A modern military EO/IR system is a remarkably dense piece of engineering:
Optics → detector → electronics → processing → stabilization → thermal management → mechanics
Every one of these blocks consumes size, weight and power. And the challenge becomes particularly severe when the system must also include: visible imaging, thermal imaging , laser range finding ,laser designation ,stabilization ,GPS/INS ,image processing ,automatic tracking. The market is therefore moving toward a very different question: How much capability can we put into every gram of payload?
That question is becoming particularly important for small UAVs, launched effects, robotics, soldier-worn systems and increasingly small satellites.
1. Why Is the Industry Trying to Make EO/IR Sensors So Small?
The platform is getting smaller — and much more numerous. A sensor weighing 5 kg may be perfectly acceptable on a large tactical UAV, but it becomes impossible on a 2–5 kg system. At the same time, military UAV procurement is moving toward much larger numbers. The U.S. Army's early Raven program was measured in thousands of systems; the Ukraine conflict has demonstrated the possibility of producing and consuming hundreds of thousands to millions of small drones per year. Ukrainian production, for example, increased from roughly 5,000 drones annually in 2022 to approximately 4.5 million FPV drones in 2025.
This changes the economics of the sensor. The future requirement is not simply to put the best EO/IR system on the best UAV. It is increasingly to provide useful sensing capability at a price, weight and power level compatible with mass deployment.
The U.S. Army is already signaling this direction. Its lightweight LWIR program targets a complete thermal sensor module weighing only 5–10 g including the lens, while maintaining operation at ≥30 Hz. At the same time, the Army is seeking modular small-UAS payloads combining EO/IR imaging with laser range finding and laser designation. The resulting technology challenge is therefore not simply miniaturization. It is capability density:
How much sensing capability can we put into every gram of payload?
That question is becoming increasingly important as the number of UAVs deployed on the battlefield grows.
The next generation of military EO/IR sensors will not be designed for a world with thousands of UAVs. They will increasingly be designed for a world with hundreds of thousands or millions.
2. The Fundamental Miniaturization Problem
There are five major engineering constraints in this demand from the field.
2.1. Optics: Long-range imaging requires aperture and focal length. A larger aperture collects more photons and improves resolution, but increases with lens diameter => mass => mechanical structure => stabilization requirements. You cannot simply shrink a 100-mm-aperture optical system to 20 mm and expect identical performance.
2.2. Detector: The detector must convert a very small amount of incoming photons into a useful signal. For thermal imaging. Thus, the choice is between uncooled LWIR and cooled MWIR/LWIR has enormous consequences. Uncooled detectors can be extremely small and low power. Cooled detectors generally provide substantially better sensitivity and long-range performance—but require a cooler, power and thermal management.
This is one reason the industry is investing heavily in high-operating-temperature MWIR detectors. LYNRED, for example, is developing compact MWIR technologies specifically around the SWaP requirements of UAVs and small gimbals. Its 2026 ARGO SL is explicitly positioned as a compact MWIR module for tactical drones and soldier-borne systems.
2.3. Electronics: The detector is only useful if the electronics can process the data.
Modern systems increasingly integrate image correction, stabilization , compression , tracking , target detection , AI classification directly into the payload. This critically changes the scaling equation.
The sensor is no longer simply: camera → video output. It increasingly becomes: sensor → processor → intelligence
Why does processing make the sensor smaller?
A conventional architecture may require: Sensor → high-speed interface → cable → external processor → storage/communications. Each of those interfaces requires connectors, PCB area, power conditioning, high-speed electronics and often additional cooling. If processing moves into the sensor itself: Sensor + processor + memory → compact payload, the external system no longer needs to carry a separate processing computer or move enormous amounts of raw image data through high-speed interfaces.
For example, a 640 × 512, 14-bit thermal detector operating at 60 Hz produces roughly:
before additional metadata or processing. A sensor that can identify a vehicle, track it and transmit only the relevant coordinates and a compressed image can reduce the downstream data requirement by orders of magnitude. That has consequences beyond electronics: Less raw data → smaller interfaces → less cabling → lower communication power → smaller processor elsewhere → lower payload mass
This is particularly important for small UAVs, where the processing computer, data link and power system can otherwise become a significant fraction of the payload.
AI also changes what the sensor needs to measure: There is another, more subtle effect. If the final objective is target detection, the sensor does not necessarily need to deliver the same amount of information required for a human watching a full-resolution video stream. An onboard processor can dynamically determine where information matters. For example: Full image: → detect potential target → identify region of interest → increase processing/measurement there → track target → transmit only relevant information. This creates an opportunity for sensor-compute co-design. The detector, optics and processor can be designed together rather than treating the camera as a generic video source. That could eventually allow smaller sensors to achieve useful mission performance without reproducing all the capabilities of a much larger conventional EO/IR system.
But there is a trade-off: Onboard processing does not mean computation is free. An AI processor consumes, electrical power, generates heat, requires memory, requires packaging, and potentially requires additional thermal management. Therefore, simply adding a powerful GPU would defeat the purpose of miniaturization. The trend is instead toward specialized low-power ASICs, NPUs and sensor processors that perform only the operations needed by the payload. This creates an important technology loop: Better algorithms → less data → less processing → lower power → less cooling → smaller electronics → smaller sensor.
And increasingly: Better AI → less information needs to leave the sensor → smaller communications system → smaller overall platform. This is why the electronics trend is not merely about making processors smaller. It is about moving intelligence closer to the photons.
2.4. Stabilization: This is an underrated problem. A long focal-length lens magnifies angular disturbances. A few tens of microradians of platform movement can move a distant target by meters. Consequently, a military EO/IR payload needs, precision bearings , gimbal motors , position encoders , inertial sensors , control loops , vibration isolation. Miniaturizing these mechanisms without losing pointing stability is extremely difficult.
The market is attacking the problem in three complementary ways.
2.4.1. Make the mechanical gimbal smaller and smarter: Instead of using large conventional motors and bearings, manufacturers are using direct-drive motors, miniature bearings, compact encoders and MEMS inertial sensors. The objective is to remove mechanical mass without sacrificing control bandwidth.
NextVision is a good example of how far this approach has progressed. Its product family currently spans from approximately 125 g to 2.15 kg, while retaining stabilized EO/IR capability across the range. Its 125-g DragonEye2 combines visible and 640×480 thermal imaging with 2- or 3-axis stabilization in a 40 × 40 × 65 mm package.
The important engineering achievement is not simply the small motor. It is the entire stabilized optical assembly becoming smaller. This trend becomes particularly interesting when combined with optical stabilization. Because instead of moving the entire camera, a system can potentially use a small fast-steering mirror or optical element to correct high-frequency jitter while the main gimbal handles slower, larger movements.
The architecture becomes:
· Large platform motion-> miniature gimbal
· High-frequency vibration -> electronic/optical stabilization
· Residual image motion-> digital image correction.
This creates a hierarchy of stabilization rather than asking one mechanical system to do everything.
Why this matters for miniaturization: The key is that stabilization performance does not have to scale entirely with mechanical precision. As MEMS gyros, encoders, processors and image algorithms improve, part of the stabilization problem can move from: metal + bearings + motors to: sensors + computation + software. That is a powerful trade.
And we can already see the result in the market: NextVision's current stabilized-camera portfolio ranges from 125 g to 2.15 kg, while Trillium's HD25 provides stabilized EO/IR imaging at around 350 g. The future therefore isn't necessarily a smaller traditional gimbal. It is more likely to be: A smaller mechanical platform carrying a much smarter stabilization system.
The fundamental limitation remains: There is, however, a limit to how much can be moved into software. Digital image stabilization can correct the image after the photons have already been detected, but it cannot recover information that was lost because the image moved across too many detector pixels during exposure. Similarly, if the optical line of sight itself must remain precisely pointed at a target—for example for laser ranging or designation—software cannot replace physical pointing accuracy.
So the hierarchy is likely to remain: Mechanical stabilization → optical/electronic stabilization → digital correction. With each layer removing some of the burden from the layer above it. That is the important miniaturization trend: the industry is not eliminating stabilization; it is distributing stabilization across mechanics, MEMS sensors, optics and computation.
2.4.2. Move part of the stabilization from mechanics into electronics: This is probably the more important trend for further miniaturization. A conventional architecture might require the gimbal to physically remove almost all platform motion. But modern systems can allow some residual motion to remain and correct it electronically using: IMU → image motion estimation → digital correction → stabilized image
Trillium's 350-g HD25-LV, for example, uses a co-located IMU, onboard electronic stabilization and scene tracking, while providing EO/IR imaging in a compact Group 1/2 UAS payload. This changes the optimization problem. Instead of designing the mechanical system to achieve: 100% mechanical stabilization the engineer can divide the task:
· Mechanical stabilization handles large, low-frequency motion
· Electronic stabilization handles smaller, higher-frequency residual errors
That can eliminate some mechanical precision and therefore reduce motor size, bearing requirements and mechanical structure.
2.4.3. Combine mechanical and electronic stabilization: This hybrid architecture is probably the most practical path for small military systems.
A miniature gimbal can provide the coarse line-of-sight control, while electronic stabilization removes the remaining jitter. This allows the mechanical system to be optimized for range of motion and large disturbances, rather than trying to correct every tiny vibration mechanically.
There are already commercial examples of this approach. The ASIO 155 micro-gimbal, for example, combines two-axis gyro-stabilized direct-drive mechanics with a third axis of electronic stabilization and specifies approximately 25 µrad angular resolution. It also integrates EO/IR processing, target detection and tracking.
This is an important architectural shift: Large system: Mechanical precision → mechanical precision → mechanical precision. Small system: Mechanical stabilization + MEMS sensing + computation. The latter allows some of the precision that previously had to be manufactured into the mechanical system to be created algorithmically.
2.5. Thermal management: This becomes particularly important when you combine: detector + processor + laser + gimbal electronics inside a tiny enclosure. A 10-W payload on a small UAV is a significant thermal load. A 50-W payload is much more difficult. And a cooled MWIR system introduces an entirely different thermal architecture. The result is that power and heat often become hidden limits on miniaturization.
How Manufacturers Are Attacking the Problem: The interesting part of the market is that manufacturers are attacking different constraints simultaneously.
2.5.1. Smaller detectors: This is probably the most fundamental development. LYNRED's ATTO640D-02, for example, is designed specifically for SWaP applications and uses wafer-level packaging and high-volume assembly. It provides 640×480 thermal imaging with <300 mW power consumption in digital mode. That is an important change in philosophy: Instead of building a smaller thermal camera around a conventional detector, build the detector itself for a small camera.
2.5.2. Higher-temperature MWIR: Cooled MWIR traditionally meant: detector + cryocooler + control electronics + thermal isolation. That is extremely difficult to put onto a micro-UAV. Higher-operating-temperature detector technologies attempt to reduce the cooling burden.
NASA and JPL have pursued similar approaches with their HOT BIRD detector technology, targeting MWIR operation at substantially higher temperatures than conventional cryogenic systems. This is strategically important because eliminating or shrinking the cooler can remove one of the largest subsystems from an IR payload.
2.5.3. Wafer-level and advanced optical packaging: Another important direction is manufacturing optics and detectors more like semiconductor components. LYNRED's ATTO platform explicitly uses wafer-level packaging and high-volume assembly processes.
The advantage is not simply smaller size. It potentially enables:
lower cost + repeatability + automated assembly + smaller packaging. This is essential if EO/IR sensors are eventually going to be deployed by the thousands rather than tens.
2.5.4. Integrated processing The sensor itself increasingly contains computing. Teledyne FLIR's Hadron 640 series is a good example at the module level: the 56-g module combines visible and thermal imaging and consumes about 1.8 W in steady state. It is explicitly designed for integration into small platforms. This is important because the OEM does not have to build a complete thermal camera from scratch. The trend is: detector → camera core → EO/IR module → stabilized sensor → sensor + AI The boundary between "camera" and "computer" is disappearing.
3. What Does the Market Actually Look Like Today?
Here is the more interesting part. There is already an enormous range between ~100 g micro-gimbals and multi-kilogram military turrets.
Representative publicly documented systems:
These figures reveal something important: there isn't one miniaturization curve. There are really several different markets. The NextVision DragonEye2, for example, puts 1080p visible imaging and 640×480 thermal imaging into a 125-g stabilized package. Trillium's HD25-LV goes to 350 g, while adding a 1280×720 visible camera, 640×512 LWIR, onboard processing and stabilization. Its average power consumption is around 10 W. NextVision's Raptor reaches 640 g with 1080p EO and 1280×720 IR, while the larger Condor reaches 1.5 kg and adds dual IR sensors and an optional eye-safe LRF. At the higher-performance end, Safran's JIM Compact combines cooled MWIR, HD visible imaging, low-light imaging, LRF, GPS and digital compass in a package below 2 kg. Its LRF reaches 12 km. And Thales' Sophie Ultima combines thermal imaging, a daylight laser rangefinder and other targeting functions at approximately 2.5 kg, with a stated 15-km LRF.
4. The Most Interesting Number: 125 g
The 125-g DragonEye2 is particularly interesting because it demonstrates how much capability can already be packed into a very small stabilized payload. It provides: 1080p visible imaging ,640×480 thermal imaging ,40× EO zoom ,4× thermal zoom, three-axis stabilization options ,IP54 protection in approximately 40×40×65 mm. It has also been demonstrated on very small tactical UAVs. One reported integration placed the DragonEye2 on a 1.58-kg fixed-wing UAV that flew an eight-hour ISR mission. That is a very useful illustration of what miniaturization changes. The sensor isn't simply lighter. The entire aircraft architecture becomes different.
5. But Look at What Happens When We Add a Laser: The smallest EO/IR cameras are not necessarily the smallest complete military EO/IR systems. The reason becomes clear when we progressively add laser functionality.
Consider CONTROP's STAMP family. The T-STAMP-XR combines a cooled 640×512 MWIR camera, HD visible camera and three-axis stabilization in a 178 mm, 3.5 kg package. The T-STAMP-XED adds an ~8 km laser rangefinder while remaining in approximately the same 178 mm envelope and only slightly increasing system mass.
But adding a laser designator produces a very different result. CONTROP's T-STAMP-XD increases the gimbal diameter to approximately 220 mm and system mass to roughly 5.8 kg, while adding laser designation capability to the EO/IR and ranging functions. CONTROP itself described the transition from the 178 mm XR to the larger 220 mm XD as a consequence of integrating the designator.
The reason is fundamental. A laser rangefinder only needs to measure the return from a target. A laser designator must deliver a controlled, high-energy optical beam onto that target and keep it precisely aligned with the stabilized line of sight. This places additional demands on the laser source, beam-conditioning optics, thermal management, mechanical structure and—most importantly—the stabilization and bore-sighting architecture.
The progression is therefore not simply EO/IR → EO/IR + laser. It is EO/IR → measurement → targeting.
And the transition from measurement to targeting is where miniaturization becomes substantially harder.
6. What Are We Actually Sacrificing? : The industry's miniaturization race does not mean that performance is preserved perfectly. Instead, engineers decide which capabilities can be sacrificed.
The trade-off looks roughly like this:
This is why saying: "This sensor weighs 350 g and that one weighs 3.5 kg." doesn't tell us very much by itself. The real question is: What capability did the 350-g system give up to achieve a ten-times reduction in mass?
The market is not sacrificing performance randomly. It is sacrificing the performance at the edge of the mission envelope. The trend is toward designing sensors around the minimum aperture, focal length, cooling capacity, stabilization accuracy, laser energy and optical channels required to accomplish a specific mission. Where physics allows it, some of the lost hardware capability is being replaced by computation: electronic stabilization, image enhancement, onboard detection and tracking. Where physics does not allow substitution—photon collection, beam divergence, laser energy or fundamental pointing accuracy—the system simply accepts a shorter effective range or lower performance envelope.
This creates an important distinction between miniaturization and capability density. A 350 g sensor is not necessarily a smaller version of a 3.5 kg sensor. It is often a different optimization of the same physics. The question is therefore not how much performance did engineers remove? but how much useful mission capability did they preserve per gram and per watt?
The major trend in the market is that: The industry is increasingly replacing physical performance with computational performance for further miniaturization.
A Very Clear Example: 350 g vs 1.5 kg. Compare the Trillium HD25-LV with the NextVision Condor. The HD25-LV weighs about 350 g and provides: 1280×720 EO ,640×512 LWIR ,12× EO zoom ,2× IR zoom ,stabilization ,onboard processing ,It consumes around 10 W average. The Condor is approximately 1.5 kg.
But the extra mass buys: 1080p EO ,1280×1024 thermal imaging ,dual IR sensors ,x80 EO zoom ,x9 optical IR zoom ,three-axis stabilization ,optional LRF ,illumination capability ,IP54 environmental protection. So the additional kilogram is not "wasted mass." It buys aperture, optical zoom, thermal performance, stabilization and additional sensing functions.
The Most Extreme Direction: 5–10 g Thermal Imaging: The really interesting development is below the current 100–300-g class. The U.S. Army is explicitly investigating 5–10 g LWIR sensor modules including the lens. The topic generated Phase I awards in 2025. For example, OptX Imaging Systems received a $249,997 Army SBIR Phase I award specifically for A244-024. Its award abstract says the company is developing an optics approach for small-pixel uncooled LWIR imaging systems, with the explicit goal of eventually achieving the 5–10 g overall module weight including the lens. This is not a commercial product yet—it is a development target—but it is important because it tells us where the military believes the technology could go. The Army wants high-resolution thermal sensing at ≥30 Hz within this extremely small mass budget. That would represent roughly: 350 g today → 10 g target or a potential 35× reduction in payload mass. The implication is enormous. A 10-g thermal sensor can potentially be carried by platforms that cannot carry even a 100-g gimbal.
That opens the door to: distributed UAV sensing ,swarms ,expendable reconnaissance platforms ,soldier-borne sensors ,unattended sensors ,miniature robots ,The architecture starts changing from: One expensive UAV + one powerful sensor toward: Many inexpensive platforms + many small sensors
7. Space Is Following the Same Pattern: The same physics driving EO/IR miniaturization on UAVs is also reshaping spaceborne infrared instruments. NASA's Compact Infrared Radiometer in Space (CIRiS) is a particularly good example. Developed by Ball Aerospace with NASA funding, CIRiS was designed as a 6U CubeSat-compatible LWIR radiometer using an uncooled microbolometer. The instrument measures less than 20 × 20 × 10 cm, weighs approximately 1.8 kg and consumes about 9.5 W including heaters. It operates across approximately 7.5–13 µm and performs its own radiometric calibration in orbit.
The key innovation was not simply a smaller detector. CIRiS eliminated the cryogenic cooler entirely and incorporated onboard calibration using deep-space views and internal carbon-nanotube-coated reference panels. NASA's objective was to demonstrate that useful, calibrated thermal measurements could be produced from a spacecraft small enough to be built around a CubeSat architecture.
NASA Goddard is pushing the concept further by attacking the optical architecture itself. Its Compact Thermal Imager technology uses strained-layer superlattice infrared detectors, and subsequent work has demonstrated spectral filters attached directly to the detector assembly. NASA says this eliminates substantial mass associated with conventional filter/lens assemblies and can enable smaller focal planes and, where cooling is required, smaller and more efficient coolers.
The broader lesson is the same as in military EO/IR: miniaturization increasingly comes from eliminating hardware rather than merely shrinking it. Eliminate the cooler. Integrate the filter with the detector. Move calibration into the instrument. Move processing onboard. Reduce the number of optical interfaces. The result is a different architecture: small platform + small sensor + onboard processing + software compensation.
That is why the space market is relevant to military EO/IR. Both are being driven toward the same fundamental question: how much useful sensing capability can we put into a very small mass and power envelope?
Comparison with traditional space IR suites.
8.The Industry Is Moving From Components to Sensor Engines: This is where we think the market becomes strategically interesting.
The old architecture was: Detector→ optics→ electronics→ processor→ external computer
The emerging architecture is: Detector + optics + processing + AI. inside one compact sensor engine. LYNRED's current SWaP products are an example at the detector/core level, while Teledyne FLIR's Hadron family demonstrates how visible and thermal sensing can be packaged into a 56-g OEM module. NextVision is doing something similar at the stabilized-payload level, with products spanning roughly 125 g to 2 kg.
This creates an increasingly modular ecosystem optimized for miniaturization:
Detector companies→ IR cores
Optics companies → miniature lenses / optical assemblies
Laser companies → LRF / designation modules
Electronics companies → processors / ASICs
EO/IR integrators → complete sensor engines
The competitive advantage increasingly comes from how efficiently these pieces are integrated, rather than simply from having the best detector.
What Is Actually Driving the Market? we see five major demand drivers.
8.1. Small UAV proliferation: The first driver is simply the number of small unmanned aircraft being deployed. For a large aircraft, adding a 2 kg sensor may be relatively insignificant. For a 2–5 kg UAV, it can completely change the aircraft's design. A small UAV has a tightly constrained: Payload mass ,electrical power ,aerodynamic drag ,battery capacity ,endurance , center of gravity. This makes the sensor's SWaP directly connected to the performance of the aircraft.
A good historical example: Raven
The U.S. Army's RQ-11B Raven illustrates how this evolved. The Raven itself weighs roughly 4.2 lb (1.9 kg) and is a hand-launched reconnaissance aircraft. In 2012, AeroVironment introduced a 450 g miniature stabilized EO/IR payload containing visible and thermal imaging plus a laser illuminator. The new payload replaced two separate sensor payloads with one package. That's already an important lesson: When the aircraft is small, payload integration becomes an aircraft-level problem. And the direction is continuing.
The U.S. Army's current modular-payload effort explicitly asks industry to develop small-UAS payloads including: EO/IR ,laser rangefinder ,laser designator ,communications relay ,EW ,SIGINT ,using a common interface.
So the market isn't merely asking for "a smaller thermal camera." It is asking: How many different sensing and mission functions can we fit into the payload envelope of a small aircraft?
Why this matters for EO/IR: Suppose a UAV has a 500 g payload limit. You could use: 400 g camera + 100 g electronics. Or perhaps: 150 g EO/IR + 50 g LRF + 50 g processing + 50 g navigation and still have mass available for other functions. That creates a strong incentive for integration and miniaturization.
8.2. Attritable systems: The second driver is different. Small UAVs are increasingly being treated as expendable or attritable assets, rather than miniature versions of expensive aircraft. That fundamentally changes the sensor economics. A $5 million aircraft can justify a very expensive EO/IR system. A $20,000–$100,000 UAV cannot necessarily carry a $1 million sensor. The sensor therefore has to become: small + inexpensive + manufacturable at scale. The Ukraine experience provides an extreme illustration of this change.
The Australian Army Research Centre describes Ukraine's evolution toward high-volume, low-cost UAS, noting that Ukrainian FPV production reached approximately 83,000 per month in 2023, while the number of domestic UAS manufacturers expanded from seven to more than 200.
Old model: Few expensive platforms→ expensive sensor→ maximum performance→ extensive maintenance→ low production volume
Emerging model: Many inexpensive platforms→ inexpensive sensor→ "good enough" performance→ rapid replacement→ high production volume
This changes the question for the EO/IR industry. Instead of: "How do we build the best sensor?" the question becomes: "How do we build a sensor that is good enough, cheap enough and small enough to put on thousands of platforms?" And there is another important consequence. If the UAV itself may be lost, it becomes much less attractive to put a huge amount of value into the sensor. That creates pressure for: commercial detectors ,standardized interfaces ,modular payloads ,mass-manufacturable optics ,COTS processors ,software-defined functionality.
8.3. Autonomy: This is probably the most important technological driver. Historically, a small UAV essentially acted as a remote camera:
Sensor → video link → human operator
The operator had to look at the imagery and determine what mattered. That creates a problem. A 640×512 thermal camera running at 30 Hz produces:
And that's only one sensor.
Add: EO ,IR ,telemetry ,multiple cameras ,several UAVs and the data volume rapidly becomes enormous. DARPA's current Semantically-Aware ISR program explicitly identifies this problem. It says full-motion EO/IR and other sensor streams can exceed tactical communications capacity and is targeting 90% reduction in transmitted data, with an objective of 95–99% in mission-suitable scenes, while keeping the processing around 2 W incremental compute on representative Group 1/2 UAS.
The old architecture: Camera->Transmit everything->Ground station-> Human watches video-
The emerging architecture: Camera-> Detect->Classify->Track->Geolocate-> Transmit only relevant information
The sensor therefore becomes an information processor, not simply an image generator.
A real example: Black Widow. The U.S. Army's Black Widow small-UAV architecture provides a concrete example. It combines the Teledyne FLIR Hadron 640R+ EO/IR module with a Qualcomm RB5 computing platform. The Army describes onboard capabilities including: Obstacle avoidance ,AI-based target identification ,classification ,tracking ,navigation ,operation in GPS-denied environments.
And the Hadron 640 itself illustrates the SWaP trend: 56 g, 640×512 thermal, 64 MP visible, 60 Hz, 1.8 W steady-state, and direct compatibility with AI software and processors such as Qualcomm and NVIDIA platforms.
The camera is not being sold simply as: "56 g thermal camera." It is being positioned as: 56 g sensing front-end + AI ecosystem. That is a fundamental change.
8.4. Distributed sensing: This is where the previous three drivers start interacting.
If sensors become: smaller ,cheaper ,lower power ,more autonomous you can deploy many of them. And that changes the optimization problem.
Imagine two approaches.
Approach A — one powerful sensor: 1 × high-end EO/IR, Long range, High resolution, Excellent stabilization, Large aperture, High cost
versus:
Approach B — distributed sensors: 10 × small EO/IR, Shorter range, Lower resolution, Much cheaper, Much lighter. But now the ten sensors can potentially observe: Different directions , different locations , different spectral bands , different times and their information can be combined. The military value therefore becomes:
Instead:
with additional gains from sensor fusion. This is why the industry is increasingly interested in capability density, rather than simply sensor performance.
The U.S. Army's current modular UAS payload effort is already reflecting this philosophy: EO/IR/laser payloads, EW, SIGINT and communications payloads are all being considered as modular functions that can be integrated onto small UAS.
8.5. Commercial technology transfer: This one is perhaps the least visible driver, but it may be one of the most important. Military EO/IR used to rely heavily on specialized components. Today, many of the enabling technologies are being developed for much larger commercial markets: Smartphones , automotive cameras , automotive LiDAR ,robotics ,autonomous vehicles ,industrial machine vision ,consumer electronics . That gives military sensor manufacturers something they never had at the same scale: commercial volume.
Example 1: SPAD: Take SPAD detectors. The fundamental technology is extremely relevant to military LRF and LiDAR applications because SPADs can detect extremely weak optical signals. But the commercial automotive market is pushing the technology toward: smaller pixels ,stacked architectures ,integrated processing ,higher photon detection efficiency ,lower cost.
Sony's IMX479 is a very good example. Sony integrates the SPAD pixels and distance-processing circuits into a stacked single-chip architecture.
The 2025 device has: 164k SPAD pixels ,10.08 µm pixel size ,37% photon detection efficiency at 940 nm ,520 dToF pixels ,up to 20 fps ,5 cm distance resolution ,up to 300 m stated detection distance.
The automotive market is paying to industrialize technologies that are also valuable to military ranging and sensing. That can eventually lower the cost and size of military components.
Example 2: Thermal imaging. The same thing is happening with uncooled LWIR. Teledyne FLIR's Hadron 640 is a good example of a commercialized thermal module architecture:
56 g, 640×512, 60 Hz, 1.8 W, and integrated visible + thermal sensing. Even more interesting is LYNRED's ATTO640D. It uses: 640×480 resolution , 12 µm pixels , up to 60 Hz , wafer-level packaging , standard surface-mount packaging , <300 mW at 60 Hz in digital output mode. This is exactly the sort of technology that changes the economics of small military EO/IR.
Instead of building a bespoke detector package for every military application, manufacturers can increasingly leverage high-volume packaging and semiconductor manufacturing techniques.
Next: Part 2 looks at who is building these sensors, from LYNRED and Teledyne FLIR to NextVision, CONTROP and the Chinese component makers, and where the market goes by 2040.