Archer identifies.
Bolt intercepts.
A $500 FPV drone costs a defender up to $125,000 to shoot down. Bolt is a small kinetic interceptor built to a sub-$6,000 unit cost, carried and launched by the Archer carrier. No explosive warhead. No datalink after release. No ground radar dependency.
The cost-exchange ratio is broken.
Every fielded kinetic interceptor costs more than the drone it kills. Ukraine theater data puts the exchange ratio between 13:1 and 2,750:1 in the attacker's favor. Against a threat that scales, a defender who wins every engagement still loses the campaign on ammunition cost.
"We can't shoot a $130,000 missile at a $1,000 drone."
Gen. James Mingus, Vice Chief of Staff of the Army$125,000 against a $500 target
A Coyote Block 2 runs up to $125,000 per engagement against a $500 FPV drone. That is 250 to 1. DoD budgeted $3.1 billion for counter-UAS in FY2026 and the exchange ratio did not move.
Fiber tethers have no RF signature
A fiber-optic guided FPV drone carries no command link to jam and no emitter to geolocate. Electronic attack has nothing to work against. Kinetic defeat is the remaining mechanism.
No Group 1 program of record
There is no US program of record for purpose-built Group 1 kinetic defeat. The Army's November 2025 RFI sought missile components at 40 to 70 mm diameter, tolerant of 30 g launch, at TRL 5 and above. That RFI describes the physical envelope of a round nobody is building.
400 Shaheds a day, scaling toward 1,000
Russian one-way attack drone production exceeds 400 per day and is scaling toward 1,000 by late 2026. Saturation is doctrine, not a surge. A magazine that costs six figures per round runs dry before the raid does.
Electric interceptors cannot catch fast FPVs
The Anduril Anvil tops out near 45 m/s. A fast FPV drone runs 33 m/s and maneuvers. Quadcopter aerodynamics set that ceiling, and no amount of software moves it. Merops Surveyor solves it with a propeller at 78 m/s and is built for the Shahed class, not the FPV class.
Semi-active guidance needs the radar to stay up
Mach Industries Dart homes on energy reflected from a ground radar. That keeps the round cheap and hands the adversary a single point of failure to jam, to geolocate, and to shoot. It is not fire-and-forget.
Cost per engagement against a $500 FPV drone. Bar length is proportional to cost. Competitor pricing is estimated from public reporting and open sources, not from vendor disclosure. Anduril has declined to publish Anvil unit pricing. The Bolt figure is a design target at volume, not a quoted price.
Reusable carrier, expendable round.
Archer searches, identifies, positions, and holds release authority. Bolt is what it launches. Only the round is consumed, and the round stays small because Archer already closed the distance. A short flyout buys a smaller motor, a lighter airframe, and guidance hardware that does not have to survive a long midcourse.
Archer searches with radar, identifies with EO
Radar covers the search volume far faster than an imager can scan it. Radar finds the target, a gimballed thermal imager identifies it, and Archer hands Bolt a cue. Archer is the command and control interface, not Bolt.
Bolt is fire-and-forget after release
No datalink. No midcourse update. No operator in the loop after launch. The only input into the guidance loop is the round's own seeker.
No explosive warhead
Bolt defeats the target by direct body impact, with kinetic energy well past what a Group 1 airframe survives. The solid rocket motor propellant is energetic and is spent during boost. Nothing terminal is.
GPS appears nowhere in the kill chain
Archer's radar measures the target in Archer's own body frame, and Bolt launches from that same frame. Absolute position and heading cancel out of the engagement geometry. GPS denial is not a degraded mode for this round. It is irrelevant to it.
Dispersal is the deconfliction mechanism
Carriers positioned around what they defend fire inward on different bearings, so rounds cross the raid rather than each other. The dispersed posture that survives counterbattery is the same posture that deconflicts fires.
One operator decision
The system surfaces a single alert with a recommendation. The operator ignores, monitors, or engages. Everything after authorization is autonomous, and there is no abort after release, so collateral control is entirely pre-launch and Archer holds it.
Concept of operations.
Engagement sequence, side view. Not to scale. Modeled results, not test data.
Archer on station
The carrier holds close to what it defends rather than far forward of it. A round has to be launched a full time of flight before the target arrives, and our engagement modeling sets how close that keeps the carrier.
Radar searches, thermal identifies
The search radar covers the sector and the gimballed thermal imager resolves the track into an identification. Onboard processing classifies the target and the system surfaces one engagement recommendation.
Operator authorizes
One decision, made before the shot. Release authority stays with the operator, and it is the last point at which the engagement can be stopped.
Archer positions to a head on aspect
Aspect is a requirement, not an optimization. Engagement geometry drives where the carrier has to be before it releases, and positioning is one of the jobs the carrier exists to do.
Bolt separates, then the motor lights
Igniting on the rail is not survivable for the carrier at any airframe size we modeled. Bolt separates first and lights clear of it. That is an architecture decision the analysis forced, not a preference.
Seeker acquires and guides
The seeker picks the target up as a point source at range and refines the track as it closes. Thermal and radar fuse into a single state estimate, and the round flies itself to intercept on its own sensor.
Kinetic impact, carrier repositions
Body to body impact. No warhead, no fuze, no fragmentation pattern. Archer is already moving to the next engagement or back to station, and reloading on the ground is what sustains the defense against repeated raids.
Sized to the threat, priced to the magazine.
Bolt sits inside the 40 to 70 mm envelope the Army's own component RFI described. Everything about the round follows from one decision: Archer closes the distance first, so Bolt never has to fly far.
Bolt airframe layout, side view. Representative configuration. Dimensions are not published.
Locked means a settled design decision. Modeled means simulation output, never test data. Target means a requirement we design to, not an achieved value. Detailed specifications are shared with program offices and partners under NDA.
Terminal guidance.
Every component in the seeker is a production part available today. The engineering is in the integration, and that integration is where our design work sits.
Thermal imaging
An uncooled long-wave infrared imager detects drone signatures in darkness, smoke, and weather with no illuminator and no emission. It is a strapdown installation with no gimbal, which is what keeps it small enough to fly on an expendable round.
mmWave radar
A millimeter wave radar measures range and closing velocity independent of light and thermal contrast, and reads rotor signatures for classification. It holds engagement geometries the thermal channel cannot cover on its own.
Sensor fusion
Thermal supplies angle, radar supplies range and closing rate, and the two fuse into a single state estimate for the guidance loop. If one channel degrades, the other carries the engagement at reduced performance rather than dropping the track.
Detection chain
At acquisition range the target is a point source, so the chain is built for point sources rather than for recognizable imagery. It refines the track as the round closes and the target resolves.
Onboard processing
Detection, classification, and guidance all run on the round. No datalink, no ground processing, and no dependency on anything outside the airframe once it is away.
Control
Canard control surfaces flown by an autonomous terminal guidance law. Our modeling identified which actuator characteristic actually drives accuracy, and we select hardware against that rather than against raw force.
The only input is the round's own sensor
No datalink, no midcourse update, no operator in the loop after release. That property is enforced and tested in our simulation rather than assumed.
The cost belongs in the same paragraph: there is no abort after release, so collateral control is entirely pre-launch and Archer holds it.
No warhead is not the same as no hazard
Bolt carries no explosive warhead, so a miss leaves one object of known mass on a known footprint rather than a fragmentation pattern. A commanded tumble cuts terminal energy by more than an order of magnitude and pulls the downrange footprint in substantially.
It reduces the hazard. It does not eliminate it, and we do not describe it as safe. Reaching a casualty-safe threshold needs a drag device we have scoped and not built.
Magazine depth.
We modeled the many-on-many problem directly rather than reasoning about it: carriers positioned around a defended area, launches scheduled as discrete events, and time of flight and hit probability drawn from the engagement model rather than assumed. Three results came out of it, and two of them changed the design.
Total magazine has to beat the raid
Carry fewer rounds than the raid has targets and it is a hard failure that no launch cadence recovers. Above that line, geometry binds rather than magazine, and rounds expended per kill stays near one across the whole sweep.
Launch cadence buys almost nothing
Varying the interval between launches barely moves the outcome. Do not spend carrier mechanical complexity on launch rate. Spend it on reloading turnaround, which is what sustains a defense against repeated raids.
Carrier standoff has a ceiling
A round has to be launched a full time of flight before the target reaches what you are protecting, not merely be in range when it arrives. That puts a hard limit on how far forward a carrier can usefully sit, and we found it by modeling rather than by assuming.
Modeled results, not test data. Repositioning between engagements is not modeled, so these are static lower bounds on performance. Magazine depth on a carrier is set by endurance rather than by mass: a carrier holds far more rounds than it can hold while staying airborne long enough to matter.
Maturity.
Martel is at TRL 3, pre-hardware. Every result on this site is modeled and nothing has come off hardware. We publish maturity by area rather than as one program-level number, because the areas are not at the same maturity and a reviewer deserves to see which is which.
| Area | Where it stands | Evidence |
|---|---|---|
| Engagement modeling | Monte Carlo engagement simulation, archived and reproducible runs | Modeled |
| Round configuration | Airframe, mass properties, and seeker architecture settled | Locked |
| Fire-and-forget behavior | Enforced and tested in simulation | Modeled |
| GPS-denied operation | Navigation dependency audit closed. GPS is absent from the kill chain | Modeled |
| Launch architecture | Carriage and release approach settled by analysis | Modeled |
| Hardware | No article built. Bench and flight test are funded Phase II tasks | Open |
| Component selection | Sensing and propulsion settled. Several supporting parts still in trade | Open |
| Unit cost | Cost model in build against a $6,000 target. Vendor quotes coming in | Target |
One baseline, and every number traces to it
Every figure in every simulation, document, and cost line reads from a single controlled baseline. Nothing is hardcoded anywhere else. Each value carries a status tag, and a build that depends on an unresolved value fails a check before it can be produced.
Simulation runs are archived with a manifest and a seed, so any published result can be traced back to the run that produced it.
We would rather find it than have you find it
Our own analysis has overturned conclusions we previously held, including one on what actually drives terminal accuracy. Each correction went into the record with the run behind it, and the design changed as a result.
That is the reason for the provenance discipline. A program office reading a claim from us should be able to ask where the number came from and get an answer.
The Group 1 gap.
Electric collision-kill multirotors are proven, short-ranged, and speed-limited. Missile-class interceptors are fast and cost six figures. A sub-$6,000 solid-motor round with a fire-and-forget seeker sits between them, and that gap is where the Group 1 threat lives.
| System | Speed | Cost per shot | Guidance | Ground radar | Maturity |
|---|---|---|---|---|---|
| Raytheon Coyote Block 2 | Missile class | up to $125,000 | Multi-mode | Yes | Fielded |
| Anduril Anvil | ~45 m/s | $15,000 to $50,000 est. | EO/IR and radar fusion | No | Fielded |
| Merops Surveyor | ~78 m/s | $15,000 | AI machine vision | No | Combat proven |
| Perennial Bumblebee V2 | FPV class | Undisclosed | AI optical | No | Contracted |
| Mach Industries Dart | Solid motor | Undisclosed | Semi-active RF | Continuous | Flight testing |
| Martel Defense Bolt | Solid motor | $6,000 target | LWIR and mmWave, fire-and-forget | No | TRL 3, pre-hardware |
Pricing and speed figures for other systems are estimated from public reporting and open-source analysis, not from vendor disclosure. Anduril has declined to publish Anvil unit pricing. Merops Surveyor has a stated path to $3,000 to $5,000 per unit at scale and is optimized for the Shahed class rather than the FPV class. Bolt figures are modeled or target values as marked elsewhere on this page, and Bolt has flown in simulation only.
Speed the electric interceptors cannot reach
Solid motor propulsion puts Bolt in a different speed class than an electric interceptor. Anvil is held near 45 m/s by quadcopter aerodynamics. That is a physics ceiling, not a software problem, and it is why Anduril built a warhead variant.
No ground radar dependency
Mach Dart requires continuous ground radar illumination to home. That radar is jammable, geolocatable, and targetable, and it bounds engagement geometry to its own line of sight. Bolt homes on its own seeker. Dart does not.
Open interfaces on a proprietary system
Archer and Bolt are proprietary Martel designs. The interfaces are not: cue and handoff, sensor track input, engagement reporting, and the Bolt to Archer rail interface are documented and publishable. Nothing in our position restricts the Government's ability to integrate or compete the system.
Current pursuits.
We fund development through non-dilutive Government channels. Feasibility work to date has been self-funded and performed by the firm and the principal investigator, with no prior or ongoing federally funded SBIR or STTR work behind it.
Air Force Direct to Phase II
Very Low-Cost Kinetic Defeat for C-sUAS. The topic sets a strict sub-$6,000 unit cost, an 800 m threshold range with a 5 km objective, fire-and-forget operation, GPS-denied capability, and simultaneous engagements. Bolt and Archer are the response.
Navy SBIR
Technical and cost volumes submitted. Separate scope and a separate architecture argument from the Air Force effort.
DHS Request for Information
Counter-UAS capability response submitted. Homeland defense applications favor kinetic energy over explosive effects, which is the posture Bolt was designed to.
JIATF-401 and AFLCMC/ES
JIATF-401 is chartered to put kinetic capability at every location and has emphasized kinetic energy over explosive effects for domestic installations. That is a direct match to a warhead-free round. Engagements are in progress and no award is claimed.
Prior DoD contracts: none. We qualify as a nontraditional defense contractor under 10 USC 4022(d).
Who we are.
Martel Defense is a defense technology company founded by Caleb Ross to close the cost-exchange ratio on Group 1 drone defeat. Our engineering leadership comes out of Raytheon Missiles and Defense, Blue Origin, and Collins Aerospace, covering guided weapons development, propulsion and airframe design, flight systems, embedded systems, and RF sensing.
We are early stage and pre-hardware. The architecture is settled, the engagement physics are modeled and archived, and the open items are named rather than papered over. Feasibility to date is analysis, and we describe it as analysis.
Components are sourced from NDAA Section 889 compliant domestic suppliers. Chinese-sourced thermal imagers are disqualified from the bill of materials on that basis, not on preference. Supply chain documentation is available on request.
Archer and Bolt are proprietary systems owned by Martel Defense. The interfaces between them and the outside world are documented and open.
Caleb Ross
Founder and Chief ExecutiveAerospace and propulsion engineering. Leads systems integration, engagement simulation, and autonomy. Sets company strategy and program capture, and serves as principal investigator on our SBIR work.
Prior: Raytheon
Nicholas Crowder
Co-founder, Propulsion and AirframeOwns motor selection, structural design, and the airframe. Responsible for the load cases Bolt is built to and for the propulsion trades behind the current configuration.
Prior: Blue Origin
David Stinson
Co-founder, Electrical and SensorsOwns the seeker electronics and the sensor modeling behind acquisition performance. Responsible for sensor selection and the NDAA Section 889 sourcing position.
Prior: Collins Aerospace
Contact Us
We are in conversation with program offices, operational units, and development partners. If this addresses a gap you actually have, we should be talking. If it does not, tell us why and we will fix the design or the pitch.
caleb.ross@marteldefense.comEmail reaches the same inbox as the form.