentagon to rapidly adopt AI technology to offset China’s advantage in industrial production capacity.
IAAS could provide critical capacity to the Navy’s surface fleet. For instance, IAAS can act as communication nodes, relaying information over the horizon (OTH) to manned and unmanned surface ships. A resilient, airborne network of line-of-sight communications would mitigate threats posed by passive detection, interception, or jamming of satellite and OTH communication pathways. IAAS can be equipped with passive or active sensors to detect threats and extend the range of a ship’s surveillance area.
IAAS can augment kinetic missions as well. Equipping a team of drones with munitions could allow for precision strikes, suppression of enemy air defenses, or close-air support of ground troops. As the conflict with the Houthis has shown, the Navy needs less expensive ways to defend commercial shipping than precious surface-to-air or air-to-air missiles. Sacrificial drones that could maneuver themselves between manned platforms and threat UAS or cruise missiles would be a crucial use-case for IAAS. Nonkinetic effects, such as airborne electronic attacks, could aid kinetic strikes from manned aircraft by protecting them from detection or targeting by enemy platforms. AI drones could be equipped with radio frequency signal repeaters and spoofers to saturate the radio frequency environment and complicate targeting solutions.
Combat Logistics
IAAS also could benefit the Navy’s strained strategic sealift force.The Government Accountability Office reported in 2019 that the MSC has a fleet of just more than 100 active and reserve ships, some of which are more than 50 years old. Further, the Navy’s push for a more distributed fleet will place increasing demands on MSC assets. In 2022, the Navy demonstrated IAAS blue-water logistics capabilities from tech startups Shield AI and Skyways, which can carry small cargo loads of less than 50 pounds up to 200 miles. Impressively, another drone by Elroy Air can fly 300 pounds of cargo up to 300 miles.
Strategic sealift forces could take an Amazon-like approach to the seas. Using MSC ships as centrally located sea hubs, drones could fly cargo delivery missions back and forth to ships requiring resupply. IAASs would not be sufficient to deliver fuel, munitions, or heavy equipment. However, light-lift cargo of less than 50 pounds accounts for 90 percent of Navy logistics deliveries, according to Naval Air Systems Command. Intelligent, autonomous surface ships may one day be able to fulfill this role for heavier cargo and fuel. In the meantime, IAAS could augment underway replenishments, open bandwidth for human helicopter pilots to focus on combat missions, and reduce the time-distance problem for MSC ships to supply the Navy during combat operations.
Public-Private Partnership
Companies such as Shield AI, Anduril, Skyways, Elroy Air, and others are developing innovative, low-cost drones and powerful AI algorithms to fly them. For example, Shield AI produces a Group 3 UAS called V-BAT, which can team up with other V-BATs to coordinate tasks. A single V-BAT is expected to cost around $500,000. For the price of a single F-35 stealth fighter, the Pentagon could purchase nearly 200 V-BAT drones. Foregoing a Ford-class carrier could easily turn every ship in the Navy’s inventory into a drone carrier for the hybrid fleet.
The real value the private sector provides is under the hood of their flying machines. Reinforcement learning, a subset of deep machine learning, enables a computer system to learn how to accomplish a task by simulating an action and then observing the results of that action millions of times. At first, the AI is incompetent and naïve. But run the simulation a million times, and the AI now performs that task far better than any human, potentially in a matter of days or weeks. A reinforcement learning system dubbed “Falco” beat an experienced Air Force fighter pilot in a simulated dogfight during DARPA’s AlphaDogFight challenge in 2020. The DoD lacks the expertise needed to develop AI like this itself, but it is taking strides through initiatives like Replicator, as well as innovation hubs like AFWERX and NavalX, to purchase the capabilities.
To Err Is Human?
Building individual and institutional trust in AI is a significant obstacle to the implementation of this vision. Paul Scharre, author of Four Battlegrounds: Power in the Age of Artificial Intelligence, highlighted the early failing of the automated Patriot missile system, which shot down friendly forces on two occasions during the 2003 Iraq invasion. “If warfighters don’t trust a technology, they won’t use it,” he wrote.1
Algorithms today are much more capable than their predecessors, but there must be strong institutional safeguards before fielding this technology in combat. To that end, Deputy Secretary Hicks established the “Responsible Artificial Intelligence Strategy” in June 2022 and DoD Directive 3000.09, titled “Autonomy in Weapons Systems,” in January 2023. These frameworks provide a foundation for the responsible military development of artificial intelligence but may prove challenging to implement in practice.
Directive 3000.09, for example, requires that AI be designed with “technologies and data sources that are transparent to, auditable by, and explainable by relevant personnel [emphasis added].” Neural networks, by design, are difficult to explain. A given input travels through many hidden layers like artificial neurons, producing an output response or action. Inspired by the workings of a human brain, machine learning is often shrouded in a black box.
Trusting AI, then, will require a different model. The traditional test and evaluation, verification, and validation process may not allow DoD complete certainty of how its autonomous platforms will perform.2 However, DoD already takes this as a given whenever it trains a new service member. Humans often make errors early in their training but less so as they gain more guidance and experience. Assuming a basic level of safety can be guaranteed to people and equipment, operators may need to tolerate AI’s mistakes as the cost of training the machines. But as training improves, AI has already demonstrated the capacity to perform discrete tasks much better and faster than humans.
Finally, the Navy will more quickly build trust in AI if it distinguishes between standards for lethal and nonlethal effects. Given the hesitance to embrace “killer robots” because of ethical and legal concerns, AI drones that operate lethal combat missions should be scrutinized more closely before participating in combat. However, many Group 3 IAAS are unlikely to carry large, destructive payloads. Their primary utility lies in nonlethal effects. Drones conducting these missions should be deployed more rapidly and given some leeway to improve over time by learning from their mistakes.
The trust factor is not the only challenge. IAAS will require manpower and training, though much less than a typical aviation squadron. An air vehicle officer, currently filled by the warrant officer ranks to operate MQ-25A Stingray, should be trained on various IAAS types and models. Alternatively, the Navy’s newly announced Robotics Warfare Specialists could be responsible for maintaining the drones, updating the software, and programming them for the proper mission. Once launched, these professionals would monitor the mission and send updates to tasking or commander’s guidance, serving as the human in the loop.
Free space on smaller ships is also at a premium. While Group 3 and 4 UAS can be small, hosting a dozen on a ship could still be challenging. However, IAAS can offer valuable capabilities and are worth finding space for. In September 2023, the Navy posted a request for information regarding “Aircraft Launch and Recovery Equipment (ALRE)” for Group 3-5 drones that would “operate off of noncarrier vessel[s].” The Navy is clearly exploring further integration of air and sea domains. Making every ship a carrier aligns with the strategic direction of the Navy’s future air force.
Finally, having every ship act as a carrier would involve a mindset shift. While surface ships have long had to consider the air domain in the missile age, launching and supporting drones is a new and distinct mission set. Surface warfare officers will need to become familiar with some airborne tactics and understand how the ship’s position will enable the desired tactical effects. Despite these challenges, the burden of adapting the surface fleet’s modus operandi to drone warfare is outweighed by the numerous benefits IAAS offers.
Future War
Despite the ambitious two-year timeline of Replicator and Secretary of the Navy Carlos Del Toro’s declaration that “the hybrid fleet is already here,” there remains a long road ahead for fully integrating intelligent autonomous systems into the manned fleet. And glaring shortfalls in long-range munitions may necessitate prioritizing current capabilities over untested and unproven dreams of robotic warfare in the short term. Yet, the speed of warfare is accelerating; the Navy’s challenge is to layer acquisition strategies for both short- and long-term requirements to meet the demands of future wars.
A new vision of naval air power is in order. The Navy must transition from a platform-based mindset—focusing on the ships, submarines, and aircraft—to one that is effects-based. AI-enabled airborne drones offer capabilities that could usher in a new era of naval aviation. Making every ship a carrier will revolutionize and unite the air and sea domains in ways unseen since the birth of the aircraft carrier.
1. Paul Scharre, Four Battlegrounds: Power in the Age of Artificial Intelligence (New York: W. W. Norton & Company, 2023), 249–53.
2. Scharre, Four Battlegrounds.