A recent Morgan Stanley Aerospace & Defense research update highlights what may be one of the most significant shifts in U.S. defense spending in decades. The FY27 budget request, released in early April, calls for approximately $1.5 trillion in total national defense spending—the largest in history. More notably, roughly $758 billion, nearly half of the total, is directed toward modernization, including procurement and research and development.
More telling than the headline number, however, is how that funding is allocated. Nearly half, roughly $758 billion, is directed toward modernization, spanning both procurement and research and development.
This is not simply an expansion of existing programs. It reflects a deeper transformation in defense systems themselves. The emphasis is shifting away from incremental improvements to legacy platforms and toward a fundamental re-architecture of how systems are designed, integrated, and operated.
Missiles and munitions are scaling rapidly. Space and missile defense initiatives such as Golden Dome are accelerating. Multi-domain operations are no longer aspirational concepts but operational imperatives. Across each of these areas, a common thread is emerging: modern defense systems are becoming increasingly software-defined—and, just as importantly, increasingly dependent on artificial intelligence operating at the edge.
Yet while much of the public conversation remains focused on platforms and hardware, the real challenge lies elsewhere. At this scale, modernization is no longer primarily constrained by what can be built. It is constrained by what can be integrated, secured, and trusted, especially when intelligence itself is embedded directly into mission systems.
The Shift Beneath the Surface
For decades, defense acquisition cycles were dominated by physical systems. Aircraft, ships, and ground vehicles defined capability, and software played a supporting role. That model is now breaking down.
Today’s systems are better understood as distributed networks of sensors, processors, and effectors, all connected through software and increasingly mediated by AI-driven decision-making. A missile defense system, for example, is no longer just an interceptor. It is a coordinated architecture that integrates space-based sensing, terrestrial radar, real-time data fusion, and decision engines capable of operating under extreme time constraints.
The scale of investment now flowing into modernization suggests that the Department of Defense recognizes this shift. However, funding alone does not resolve the underlying complexity. As systems become more interconnected and more software-intensive, the difficulty of integrating them grows exponentially. When artificial intelligence is introduced into the equation, that complexity increases again.
The bottleneck is no longer production. Nor is it simply technological innovation. It is the ability to deploy and operate software, and now AI, within environments that demand absolute reliability, security, and predictability.
Golden Dome and the Evolution of System-of-Systems Warfare
Missile defense provides a particularly vivid illustration of this transformation. Programs like Golden Dome are often described in terms of capability, interceptors, sensors, and coverage. But the defining characteristic of such systems is not any individual component. It is the architecture that connects them.
Golden Dome is, by design, a system of systems. It relies on the seamless coordination of assets across multiple domains, each operating under different constraints and timelines. Increasingly, artificial intelligence is expected to play a central role in this coordination, helping to process vast streams of sensor data, identify and prioritize threats, and support decision-making at speeds that exceed human capacity.
This introduces a fundamental challenge. AI is not a deterministic technology. Its outputs are probabilistic, shaped by data and models that evolve over time. Integrating such systems into mission-critical environments, where outcomes must be predictable and certifiable—requires a level of architectural discipline that traditional approaches cannot provide.
The problem is no longer just connecting systems. It is ensuring that these systems, including their AI components, can operate together securely, predictably, and with assurance.
The promise of AI in defense is often framed in terms of capability: faster decisions, greater autonomy, and enhanced situational awareness. These benefits are real, but they come with significant constraints—especially when AI is deployed at the edge.
Unlike centralized data center environments, edge systems operate under tight resource constraints and often in contested or disconnected conditions. They must process data in real time, support mission-critical functions, and remain secure against adversarial threats. Introducing AI into this environment adds layers of complexity.
AI workloads are computationally intensive and inherently non-deterministic. They require continuous updates and retraining, and they introduce new attack vectors, from data poisoning to model manipulation. At the same time, the systems in which they are deployed must meet stringent requirements for safety, certification, and reliability.
This creates a tension that cannot be ignored. Mission-critical systems demand predictability and assurance. AI introduces variability and uncertainty. Bridging this gap is one of the central challenges of modern defense architecture.
The Limits of Current Approaches
Efforts to modernize defense systems have increasingly embraced modularity and openness, particularly through the adoption of Modular Open Systems Architecture (MOSA). The intent is to enable interoperability, reduce vendor lock-in, and accelerate innovation.
While MOSA is a critical step forward, it does not fully address the challenges introduced by AI. Open systems, by their nature, increase the number of components and interfaces that must be managed and secured. When AI is layered on top, the attack surface expands further, and the complexity of ensuring correct behavior increases significantly.
Traditional approaches to security, often implemented as overlays or afterthoughts—are insufficient in this context. Similarly, approaches that treat AI as just another application fail to account for its unique characteristics.
What is needed is not simply more modularity, but a deeper integration of security, control, and assurance into the architecture itself. This is the premise behind the evolution from MOSA to what can be described as MOSA.ic.AI.
From MOSA to MOSA.ic.AI
MOSA.ic.AI represents an extension of modular architecture principles into the AI domain. It acknowledges that AI workloads must be treated differently isolated, governed, and integrated in ways that preserve system integrity.
In this model, AI is not allowed to operate freely within the system. Instead, it is contained within defined boundaries, subject to policy controls, and continuously monitored. Its interactions with other components are mediated through secure interfaces, ensuring that failures or anomalies cannot propagate unchecked.
This approach enables AI to be deployed in environments where assurance is paramount. It allows systems to benefit from AI-driven capabilities without compromising the reliability and security on which mission success depends.
The Role of the Software Substrate
Enabling MOSA.ic.AI in practice requires a foundational layer that is often overlooked. Beneath applications, middleware, and even operating systems lies a more fundamental construct: the secure software substrate.
This substrate provides the mechanisms that make everything else possible. It enforces separation between workloads, ensuring that critical functions are insulated from less trusted components. It enables deterministic execution, allowing systems to meet real-time performance requirements. It embeds security at the lowest levels, supporting a Zero Trust posture that is essential in contested environments.
Perhaps most importantly, it provides a path to certifiability. By isolating components and controlling their interactions, the substrate allows systems to evolve incrementally, without requiring complete recertification each time a change is made. This is essential in a world where both software and AI must be updated continuously.
Without such a substrate, the ambitions of modernization are difficult to realize. Systems become brittle, integration becomes costly, and the risks associated with deploying AI grow unmanageable. With it, a different picture emerges—one of systems that are adaptable, resilient, and capable of evolving over time.
The significance of this architectural layer becomes even clearer when viewed in the context of current investment priorities. The rapid expansion of missile programs, the growth of space-based systems, and the emphasis on multi-domain operations all point toward a future in which integration and interoperability are paramount.
In each of these domains, AI is expected to play a central role. Yet in each case, the ability to deploy AI safely depends on the same underlying capabilities: isolation, determinism, security, and assurance. These are not features that can be added at the application level. They must be built into the foundation of the system.
This is why the concept of a secure software substrate is not merely technical. It is strategic. It determines whether the investments being made today will translate into operational capability tomorrow.
From Potential to Reality
Artificial intelligence has long been viewed as a transformative force in defense. But transformation does not occur at the level of algorithms alone. It occurs when those algorithms can be deployed, trusted, and sustained within real-world systems.
That transition, from potential to deployment, is where the greatest challenges lie. It is also where the greatest opportunities exist.
As defense systems continue to evolve, the focus will increasingly shift from what AI can do to how it can be integrated. The success of this integration will depend not on any single technology, but on the architecture that brings them together.
In that sense, the story of modernization is not just about new capabilities. It is about building systems that can accommodate change, manage complexity, and maintain trust under the most demanding conditions.
The $758 billion being directed toward modernization is a powerful signal of intent. Whether that intent is realized will depend on the foundations that are put in place today. Because in the end, the future of defense will not be defined solely by the systems that are built—but by the invisible layer that allows those systems, and the intelligence within them, to work together reliably, securely, and at scale.
How Lynx Enables Modernization at Scale
Modernization at this scale is not just about deploying new systems—it is about ensuring those systems can integrate, evolve, and operate with assurance in increasingly complex environments.
Lynx enables this by providing the secure software substrate required to safely deploy mixed-criticality workloads, including AI, at the edge. Through built-in isolation, deterministic performance, and Zero Trust enforcement, Lynx makes it possible to realize architectures like MOSA.ic.AI in practice, not just in concept.
In doing so, it helps turn modernization funding into operational capability.
Get in touch with our team Contact Us