Saber Defense • Helix Swarm™
The offensive role of the Helix Swarm dual-role attritable flying drone
Build the Web. Own the Sky.
1,500-drone packages organized as 5 × 300-drone hierarchical clusters (MASAC commander + MADDPG executors) penetrate, saturate, and prosecute high-value targets 100–300 km behind contested lines. ≤5 kg, 1200 mm wingspan, 12–20 hr loiter, tube/canister-launched. Same airframe as Helix Shield — toggled by effector cartridge and MARL policy.
Company Vision
“To become the preeminent U.S. provider of autonomous attritable swarm systems by delivering unmatched endurance, decentralized intelligence, vertical integration, and mission adaptability — redefining the future of tactical and operational advantage for the U.S. and its allies.”
Deliver physics-limited offensive dominance through hierarchical MARL — one airframe, two roles, brigade-scale effects.
1,500-drone packages organized as 5 × 300-drone hierarchical clusters overwhelm enemy air defenses through coordinated saturation, decoy feints, and dynamic re-tasking — all without external C2.
Hierarchical MARL (MASAC commander + MADDPG executors) autonomously identifies, prioritizes, and destroys high-value targets and command nodes 100–300 km behind contested lines.
Emergent self-sacrifice tactics force the defender to expend $1M–$3M interceptors against $49K–$59K attritable Helix Swarm drones. The cost-exchange ratio is permanently inverted.
Identical airframe, ASIC, mesh, and AeroHydroCell as Helix Shield. The only difference is the effector cartridge loaded and the MARL policy assigned at mission planning.
Identical airframe to the Shield role. The Sword configuration loads offensive effector cartridges and assigns hierarchical MARL policies for penetration and strike.
MASAC commander policy explores tactics; MADDPG executors deliver deterministic precision. MARL-LAC layered autonomy with dual rewards (mission success + survival). Trained in 10,000+ adversarial physics-twin simulations against the Helix Shield itself.
EO/IR + RF + acoustic + edge LiDAR + onboard terminal seeker. Same passive multimodal stack as Shield role — re-tasked for target acquisition and BDA instead of threat-web tracking.
90-second tool-less swap on the flight line: HE shaped-charge, EW jammer, ISR/BDA, decoy/spoof, cyber-takeover. Mission-configurable at the cluster level — one airframe, five offensive doctrines.
~9 W average inference, ≤6 ms decision latency. Hybrid ANN-SNN handles terminal guidance and collision avoidance with 3–5× lower power than dense ANN-MARL. Zero dependence on GPS or external C2.
ML-KEM-768 + ML-DSA-65 + AES-256-GCM mesh for theater-scale offensive coordination. Launched directly from ground tubes, naval canisters, or airdrop pods — no fixed ground infrastructure required.
$49K–$59K ASP, ~$8.5K COGS at Y5 scale (≥ $7,900 floor). 10,000 drones/year at Y5 from a single U.S. line; 60,000/yr at multi-site Arsenal scale (Phase II/III reference). 100% sovereign vertical integration — airframe, ASIC packaging, AeroHydroCell, integration.
≤5 kg
Mass
Attritable, tube/canister-launched
1200 mm
Wingspan
Folded for canister stowage
100–300 km
Range
Per-drone operational radius
12–20 hr
Loiter
AeroHydroCell v2.0 boosted
$49K–$59K
ASP
Y1 → Y5 average selling price
300×5
Cluster
1,500-drone hierarchical package
One airframe. Five mission-configurable cartridges. Tool-less field reconfiguration in under 90 seconds — a tactical flexibility no single-mission FPV or loitering munition can match.

Cartridge Bay Cutaway — Single Cartridge Loaded, Hot-Swappable Mission System
Standard
Anti-armor / anti-personnel
Copper-lined shaped-charge with MARL-controlled terminal guidance and impact/proximity fusing.
SEAD
Suppress enemy C2 & GPS
High-power RF module with MARL-adaptive frequency hopping driven by real-time sensor fusion.
ISR / BDA
Intelligence & damage assessment
EO/IR seeker with encrypted downlink. Extended loiter patterns commanded by hierarchical MARL.
Deception
Saturate enemy defenses
Radar/IR signature generators with chaff dispensers. MARL coordinates decoy patterns with strikers.
Non-Lethal
Hijack enemy UAS
Software-defined RF injection enables non-destructive capture of adversary drones for exploitation.
The Helix Sword role fields the same 1,500-drone packages as Shield — organized as 5 × 300-drone hierarchical clusters — but tasked into offensive doctrine:
This dual-reward architecture enables emergent self-sacrifice — agents voluntarily absorb interceptors to create openings for strike elements.
Helix Sword swarms train in 10,000+ parallel physics simulations against the Helix Shield MARL itself. Because both roles share the same airframe, ASIC, and mesh, this self-play produces the most battle-hardened offensive AI in existence — every Sword tactic must beat the world's best defensive web.
Through self-play, Helix Sword develops behaviors no human programmer could script:
Helix Sword enables a three-phase closed-loop approach against deeply buried, hardened facilities:
12–20 hr loiter per drone enables 24/7 pressure. AeroHydroCell v2.0 extends endurance in variable-humidity environments without resupply.
Helix Sword dramatically reduces the need for high-risk SOF insertions in contested, GPS-denied environments:
Their limitation: Scripted or simple RL autonomy. Individual loitering munitions lack true swarm coordination — each drone operates semi-independently with pre-programmed waypoints.
Helix Edge: True hierarchical MARL with 1,500-drone clusters produces emergent coordinated tactics (decoys, flanking, saturation) that overwhelm any defense. $49K–$59K ASP at volume scale; competitors run $50K–$100K+ without swarm intelligence.
Their limitation: Platform-centric autonomy on V-BAT, CCA, Hornet. Focused on piloting individual aircraft, not mass-attritable hierarchical swarm tactics. GPU-heavy compute; no vertical manufacturing.
Helix Edge: Helix Swarm is a co-designed airframe + 16nm SNN ASIC + FPGA bridge (Phase I; 7nm ASIC deferred to Phase II) + MARL-LAC + AeroHydroCell v2.0. Same platform delivers Shield + Sword roles. Hivemind on a Hornet costs $400K+ per shot; Helix Swarm costs $49K–$59K with 12–20 hr loiter.
Their limitation: High-cost attritable aircraft ($2–20M per unit for CCA; Roadrunner-M is recoverable single-shot). Lattice C2 lacks native swarm AI for mass tactics.
Helix Edge: ~50× more agents at <1/40th the cost. True hierarchical MARL overwhelms with numbers + coordination, not individual platform capability. Economics make brigade-scale strikes viable where CCA cannot.
Helix Swarm is the only single-platform dual-role attritable flying drone with hierarchical MARL at 1,500-drone scale. Competitors remain siloed:
DoW attritable autonomy spending: $3–5B FY2026 accelerating to $15B+ by 2030. Replicator 2 explicitly prioritizes mass-producible, AI-coordinated systems. Helix Sword is purpose-built for this exact requirement.
Human-on-the-loop architecture for all lethal decisions. Verifiable ethical AI layers with full audit trails. Compliant with DoW Ethical AI Principles and NATO autonomous-weapons frameworks.
Phased OTA live-fire validation at secured U.S. facility. High-fidelity physics twins with real-world correlation data. Incremental complexity ramp from 10-agent to 1,500-agent operations.
Mitigated by 100% U.S. vertical integration: airframe, ASIC packaging, AeroHydroCell, and final integration. Secured U.S. facility handles warhead loading and live-fire only. Sovereign U.S. supply chain throughout.
Y1 (2026)
Y2 (2027)
Y3 (2028)
Y4 (2029)
Y5 (2030)
The Department of War's flagship initiative to rapidly field thousands of low-cost, attritable, autonomous systems — and the strategic tailwind powering Helix.
~$1B
Initial Funding
FY2024–2025 allocation
18–24mo
Sprint Timelines
Concept to fielding
JIATF-401
C-sUAS Lead
Replicator 2 organization
FY2026
Funding Decision
Full Replicator 2 budget
Launched August 28, 2023, by then-Deputy Secretary of Defense Kathleen Hicks, Replicator was designed to counter China's massed uncrewed systems advantage in the Indo-Pacific by changing the economics and speed of acquisition.
Goal: Field "multiple thousands" of ADA2 systems by August 2025 (18–24 month sprint).
Tranches:
Status (March 2026): Hundreds of systems delivered and transitioned to warfighters by August 2025. Software layer selected from Aalyria, Viasat, Higher Ground, and IoT/AI providers. Focus shifting to end-user adoption by services and SOCOM.
Announced September 27, 2024 via Secretary Lloyd Austin memo — shifts focus to countering the small UAS threat to critical assets and force concentrations.
Now often referred to with a secondary "Department of War" designation under Executive Order 14347. Congress continues to debate oversight and specific capability details.
Primary Threat Driver: China's rapid uncrewed systems buildup and mass production advantage in the Indo-Pacific.
Progress: Officials highlight accelerated acquisition, software collaboration, and hundreds fielded. Replicator has driven cultural change — iterative risk tolerance, commercial integration, and software-centric lethality.
Evolution: Transitioning from DIU-led prototype effort to service/SOCOM execution. Moving from concept to hundreds fielded (Replicator 1) and first operational C-sUAS buys (Replicator 2), with FY2026 funding decisions imminent.
Helix Offensive SWARM aligns directly with Replicator 1's vision: attritable autonomous offensive systems with swarm collaboration software and lethal effects in dynamic environments.
Saber's dual-role Helix ecosystem — defensive MARL-LAC webs + offensive hierarchical MARL swarms on a single airframe — is uniquely positioned to capture significant share in both Replicator phases from a single secured U.S. manufacturing campus. Shared IP and training twins create cross-tranche advantages no competitor can match.
This program is the strategic tailwind that accelerates Helix to prime-contractor status and the $316M+ consolidated revenue target by 2030.