Thesis: Defense tech is being restructured around software-defined, attritable, AI-piloted systems where the unit of value is shifting from the physical platform to the autonomy stack, mission data, and C2 layer that orchestrates distributed effects at machine speed.
---
1. Attritable autonomous platforms (air, land, sea)
Companies: Anduril (Ghost, Roadrunner, Fury, Altius), Shield AI (V-BAT), AeroVironment (Switchblade, Puma, JUMP 20), Kratos (XQ-58 Valkyrie, BQM-167), Skydio (X2D/X10D), Teledyne FLIR (Black Hornet)
Dynamics: Hardware is becoming a high-volume, lower-margin carrier; differentiation is moving to autonomy software, sensor fusion, and production speed rather than platform survivability alone.
---
2. Mission autonomy, AI, and C2 software
Companies: Palantir (Gotham / Maven Smart System), Anduril (Lattice), Shield AI (Hivemind), Scale AI (Donovan), EpiSci, Applied Intuition
Dynamics: This layer is becoming the new system of record for tasking, ISR fusion, targeting, and mission planning. Contracts are software-heavy, sticky, and feed on operational data flywheels.
---
3. Battlefield networking, electronic warfare, and counter-UAS
Companies: Epirus (Leonidas), Anduril (Pulsar), BlueHalo (Titan / LOCUST), D-Fend Solutions (EnforceAir), Hidden Level, Dedrone
Dynamics: The cheap-drone threat is driving rapid procurement of detection, non-kinetic defeat, and software-defined electronic attack. Spending is shifting toward non-traditional vendors that can field quickly.
---
4. Edge compute, sensors, and open-architecture mission systems
Companies: Mercury Systems, Parry Labs, Curtiss-Wright, NVIDIA (Jetson/Orin), L3Harris
Dynamics: Low-SWaP processing and open standards like MOSA/SOSA are enabling autonomy software to port across platforms. Incumbent ruggedized hardware providers are being pressured by faster commercial silicon refresh cycles.
---
5. Synthetic environments, training, and validation
Companies: Applied Intuition, Scale AI, Bohemia Interactive Simulations (BISim, now BAE), Red 6, CAE
Dynamics: Autonomy programs cannot be validated with live tests alone; synthetic data, simulation, and red-team evaluation are becoming acquisition gate infrastructure for AI-enabled systems.
---
Where the most value accrues
The mission autonomy, AI, and C2 software layer wins the most structural value.
It owns the observe–orient–decide–act loop, controls the data flywheel, sets interoperability standards, and generates software-like gross margins. Platforms are increasingly swappable carriers; the OS layer taxes all of them. Near-term budget still favors hardware, but long-term value pools will concentrate in the software/C2 layer.
---
White-space opportunities
- Neutral, cross-vendor autonomy orchestration/runtime layer for mixed fleets under degraded comms No current player cleanly owns a productized runtime that can deconflict, re-task, and verify behavior across multi-vendor drones and loitering munitions — e.g., a U.S. Army Ghost-X, a Navy V-BAT, and a partner nation drone — under GPS and comms denial. Existing offers like Lattice and Hivemind are largely vendor-centric. Less sure: DARPA/DoD efforts like CODE or Golden Horde may have partial research prototypes, but no neutral commercial owner has emerged.
- Edge-native operational data logistics and after-action model fine-tuning for denied environments There is no clear player that captures, curates, provenance-tracks, and synthetically regenerates multi-sensor tactical data at the edge to fine-tune autonomy models without raw cloud upload. Current players such as Palantir and Scale mostly centralize data; the disconnected, edge-native learning loop for autonomous systems remains open. Less sure: some classified or prime internal capabilities may partially exist but are not productized across services.