Wrestling with massive amounts of data will define modern warfare as much as battling the enemy.
Inside the defense industry, companies are locked in a highly competitive race to build cutting-edge tools designed to bring secure, artificial intelligence-powered decision aids and communications to combat. The goal is to help America’s warfighters access, understand and capitalize on the mountains of data needed to navigate a complex 21st-century battlespace.
Deployable data centers are quickly becoming a critical piece of the equation. Those deployable centers — an alternative to the large, permanent, easy-to-target physical structures that Iranian forces have attacked in the Middle East in recent months — can transport data-processing power to the front lines of war.
Analysts say there is a clear benefit for troops.
“When you’re making split-second decisions, the difference between you having to do some work versus having what they’re calling an [AI] agent do it, sounds like it could be really helpful,” Jake Steckler, a researcher at GovAI, told The Washington Times.
A former U.S. Army pilot, Mr. Steckler now focuses on emerging technologies in defense at GovAI, a nonprofit think tank researching AI technologies.
“The question is always the robustness of the system. How reliable is it? Is it going to consistently provide you with the correct information without errors?” he said.
The U.S.-Iran conflict and the Russia-Ukraine war have acted as testing grounds for the most advanced weapons the U.S. has ever produced, including military AI applications. Many of these new tools leverage massive amounts of data, much of which is collected at the front lines of fighting.
In recent demonstrations at the Amazon Web Services Summit in Washington, D.C., both the distributed data approach and an AI-assisted future for U.S. warfighters were on display.
Using tools developed by companies already working closely with the military, a U.S. service member could coordinate and redirect fighter jets, rockets, and autonomous drones all in a single chat window. Military operations that would have taken hours to plan are reduced to minutes as AI agents make optimized routes and targets based on the vast ocean of data available.
Multiple defense firms have started to field options to push that computing power forward. The AI services that these data-centers-in-a-box could support have direct uses for every military branch.
The defense firm Anduril Industries has worked for the past three years on a program called Menace-I. It’s meant to connect key processing capabilities for autonomous drones, air defense radars, and even fighter jets, to human users. To do that, systems like Menace have to push close to the front lines of fighting.
“We actually see this as a modern weapon system,” Tom Keane, Anduril’s senior vice president of engineering, told The Times in an exclusive interview. “The bytes are the bullets.”
Fighting data
Anduril has combined forces with another rapidly rising defense industry power, Palantir, to connect information from each branch of the U.S. military and to allies such as Britain or Australia. Since December 2024, the two companies have worked together toward what they called “data readiness” and “processing data at scale,” according to company press releases.
Much of that connection has been based on a Palantir program called the Maven Smart System, which has quickly expanded across the Defense Department.
Multiple branches of the military now use Maven as their mandatory reporting system. It acts as a focal point for accessing the data feed from around the world. The Maven system was also contracted as part of NATO operations in March 2025, allowing allies to operate on a compatible platform.
“We’re seeing what’s going on in Iran with integrating Claude [AI] and other large language models into Maven Smart System,” Mr. Steckler said. “But that’s all happening in a very central location rather than being some sort of distributed fight in a denied environment.”
That reality has led to the push for deployable data centers.
Menace has, according to Anduril, already been used on Navy ships for mission planning and by the U.S. Air Force for AI-developed flight plans and mission briefs for the F-35 fighter jet.
And Anduril’s Lattice software is baked into Menace, just like “everything we deploy,” according to Mr. Keane.
Anduril isn’t the only company investing in the idea of more compute power at the edge, which refers to the transition point between computations, data and programs into real-world applications.
San Francisco-based competitor Armada is another mobile data center company that’s already working with the U.S. Navy. Both companies have billed their products as open-ended tools.
During the Amazon Web Services Summit, Anduril announced a new partnership with AWS — one of the key businesses involved in a multibillion-dollar cloud contract marketplace. The move lets troops and unit commands select exactly how much compute power they need and quickly field a Menace suite for that specific situation. The partnership also integrates Amazon’s military cloud technology called Outpost, a system that AWS says is already deployed around the world to nearly every combatant command.
Right now, all of those technologies are only at the military’s higher levels. Operations would change if the hardware on the front lines supported all of those new tools.
“To actually be integrated into the fight could possibly allow individual units to have this situational awareness of the battlefield that we’ve never been able to have before in real time,” Mr. Steckler said. “Especially for the future of warfare, where we’re talking about having more uncrewed systems, having the opportunity to do this will make the orchestration of those systems possible.”
The efficiency game
While new tools like Menace-I and Maven show great promise, the amount of energy and computing power required are problematic. Pushing this technology further forward, companies are trying to expand their capabilities at the edge.
Not everyone is convinced that more compute power and data further forward in combat is the answer. Eric Davis, a program manager for the Defense Advanced Research Projects Agency, or DARPA, told The Times that the agency’s research is focused on efficiency.
“If I want AI on the edge, the natural assumption is I need robust hyperscalar-level compute on the edge,” he said, pointing out that some offices in DARPA are working on that problem as well. “All of that is working at making our edge capabilities more robust from the computer perspective. There’s a second side to that, too, which is a question of do I actually need it, or are my methods perhaps inefficient?”
Mr. Keane didn’t disagree.
“The world has to make these models more efficient. We’re on a non-sustainable pathway right now,” he said. “The reason I believe so strongly in edge is physics … you have to have proximity to that data.”
Battle hardened
Attacks on infrastructure, jammed communications and other threats have made edge computing power a necessity for AI. Without it, specialists said, U.S. forces would rely entirely on information in the cloud.
“Relying on the cloud creates a lot of vulnerabilities,” said Mr. Steckler, the GovAI researcher. “The military can’t expect to operate continuously on a battlefield that’s a denied environment, a contested environment, if you’re relying on the cloud.”
Communications back to a data center can be “slower than you’d like,” Mr. Steckler said, as cloud computing can still create a bottleneck effect. That makes it vital to have computing power capabilities closer to the edge.
Ukraine has become one of the premier testing grounds for AI-at-the-edge for defense firms.
Shield AI, an aerospace and defense technology company, said it is moving to build more terminal systems, letting AI drive missiles, aircraft and other attack methods for the last few hundred yards.
“A 100% kill rate is worth a lot,” Brandon Tseng, the company’s president and co-founder, told The Times. Shield AI’s HiveMind program is already on board drones being used in Ukraine, has seen some testing in the Iran war and is part of the fully autonomous flights in the U.S. Air Force’s Collaborative Combat Aircraft (CCA) program. That program aims to provide a fully autonomous wingman fighter aircraft.
Eventually, the company believes, a fully autonomous aircraft could release AI-enabled weapons, allowing the U.S. military to conduct strikes deep behind enemy lines with no risk to a human pilot.
And it’s not just the U.S. HiveMind is being used by Taiwan, Greece, Japan and other allies and partners.
“We’ve got our CCA efforts with HiveMind. We’re doing an unmanned Lakota helicopter with Airbus for the Marine Corps with HiveMind,” Mr. Tseng said.
Eventually, new technologies like these will have to meet the hard realities of use in the field by service members.
“Now you’re trusting AI agents in life or death scenarios that aren’t even necessarily completely trustworthy in less than life or death scenarios,” Mr. Steckler cautioned. “Whether or not the technology bears out to be able to do that is another question.”
Contact the author
Please read our comment policy before commenting.