The new Defense Department budget request seeks big investments in artificial intelligence and autonomy: Large and small undersea drones, flying drone swarms, flying wingmen for both the Navy and the Air Force, and basic AI research. 

The big question now is if the Defense Department has evolved enough to spend that money in a way that reflects the rapid pace of autonomy development. 

The Air Force leads the services with $789 million requested for its ongoing program to deploy autonomous “wingman” fighter drones alongside jets, a program called collaborative combat aircraft, or CCA. The Air Force plans to spend about $28 billion on the program by 2029. 

But new to the Air Force budget this year is a smaller drone swarm program called Offensive Small Unmanned Aircraft Systems, intended to develop a family of drones that can operate in places where adversaries are deploying heavy electromagnetic warfare effects. They’re intended primarily for intelligence collection and reconnaissance, as well as some precision strikes. The effort bears many similarities to the Defense Department’s larger “Replicator” program, which also seeks to build out a large force of small, cheap, highly-autonomous drones, and counter-drone offenses. 

The Air Force is also requesting $147 million for “tactical autonomy” materials, as in sensors, software, and compute modules to run AI programs on existing or future platforms. But existing platforms don’t factor much into the Air Force ask. Among the most significant is $29 million to modernize the MQ-9 Reaper, which Air Force Special Operations Command is also trying to convert for use in a swarming team.

The Navy also has big autonomy plans and is requesting $185 million for its large  unmanned undersea program, as well as funds for medium and small undersea vehicles. They’re also looking for almost $300 million for “advanced undersea prototyping,” which will likely lean heavily on artificial and autonomy, as well as $361 million to update the MQ-4C Triton, a high-altitude intelligence and surveillance drone intended to keep persistent watch over the sea. 

The Navy’s request doesn’t include a specific line item for its largest autonomy effort, the MQ-25 Stingray carrier-based refueling drone, in part because the program’s components are spread across multiple other programs. But Navy officials have recently reiterated that they are committed to developing the drone and anticipate funding levels for this year similar to last year.

The Army is looking to pour $486 million into basic AI research and development, $35 million for robotics development, and $144 million for a broad category of “emerging technology.”

On top of those service investments, the Defense Department is requesting $582 million for data labeling services, model deployment frameworks, and more, which would be available to all military branches and agencies under an effort called “Alpha-1.” That’s in addition to a DARPA budget request of $1.7 billion and $45 million for CDAO.

While the Defense Department is showing a willingness to spend more on research and development to field AI and autonomous solutions, GAO and others have warned that unless the Pentagon changes the way it builds and buys things, it will continue to see over-budget, behind-schedule programs. Speaking at the Defense One Tech Summit last week, Chris Brose, the president and chief strategy officer for Anduril, said he’s hopeful that the Defense Department is moving toward a new philosophy of investing in technology, de-emphasizing programs of record and formal requirements and focusing more on buying new tech in regular increments. 

“I think you can begin to create programs that look more like markets where the government is actually going out to industry more regularly, every two to three years, for example,” he said.

Many of the autonomy solutions the Pentagon is going to buy will essentially be smart ammunition, he said, such as one-way attack drones. But “many of them are going to be vehicles or other things that we're going to operate for several years. We can go get new versions, the way you use or lease commercial cars and trucks. And by creating that type of demand signal, and that type of signal to industry… you can begin to create incentives that make it a little bit more market.”

Brandon Tseng, co-founder of drone maker Shield AI, acknowledged at the Tech Summit that big programs or records for manned submarines and manned fighter jets or bombers will always exist. But “there is a subset of products, AI, autonomy, drones, which are rapidly evolving, and there's wide recognition that the drone you buy now is going to look very different two years, three years from now. So we shouldn't be locked into a 17-, 20-year program of record to buy these technologies.”

Critical to the success of any of these autonomy efforts, Tseng said, is to get them to warfighters faster, including, potentially, partners like Ukraine. 

“You really do need to move more towards a continuous development approach. Are you continually getting these things in the hands of war fighters, having that feedback that then feeds back to the company, where then 24 hours later, they can give you an improved product.”