Initium Technologies Builds Edge AI Systems to Counter Cloud Vulnerabilities in Modern Warfare

Share this news:

Initium Technologies develops edge AI nodes that keep U.S. military software operational when cloud data centers are targeted or offline, shifting computation to ships and special operations teams so decisions can be made locally in contested, disconnected, or covert environments.

-- On March 1, Iranian drones struck Amazon Web Services facilities in the United Arab Emirates and Bahrain. Daniel Choupak & Conrad Scheibye, co-founders of Initium Technologies, had spent the preceding months building a company in anticipation of such an attack.

Photo courtesy of Initium Technologies

The Hidden Risks of Military Cloud Dependence

A growing share of U.S. military software now runs through commercial cloud data centers like those targeted. The arrangement is recent and deliberate. Over the past decade, the Department of Defense moved substantial data, analytics, and artificial intelligence (AI) workloads onto commercial cloud infrastructure, a shift formalized in 2022 when it awarded the Joint Warfighting Cloud Capability contract to Amazon, Microsoft, Google, and Oracle. The reasoning held up well in peacetime: Commercial providers had scale that the department could not match on its own, and buying access was faster and cheaper than replicating it.

This model, however, introduces new operational risks in conflict scenarios. Those buildings are fixed, publicly known, and commercially operated. Many of them sit in allied countries close to contested regions because that is where mission latency is lowest. By extension, however, this places them inside an adversary's reach.

A direct strike is the most visible form of failure, but also the least common. More often, connectivity is denied, degraded, intermittent, or intentionally avoided. In the case of covert operations, a unit may not transmit at all, since a signal can be direction-found and a request for compute becomes a disclosure of position.

Rethinking Architecture for Contested Environments

Choupak incorporated Initium Technologies in November 2025 with co-founder and CTO Conrad Scheibye and a team of six others, and is now headquartered in Mountain View, CA. The company develops edge-based AI systems designed to operate independently of centralized cloud infrastructure.

They argue that the U.S. has built a real and compounding advantage in military software and has leveraged that advantage by staging it on physical infrastructure designed for commercial workloads under peacetime conditions.

Those dependencies held for twenty years of counterinsurgency, against adversaries who could neither reach the infrastructure nor contest the spectrum. Against one who can do either, each is a single point of failure that no amount of software quality corrects.

Initium Technologies does not describe its programs as products. Choupak describes them as nodes on one architecture, with the same assumption that nothing past the node is guaranteed to be there. In practical terms, a node is a self-contained computing system that can operate independently wherever it is deployed. What changes from node to node is where it sits, what feeds it, and what it is asked to decide.

Deploying Edge AI Where Decisions Happen

MANTIS, or (Maritime Autonomous Neural Triage and Inference System), is edge processing for uncrewed aircraft, sited aboard U.S. Navy surface combatants and expeditionary platforms. A UAS operating over open water has no reliable connection to a distant data center, but it can connect to the ship it launched from. As a result, the node is deployed aboard the ship.

Putting it there is a choice rather than a compromise. A warship supplies power and cooling that a node in a desert or on a vehicle would otherwise have to bring with it, making the maritime environment the cheapest and least complicated place to field the architecture quickly and at scale. It is also where interoperability gets proven.

“Warships are moving, defended targets; If our carrier goes down, we have a much bigger issue than the data center that just went down with it,” Aaryan Khanna, Chief Operating Officer at Initium Technologies, mentions.

A node that is not married to any single airframe can serve whatever is flying that day, and aboard a ship, it has to speak to systems already installed rather than arriving as a closed appliance. A data center in Bahrain is fixed, publicly known, and stationary. A carrier strike group is moving and defending.

While MANTIS shows how Initium Technologies’ architecture works in the maritime domain, IRIS applies the same principle to close-quarters ground operations. IRIS is the node for special operations, an edge-AI situational awareness system for building-breach operations, aimed at the U.S. Special Operations Command, with person detection, georeferencing, FOV ray tracing, and facial recognition among its capabilities.

In these missions, disconnection is often intentional rather than imposed. A team conducting a covert entry holds an advantage that consists almost entirely of not being located, and reaching back would spend it.

The nodes look nothing alike operationally, which is what Scheibye is using them to demonstrate. None of them is the company. The count is set by how many places the mission outruns the network, and Choupak expects it to keep climbing anywhere the decision has to be made faster than the link can carry it.

This reflects a broader shift toward distributed computing in defense, where decision-making moves closer to the point of action rather than relying on distant infrastructure.

Contact Info:
Name: Aaryan Khanna
Email: Send Email
Organization: Initium Technologies
Website: http://initiumtechnologies.com

Release ID: 89199737

CONTACT ISSUER
Name: Aaryan Khanna
Email: Send Email
Organization: Initium Technologies
REVIEWED BY
Editor Profile Picture
This content is reviewed by our News Editor, Hui Wong.

If you need any help with this piece of content, please contact us through our contact form
SUBSCRIBE FOR MORE