In a battlespace that is increasingly decentralized, fragmented, and unpredictable, drones are not only growing in number, but rewriting the rules of engagement. Russia’s invasion of Ukraine is an unfortunate pinnacle example — Ukraine expects the manufacturing capacity of first-person-view (FPV) drones to triple this year to 4.5 million, and sell out of stock. These $300 (or cheaper) drones are accounting for 80% of casualties in Russia and Ukraine, and are destroying billions of dollars of weapons stockpile as Ukraine and Israel have most recently shown through discrete attacks.
In a protracted and resource-constrained conflict, efficiency is the name of the game, as Squadra Principal Dan Madden illustrated this in depth after visiting Ukraine in the fall of 2023. At that point, autonomous solutions had not yet been deployed at scale as we are seeing today, and the need for more enhanced levels of autonomy was identified and still have yet to be realized. Drones level the playing field between militaries of varying sizes, requiring more rapid detection and decision-making than ever before. They have launched a technological revolution comparable to that of the machine gun, tank, and airplane in considering their impact on the character of warfare
Inexpensive drones have become a fulcrum of sorts in modern warfare, upending traditional cost-benefit logic — not only do they destroy multi-million dollar assets, they also require similarly expensive systems to be neutralized. Cheaper systems are causing outsized damage while more expensive ones might even be considered liabilities. To some, this may not sound an alarm as it may unsurprisingly look like a reimagined case of asymmetric warfare. For centuries, traditionally weaker forces have always found ways to offset conventional military advantages. Still, this inversion begs a reconsideration of resource allocation, to include procurement, force structure, and tactics, techniques, and procedures (TTPs).
Semantics aside, it is not only about shooting drones out of the sky. What makes this possible is locating them before they become a threat and deciding how to respond at machine speed. The introduction of AI to these systems has added a new layer of sophistication, as it enables a drone to conduct operations without connecting back to its pilot (unlike FPV drones), and thus does not need to transmit a signal. The question then becomes how to counter a drone, when there is no signal to grab onto.
Detection itself is layered; it requires a variety of sensor types, each with its own advantages and drawbacks. Radar, for example, provides long-range coverage and tracks fast-moving targets through inclement weather, but reflections and signal clutter in urban environments undermine it, at present. Radio-frequency (RF) sensors are less expensive and are excellent at passively picking up signals but, as identified, signals may not always be transmitting. Electro-optical (EO) and infrared (IR) sensors are ideal for silent or low-emission targets, but require clear lines of sight to provide sharp imagery. Acoustic sensors can fill some of these gaps. They too rely on passive detection, and they are small, lightweight, and power-efficient, nimble enough to be mounted on a mobile platform, though typically have shorter range. There is no silver bullet when it comes to individual sensing modalities, particularly as frequency hopping becomes an even more delicate dance. Layering modalities, though, creates a comprehensive toolkit. Below is a comparative chart of these different sensors in the context of countering unmanned aircraft systems (cUAS), including unmanned aerial vehicles (UAV) and ground control stations (GCS).
Still, detection is but the first step. Data processing and integration follow, and warfare today has brought it to the edge. Reliance on far-away command centers to process is a thing of the past, as contested and degraded communications environments are the norm. This is where next-gen command and control (C2) systems must adapt accordingly, and where the introduction of artificial intelligence/machine learning (AI/ML) training models is indispensable. They must possess an ability to synthesize inputs from a variety of sensor types at the edge to enable a split-second decision without access to a direct line back to a human decision-maker, which is why legacy systems are no longer able to keep up. The Department of Defense (DoD) investment in Joint All-Domain Command and Control (JADC2) reflects this shift, and signals to the cUAS market that future systems must be agile, modular, and open. Mutable Tactics is addressing a different facet of decision-making integration: autonomous tactical C2, perhaps providing a window into a future where squads of robots are fighting our wars. In any case, the ability to plug into larger architectures without necessitating a complete overhaul will be key, as speed is a critical capability in and of itself. A rapidly changing battlespace has presented the moment for less expensive, flexible, and software-defined solutions.
My few years spent in the Joint Staff J-7 Office of Irregular Warfare and Competition lead me to remind readers of one of Clausewitz’s core tenets: the nature of war is constant (violence, chance, and politics), but the character of war is variable, contingent on time, technology, geography, and actors. Today’s character of war is at an inflection point. If you have spent any time in the Pentagon, you know well, perhaps too well, that lessons learned are baked into just about everything. The Joint Force is now at an interesting inflection point where many of the takeaways from the Global War on Terror (GWOT) are no longer relevant, and certainly not sufficient as it prepares for the future of war. In fact, most would argue GWOT-era thinking is dangerously outdated, from a strategic lens. Using counterinsurgency lessons learned to win a drone-driven fight would be like learning tennis to play in a soccer game. All this to say, the character of war is changing in plain sight, seen in the Armenia-Azerbaijan war of 2020, and now in Ukraine and Gaza, amongst other places. The recent Ukrainian and Israeli drone strikes on Russian and Iranian critical infrastructure — conducted up to thousands of miles from the front lines — underscores that geographic sanctuary is disappearing. Drones now enable state and non-state actors alike to deliver precise, asymmetric attacks on critical infrastructure or military targets with near-zero warning, regardless of distance or border.
The Department can no longer turn to the past to prepare for the future. Of course, I am reminded of this timeless Foreign Policy article whose title, "100% Right 0% of the Time" speaks for itself. Even with all the time and capital spent on preparing for the future of war, even the most experienced have been astonishingly wrong in predicting what it will look like. The point I extract is that preparation can only take us so far and, at the end of the day, ultimate preparedness demands adaptability and flexibility at the speed of relevance. Given the pace at which defense tech needs and capabilities are evolving, there is no choice. This CNAS piece similarly highlights what about Ukraine may or may not be transferable to a potential conflict in the Indo-Pacific, i.e., underscoring that preparation and adaptability are paramount. In accepting a high level of uncertainty, we can still anchor on what the world is showing us to be true: sheer size and numbers no longer equates to victory. A new operational reality demands the ability to make sense of incomplete, ambiguous, and deceptive information; and make decisions more efficiently than one’s adversary. And, as conflict today is showing us, this capability does not exist within the confines of linear decision cycles and hierarchical control to which legacy systems adhere.
Today’s battles extend beyond land, sea, and sky- into code, signals, and information itself. They’re being fought in the electromagnetic spectrum (EMS), and the Pentagon knows it is playing catch up. In a recent conversation with D-Fend Solutions’ Brett Feddersen, he commented on how the United States’ ability to ensure speed of deployment in the EMS depends much more heavily on policy infrastructure and regulatory hurdles than technology, and those vendors with pathways to legitimate federal accreditation will have a head start. The United States has managed to create regulation around the initial “detect” component of the cUAS kill chain, but is still working on codifying the requirements and training standards for the “defeat” component — or in his words, “mitigation” — which in part speaks to why foreign sales are moving faster than domestic. This is also due to geopolitical urgency — NATO counterparts show greater risk tolerance out of pure necessity, creating faster pathways for operational validation.
It should come as no surprise that institutional skepticism and doctrinal lag present significant friction. I had the chance to pick Nathan Mintz’s brain, co-founder of Spartan Radar, Epirus, and CX2 (the latter very recently raising a Point 72-led $31M series A); he mentioned that current tracking and C2 systems are optimized for traditional aircraft. They are fundamentally ill-equipped to handle drone swarms operating at extreme density and low altitude, the inverse of traditional aircraft. He cited the Ukrainian offensive at Kherson, where he observed up to 800 drones per square mile at peak density. No U.S. system today can task and track at that scale. Such systems are vertically segmented- sequentially layered by function and range to serve different stages of the kill chain (more on this below). The People’s Liberation Army has already updated its doctrine to inform investment in AI-driven and multi-layered approaches blending kinetic and non-kinetic attacks. If the U.S. doesn’t respond in kind, traditional warfare systems risk becoming obsolete without incorporating complementary EMS capabilities.
DoD has made some moves, at least nominally. It has designated the Army to lead the Joint Counter-Uncrewed Aerial Systems Office (JCO) and the Counter Uncrewed Systems Warfighter Senior Integration Group, while U.S. Northern Command and Indo-Pacific Command have taken the reins on leading counter-drone operations. However, these joint organizations are somewhat ineffective as the Services do not utilize them in favor of their own siloed cUAS strategies. This perpetuates incompatible data formatting and communications protocols, requiring third-party translation layers or interoperability middleware, like Parry Labs or Palantir’s Gotham. Former Secretary of Defense Lloyd Austin also signed off on a classified plan before his departure (unclassified fact sheet here) to streamline the Department's counter-drone strategy, and Air Force Special Operations Command is exploring its own playbook. The result is a somewhat patchwork ecosystem that, at the end of the day, may result in duplication of effort, operational latency, and coverage gaps.
Even so, it’s not enough. EMS spans multiple branches, each with its own priorities, budgets, and operational doctrines. A siloed and sprawling structure hinders the ability to move quickly enough to adapt to and prepare for a rapidly evolving threat. Ukrainian UAS designs are updated in weeks-long cycles, as constant iteration equates to survival in an electromagnetic battle. The ability to effectively use cUAS is not compatible with the Pentagon’s current infrastructure, which is built around closed, proprietary systems that do not easily integrate modular and software-defined solutions needed to compete. Drone warfare is not only reshaping tactics; it forces militaries and homeland security forces everywhere to favor agility and machine-speed decision-making to adapt to cluttered battlespaces, including domestic land.
The United States began experimenting with targeted interdiction of high-value targets during the Vietnam War. This effort reflected an overall need to integrate intelligence, surveillance, and strike, also precursors to what would later become the Find-Fix-Finish model. Not until the early 2000s was it codified in military doctrine, given its omnipresence in GWOT-era priorities in Iraq and Afghanistan. It was designed for hierarchical networks and human-led linear targeting, when time was measured in hours and days, not seconds and minutes.
A traditional Find-Fix-Finish structure breaks down when applied to countering drones as it assumes an operator can reliably find the threat, that there are a limited number of targets, that they will have time and clarity to track and locate precisely, and decide to neutralize. They also assume wide-area engagement zones, meaning an obsolete assumption that there will be time to spot, track, and intercept threats over large, open, and uncluttered environments like skies over oceans or deserts. With hundreds of drones in play, jamming becomes a coordination and prioritization issue, not simply a capability one, which is where tech like CX2’s comes into play.
Traditional architecture is effective for strategic air defense — think: ballistic missiles, jets at altitude, cruise missiles). The kill chain must not only be compressed for speed, but redesigned for automated or distributed decision making to counter drone threats (operating close together, at low altitudes, and with different line-of-sight challenges and engagement criteria). An updated kill chain should incorporate sensor fusion, AI/ML-enabled tracking, and autonomous decision loops that enable the system to anticipate versus react.
A detect-decide-defeat model is more adaptive for modern-day threats, where “detect” is able to continuously discriminate, not just identify, in an environment saturated with noise, spoofing, and low-size, weight, and power (SWaP) drones. It is not simply about spotting the threat, but classifying intent and prioritizing them amidst clutter. The “decide” piece must be able to operate under degraded communications in a matter of seconds, relying on a variety of sensors (as noted: EO/IR, RF, radar, and acoustic), each possessing its own advantages depending on the operational context. This demands robust C2 systems capable of real-time tasking and deconfliction to fuse data from each sensor being used to efficiently prioritize true threats (more on next-gen C2 below).“Defeat” includes graduated, non-kinetic options like jamming versus only lethal end state per “Fix’s” kinetic bias; this nuance will require the ability to adaptively select the appropriate effector (e.g. directed energy via high-powered microwave per Epirus's Leonidas or lasers like Aurelius System’s, interceptors like Allen Control Systems or ZeroMark, or jammers like DroneShield’s). Below is a chart illustrating key operational difference between the two targeting paradigms, not intended to be exhaustive.
Find-Fix-Finish operates on the logic of asymmetry of value, expending significant time and assets to eliminate a single high-value node, while Detect-Decide-Defeat reflects a new logic: asymmetry of volume. Who can process more data faster, make faster decisions, and apply scalable effects will gain operational or strategic advantage and leverage. Ultimately, nomenclature comes and goes with passing wind in the Pentagon, but the real weight is in the strategic implications of the shift described. The shift from asymmetric value to volume reframes how the Joint Force should be thinking about targeting, and reveals limitations in overly relying on a single sensing modality. This begs a fresh look at underleveraged modalities offering complementary strengths.
So, let’s be clear that acoustic sensing is not new technology. In fact, it is over a century old dating back to World War I, when militaries used sound ranging to locate enemy artillery by analyzing the time delays between sound waves reaching a distributed array of microphones or observers. During the Cold War, underwater passive sonar underpinned all anti-submarine warfare, during which U.S. and Soviet navies deployed Sound Surveillance Systems across the Atlantic to detect low-frequency signatures of adversary submarines (though this conversation pertains largely to land-based acoustics). And, in Iraq and Afghanistan, acoustic sensors were used to detect and localize gunfire and sniper shots, roadside bombs, or improvised explosive device detonations in real time.
In any case, acoustic sensing is not the shiniest nor most enticing technology, and its historical use brands it with a dreaded “legacy” tag, which may explain the lack of hype around it. I spoke with Jonathan Hunter, currently CEO at Ziz Defense, and he named it: “Acoustic is boring- it’s just sound. But it hears what you can’t see,” not to mention its ability to operate in RF-denied, GPS-jammed, and line-of-sight obstructed environments. It is also inexpensive, which is something Jonathan and the Ziz team are optimizing for. Many multimodal systems come with $2-3M price tags, while acoustic-based platforms offer affordability and modularity, making them highly complementary and primed for systems integration when it comes to cUAS. Jonathan cleverly identifies this attribute as the “Walmart of war,” commodity-based systems that really work everywhere. Norway-based Squarehead Technologies is widely regarded as the frontrunner in tactical acoustic sensing; U.S. Special Operations Command has already fielded 166 of their surveillance microphones to date.
In a landscape enamored with radar, RF, and high-end effectors, acoustic sensing remains an underestimated modality, despite proving its utility in high–threat environments. They have played a key role in Ukraine, providing passive and low SWaP detection of long-range, low-speed, and low-altitude threats like Iranian-made Shaheds which overwhelm air defenses with relative ease. Acoustic sensors not only detect threats that other sensors cannot, but provide early warning to protect sensors like radar that are vulnerable to jamming. Companies like Zvook were established in the early months of Russia’s invasion of Ukraine soon after noticing these capability gaps, and in speaking with their team they described that it was the only modality that allowed them to keep up with these unforeseen, low-speed, and low-altitude threats.
Each sensor has its drawbacks, and some of acoustics’ come in the shape of background noise in cluttered environments (resulting in false positives); they typically function best in shorter range, up to a few kilometers making them primed for detection and less for classification or localization; and atmospheric conditions may distort sound propagation (requiring adaptive calibration or multimodal fusion to maintain accuracy). As such, it is not a standalone solution, but a critical component of a multi-sensor fusion stack that enables successful cUAS. And, as with everything else, recent advances in machine learning and digital signal processing have significantly boosted their effectiveness, particularly as the Pentagon is prioritizing distributed sensing.
The case for acoustic sensing is not about novelty, but about relevance. Its value lies in what it enables that other sensing modalities cannot consistently deliver- not by doing more, but by doing differently. Innovation leaves space to reimagine what already works for the existing threat environment, which may or may not include glossy hardware, and which may include retooling old tech with new tricks.
As noted, many C2 systems are clunky, siloed, and built on decades-old architectures that are no longer suitable for modern-day warfare (e.g. U.S. Army’s Tactical Airspace Integration System, Global Command and Control System which is transitioning to JADC2-informed redesign) — especially in the context of rapidly proliferating UAVs which demand distributed response coordination. This kind of lag is becoming less an inconvenience and more a liability as the rest of the world evolves.
Selling to the Department necessitates that you speak the same language, which includes understanding that JADC2 is conceptual in nature, not a tangible product. It is a vision for how the U.S. military must integrate amidst distributed decision-making across all domains of warfare. And, yes, the irony is not lost that each service is working towards this joint effort individually (e.g. USAF with Airforce Battle Management System, Army with Project Convergence, Navy with Project Overmatch), but the startups that succeed will be able to speak to each of these dialects, understanding that there is no single JADC2 interface. A winning approach will align with the principles (i.e. real-time data fusion, autonomy, and open interfaces) while supporting translation and interoperability between different service-specific architectures. Startups have taken note and are building platforms emphasizing ease of use and flexibility, critical in GPS and RF-denied environments where centralized decision loops fail. This reduces operator cognitive load and enables faster operations as the ability of adversaries to adapt in real time has increased exponentially. Moving away from closed and monolithic systems enables seamless integration with a broader array of sensors and effectors.
AI and data fusion are at the heart of this shift. Advanced algorithms are able to sift through massive volumes of sensor data in real time and at the edge, which is necessary in cUAS where time is measured in seconds or minutes. In managing an environment rife with UAVs, the “control” piece is less important than “cooperation,” demanding that rigid top-down structures shift to adaptive cooperation among all actors contributing to the solution of common or complementary problems. Centralized control works when communications are reliable and threats are linear, but a drone-dense and EW-contested environment breaks down this model, as the Find-Fix-Finish gaps portray. This breakdown is what gives rise to a Detect-Decide-Defeat kill chain built not only for speed, but for ambiguity, adaptability, and volume.
Centralized control will be a bottleneck when adversary tactics are able to evolve in real time and data links are jammed, making distributed tasking and autonomy the new imperative. This autonomy must not only highlight threats but evolve towards true operator decision support, which calls for reliable, untampered, and training data from source to finish (a problem Hardshell is working on, and one for founders to keep in the back of their minds as the prospect of procurement nears). Companies developing new systems operate where data links are intermittent or compromised, taking the bet that speed, modularity, and intuitive design will define the next generation of C2.
Anduril’s Lattice is undoubtedly the first that comes to mind, excelling in multi-sensor and multi-effect fusion rapidly. It builds both the effectors and the “brain,” limiting the need for any third party-integration, but inviting scrutiny around ecosystem lock-in and the long-term flexibility it affords (or not). As DoD moves towards more open architecture and long-term modularity alongside increasing reliance on a growing and varied startup ecosystem, it will face questions about how genuine broad interoperability is enabling a federated joint force. Lattice is, at the end of the day, proprietary and vertically integrated, faster and easier to use out of the box but obstructing others’ participation in the innovation ecosystem.
This contrasts with platform agnostic startups built to fuse third-party sensors natively, which is relevant where operational theaters involve layered sensor networks composed of multiple vendors. Such features make it an ideal decision-maker for joint, coalition, or multi-vendor C2, and reflective of a complex and messy real world.
Operational demand, budget alignment, and technical readiness have converged to create a critical inflection point. The C2 redesign effort is underway, but progress and urgency lag that of our adversaries, and the United States cannot risk falling behind in a fight evolving faster than the DoD acquisition system can respond.
There is a clear demand pull that did not exist even three years ago. In seeking investment opportunities in sensing tech, particularly acoustic, investors should work to understand where RF, radar, or EO/IR can’t perform reliably, in tandem with where acoustic is able to uniquely provide value. And even further, look for teams advancing acoustic with machine learning (it is not worth the time without), maintaining low cost, and ensuring easy integration with primes and proven success in live combat zones.
With C2, investors should seek companies enabling modular and interoperable architectures that translate between systems and new modalities; modularity reduces vendor lock-in, ensuring rapid iteration so existing systems may evolve alongside emerging threats. These architectures must also be able to operate in GPS-denied, RF-degraded, or other contested environments. And, in general, investors should look for clear signs that the tech is being pulled into the mission beyond simply generating revenue, whether through repeat prototype pilots, or non-traditional acquisition pathways like Small Business Innovation Research (SBIR; most notably through Phase III), Other Transaction Authorities (OTA) to get tech in the field quickly, or traction with RDER. Integration with primes or other existing programs of record also matter, as they signal a startup’s ability to plug and play.
Standalone capabilities will not define the next generation of defense tech. Rather, it will be how well DoD can connect, accelerate, and scale them. Markets and adversaries will not wait for doctrine and legislation to catch up, and nowhere is this more apparent than in the cUAS fight.
and my two incredible mentors here at Squadra who made this possible and whose guidance has been indispensable, Isaac Carp and Nav Vishwanathan.