The AI boom has been built around a simple assumption: more computing, more data and better connectivity produce better AI. But that equation breaks down on a battlefield. EdgeRunner AI CEO Tyler Saltsman says military AI has to work when the cloud isn’t available, making efficiency and local computing less of a compromise and more of a requirement.
The Battlefield Doesn’t Have a Data Center
Consumer AI is increasingly tied to enormous computing infrastructure. A warfighter may have a laptop, tactical server or another edge device—and no reliable connection to the cloud.
That changes what "better AI" means.
"On the battlefield, you can’t assume you’re going to have a data center or even a reliable connection to the cloud," Saltsman told Benzinga in an exclusive email interview.
EdgeRunner AI is building its platform around that constraint, with AI designed to operate locally in disconnected and air-gapped environments. Its Battle Buddy, for example, is designed to deliver Llama 3-70B-level performance while requiring substantially fewer computing resources, according to the company.
The company’s current platform is explicitly designed for on-device use, including environments without internet connectivity. Its models can also run on relatively constrained hardware rather than depending on centralized cloud infrastructure.
That creates an unusual divergence from mainstream AI development.
Bigger Models Aren’t Always Better
The commercial AI race has made model size an easy shorthand for progress. Saltsman argues that military applications require a different yardstick.
"For military AI, the best model is not necessarily the biggest model," he said. "It’s the model that can perform the mission where the warfighter needs it."
That means an AI system capable of running locally can have an advantage that doesn’t show up in conventional model benchmarks: it can continue working when connectivity disappears.
EdgeRunner’s Battle Buddy has been designed to operate fully offline, including in air-gapped environments, while keeping sensitive information on the device. The company says its smaller models are intended to make sophisticated AI usable on laptops and other edge hardware.
The implication extends beyond defense. As AI moves into factories, vehicles and other locations where sending every piece of information to the cloud is impractical, the ability to run models locally could become increasingly valuable.
Cloud Dependence Can Become a Vulnerability
Saltsman makes the strategic argument directly: if a military AI system requires constant access to enormous cloud infrastructure, that dependence can itself become a weakness.
"The advantage comes from putting capable AI directly at the edge," he said, alongside the people and systems conducting the mission.
For investors, that creates a less obvious corner of the AI opportunity. The next phase of AI infrastructure may not only be about building larger data centers and buying more compute. It could also be about shrinking intelligence enough to put it where connectivity cannot reliably go.
The key question is whether that edge-first architecture remains a specialized defense requirement or becomes a broader computing trend as AI spreads into increasingly disconnected environments.
Image via Shutterstock
© 2026 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.



