For decades, electro-optics and photonics were often treated as enabling technologies hidden inside larger systems: a camera inside a vehicle, a laser inside a rangefinder, an optical encoder inside a machine, or an imaging sensor inside a satellite.
That is changing.
Across autonomous systems, robotics, defense, aerospace, semiconductor manufacturing and advanced industrial equipment, the ability to see, measure, locate, communicate and interpret the physical world with light is increasingly becoming a defining capability of the entire system.
The most important trend may therefore not be the development of one particular laser, detector or imaging technology. It is the convergence of optics, photonics, sensing, computation and artificial intelligence into integrated perception systems.
This is the trend we will follow at Photonics Intel.

The Short Version

• Optical sensing is becoming more central to intelligent systems. Autonomous machines increasingly depend on cameras, LiDAR, infrared sensors and other optical modalities to understand their environment.
• Miniaturization is changing where electro-optics can be deployed. Advanced optics and photonic integration are making increasingly capable sensing systems possible within tight size, weight and power constraints.
• Photonics is moving closer to the processor. Integrated and heterogeneous photonics are expanding from communications into sensing, computing and AI-related applications.
• Defence is accelerating the demand for compact, resilient sensing. Modern military programs increasingly emphasize low-SWaP EO/IR systems, autonomous sensing, sensor fusion, and processing at the edge.
• The next competitive advantage will increasingly come from the complete sensing architecture, not a single component.

Estimates represent approximate 2026 annual global economic activity associated with each trend. Categories overlap and should not be summed.

1. What Is Happening?

The role of electro-optics is expanding because machines increasingly need to perceive their environment rather than simply execute predefined instructions.  The fundamental driver behind this momentum is due to that:

·        A robot needs to determine where objects are.

·        An autonomous vehicle needs to understand its surroundings.

·        A drone needs to navigate, identify objects and operate in changing conditions.

·        A satellite needs to observe Earth, communicate with other systems and sometimes make decisions without immediate ground intervention.

·        A semiconductor manufacturing system needs to measure structures at extremely small scales.

·        A modern defence platform may need to detect, identify, track and engage objects while operating under severe constraints on size, power, communications and human intervention.

All of these problems have something in common:

The machine needs information about the physical world.

And increasingly, light is the most powerful way of obtaining that information. Market studies indicate that current around 14B USD electro-optics market size is expected to increase to 23 B USD in the next five years.

The European Commission's 2026 assessment of photonics explicitly identifies sensing and imaging alongside aerospace, defence, automotive, computing and communications as major application areas, while highlighting photonic integrated circuits as an increasingly important platform for these systems.

Recent commercial developments illustrate how photonic integration is expanding well beyond traditional optical communications. For example, Aeva is using silicon photonics to build compact FMCW LiDAR systems; QCi and Freedom Photonics are developing integrated photonic platforms for quantum and advanced sensing; and Light matter is bringing photonic engines directly into AI-computing infrastructure through near-packaged and co-packaged optics.

The result is a subtle but important shift:

Optics is no longer simply an input to the machine. In many systems, it is becoming part of the machine's intelligence.

2. What Is Driving the Change?

Several forces are pushing this transition simultaneously.

Technology is becoming smaller

The most obvious trend is miniaturization.

For many years, sophisticated electro-optical instruments required relatively large optical assemblies, substantial electronics and significant power. Today, there is increasing pressure to place similar capabilities on:

As an example you may realize the evolution of the laser range finders in the last two decades as below:

In addition, The U.S. Army, is actively developing compact EO/IR payloads for small unmanned aircraft, explicitly combining improved imaging performance with lower cost and reduced size and weight. This is more than a packaging exercise. Reducing the size of an optical system changes what the system can do.

Example of a micro drone with sensor suite : Black hornet. The sensor suite with thermal imaging and guidance camera. The system is developed by Teledyne FLIR.

A sensor that once required a large stabilized platform may eventually become practical on a very small UAV. A sophisticated ranging or imaging capability can move closer to the edge of an autonomous machine. A technology that was previously available only to large platforms can begin to appear in much smaller platforms with much larger numbers.

3. From Sensors to Perception Systems

It is useful to think about the evolution of photonics in the current market stream in three stages.

The system combines multiple sensing modalities, interprets the environment and contributes directly to an autonomous decision.

This third stage is where things become particularly interesting. Consider a future autonomous platform that combines:

Visible imaging + SWIR + thermal imaging + LiDAR + inertial sensing + AI

Each sensor sees something different. The value doesn't come simply from adding their specifications. The value comes from fusing their information giving a useful feedback towards the decision maker.

Recent defense research illustrates this direction. U.S. Army programs are explicitly investigating fusion of electro-optical and infrared sensing, event-based imaging and AI-based analysis to produce information that cannot be obtained from a single sensor modality.

This is why I believe that the next generation of electro-optical systems should increasingly be judged as “perception architectures” not isolated optical products.

4. Photonics Is Moving Into the Architecture

Photonic technology is increasingly becoming integrated with electronics rather than sitting beside it as a separate subsystem. Silicon photonics and heterogeneous photonic integration are particularly important examples.

Recent work describes heterogeneous integration of lasers, modulators, detectors and other optical functions on common platforms, with applications extending into high-speed communications, LiDAR, spectroscopy, quantum measurement and photonic computing.

This matters because integration can change the economic equation.

A traditional optical assembly might require:

An integrated photonic architecture potentially brings many of these functions into a much smaller and more manufacturable platform. That does not mean traditional optics is disappearing. Quite the opposite.

The likely future is a hybrid architecture in which bulk optics, micro-optics, photonic integrated circuits, semiconductor detectors and electronic processors are combined according to the requirements of the application.

A good example of a hybrid architecture is Ayar Labs' optical I/O system for AI and high-performance computing.

The system combines:

The important architectural point is that the laser is deliberately not integrated into the same chip as everything else. The TeraPHY optical engine is designed to sit within a system-in-package architecture alongside electronic computing hardware. It uses silicon photonics and microring modulator technology, while the laser source is physically separated and supplied remotely. Ayar Labs currently specifies more than 8 Tbps per optical engine, with claimed end-to-end optical I/O latency below 25 ns.

Why this is a hybrid architecture

Ayar Labs could theoretically aim to integrate the laser, photonic circuitry and electronics as tightly as possible.

Instead, it separates them according to their engineering requirements.

This is important because lasers and high-performance processors have very different thermal and reliability requirements. By keeping the laser external, the system can potentially improve thermal isolation and serviceability: the laser source can be replaced without replacing the processor package.

This is therefore a good example of the future not being:

Everything becomes one monolithic photonic chip.

Instead, it is:

Use integration where it creates the greatest benefit, and keep other components separate where physics, heat, reliability or manufacturability make that preferable.

Another example particularly relevant to the EO/photonics world: SiLC's Eyeonic FMCW LiDAR architecture. At its core is a silicon-photonics chip that integrates important coherent LiDAR functions onto a common platform.

The Eyeonic architecture combines:

SiLC describes its silicon-photonic platform as providing chip-scale integration and coherent detection, while the Eyeonic sensor connects the silicon-photonics chip to external scanners. Its broader Eyeonic Vision System combines the sensor with a digital processing solution based on an FPGA and supports different scanner options.

The Eyeonic Vision System is especially illustrative because SiLC explicitly combines the photonic sensor with external scanners and a digital processing solution based on an FPGA. That means the final product is not merely a photonic chip; it is a complete hybrid sensing architecture.

The current product family spans applications from short range to systems specified by SiLC for detection beyond 1 km, depending on the configuration.

Why the hybrid approach makes sense

A silicon-photonic chip is excellent for integrating functions such as:

But a LiDAR system still needs other components that have fundamentally different requirements:

So the silicon-photonics chip becomes the optical engine, rather than the entire sensor.

The key lesson from both examples: Integration is becoming hierarchical.

Instead of: Separate laser +Separate detector+ Separate modulator+ Separate electronics+ Lots of alignment

we increasingly see:

The future of photonics is unlikely to be purely integrated or purely discrete. It is increasingly a layered hybrid system in which each technology is used at the scale where it performs best.

5. Autonomous Systems Are Becoming an Important Driver

One of the most important consequences of these developments is that optical sensing is becoming essential to autonomous machines. A machine cannot act intelligently without knowing something about its surroundings. For autonomous systems, optical sensing provides several unique capabilities:

Imaging: Where are objects?

Range: How far away are they?

Spectral information: What are they made of?

Thermal information: What is their temperature or heat signature?

Motion: How are they moving?

Fine measurement: Where exactly is an object located?

Identification: What is the object?

This is why LiDAR, EO/IR imaging, 3D sensing, spectroscopy and advanced cameras are converging into a common technological landscape. Recent photonics research specifically identifies LiDAR and optical phased arrays as important applications for integrated photonics in autonomous systems.

The same trend appears in industrial robotics and automation, where laser and photonic technologies are increasingly being used for sensing, measurement and adaptive manufacturing.

A useful example of this convergence can already be seen in autonomous vehicles. Consider the perception system of the Waymo Driver, which combines camera imagery, LiDAR point clouds and radar measurements into a unified representation of the environment. Each sensing modality contributes different information. Cameras provide semantic detail, helping the system recognize traffic lights, road signs and objects. LiDAR adds precise three-dimensional geometry and distance. Radar contributes velocity information and remains effective in challenging conditions such as rain, fog and snow.

The value emerges when these information streams are fused. A pedestrian is no longer simply a collection of pixels in a camera image or a cluster of points in a LiDAR cloud. The system can estimate what the object is, where it is in three-dimensional space, how fast it is moving and whether its predicted trajectory creates a risk.  This is an important shift. The objective is no longer to build a better camera or a better LiDAR sensor in isolation. Increasingly, the objective is to build a better representation of the physical world by combining complementary sensors, photonic technologies and electronic processing into a single perception system.

6. Defense Is Pushing the Technology Harder

Defense is particularly demanding because optical systems have to operate under constraints that many commercial systems can avoid.

A military sensor may need to be:

And increasingly: It must operate with less human intervention.

Defense is becoming an important accelerator for advanced electro-optics because military platforms increasingly need more sensing and directed-energy capability without increasing size, weight and power.

A particularly good example is DARPA's Modular Efficient Laser Technology (MELT) program. Conventional high-energy laser systems typically rely on multiple high-power fiber amplifiers, complex optical benches and large beam-director assemblies. DARPA describes these architectures as large and difficult to manufacture and integrate. MELT is attempting to change this architecture by developing a compact, scalable "laser tile" using semiconductor fabrication, coherent beam combining, photonic integration and 3D packaging. The tiles could then be combined into larger planar arrays to create high-energy laser sources with substantially smaller footprints.

The interesting point is that DARPA is not simply trying to build a more powerful laser. It is trying to change the architecture and manufacturing approach so that high-energy lasers become smaller, more scalable and potentially easier to produce.

Current HEL architecture:  Fiber amplifiers → optical bench → beam combining → large beam director

versus

MELT concept: Laser tile → laser tile → laser tile → planar array → compact HEL source

This illustrates an important pattern in defence electro-optics: military requirements can justify investment in technologies that are initially difficult and expensive because the performance objective is unusually demanding. Once the underlying technology matures, however, the same advances in photonic integration, semiconductor lasers, packaging and beam combining can potentially spill over into other applications.

The same pattern is visible beyond laser weapons. DARPA's current RAAPTR program is developing chip-scale photonic architectures for next-generation RF systems, aiming to make photonic timing and signal routing compact and rugged enough for military platforms. DARPA describes potential applications in sensing, communications and electronic warfare.

Meanwhile, the 2026 AMPED program is pursuing a new class of high-power semiconductor lasers that combines photonic-crystal structures with advanced materials to increase power while maintaining beam quality, with an explicit goal of producing compact, efficient and manufacturable devices for defence applications.

The broader message is significant: Defence is not merely buying electro-optical technology; it is actively pushing the boundaries of what the technology can become to comply with the harsh realities dictated by the field.

7. Semiconductor Manufacturing Is Another Major Driver

Perhaps less visible to the general public is the role of optics in semiconductor manufacturing. With the latest momentum it gained from AI, semiconductor industry started to become one of the main driver for photonics technologies.

Modern semiconductor production depends heavily on optical technologies for:

As feature sizes become smaller and manufacturing processes more complex, the ability to measure extremely small deviations becomes increasingly important. The semiconductor industry is simultaneously one of the largest beneficiaries of photonics and one of its most demanding customers.

The convergence between photonics and semiconductor manufacturing is also visible at the industry level. SEMI (Global semiconductor and electronics industry association) is increasingly treating photonics as part of the semiconductor technology landscape, while recent research emphasizes the move toward integrated optical/electronic systems and photonic integration.  It is increasingly treating these technologies as part of the broader semiconductor ecosystem. In 2026, for example, SEMI's 3D & Systems Summit specifically included co-packaged optics alongside chiplets, hybrid bonding and heterogeneous integration in its semiconductor technology program. The implications extend beyond lithography itself.

The implication is profound: Photonics is no longer only a tool used to manufacture the semiconductor. It is increasingly becoming part of the semiconductor system itself.

A striking example: High-NA EUV lithography

One of the clearest examples is High-NA EUV lithography.

ASML's latest High-NA EUV platform uses an optical system with a numerical aperture of 0.55, compared with 0.33 for the previous generation. Its projection optics use large, ultra-precise multilayer mirrors to focus 13.5-nm EUV radiation and achieve a stated resolution of approximately 8 nm.

The optics are extraordinary in themselves. ASML states that the EUV mirrors require surfaces polished to a smoothness of less than one atom's thickness; each mirror also contains more than 100 carefully engineered reflective layers.

But the more important point is what happens after the pattern has been printed.

How do you know the chip was printed correctly?

This is where another class of photonic technology becomes critical: optical metrology.

ASML's YieldStar systems use optical diffraction measurements to determine whether successive layers on a wafer have been positioned correctly. The latest systems can make nanometer-level overlay and critical-dimension measurements, directly measuring structures within the chip and providing feedback to the manufacturing process.

The basic manufacturing loop therefore looks something like this:

Design=> Optical lithography => Pattern printed on wafer => Optical metrology

Measurement of overlay / dimensions => Process correction => Next wafer

This is extremely important because advanced semiconductor manufacturing is not simply about producing a very good optical image once. It is about repeatedly producing billions of structures with nanometer-level consistency. And that means optics is effectively part of the feedback-control system of the semiconductor fab.

The AI connection

The AI boom is making this relationship even more important.

SEMI estimates that global 300-mm fab equipment spending will reach about $133 billion in 2026, driven strongly by AI-related demand and advanced-node capacity expansion. SEMI expects this spending to exceed $150 billion in 2027.  This investment is important for photonics because advanced semiconductor manufacturing requires increasingly sophisticated lithography, metrology, inspection and process-control equipment. ASML itself reported that AI-driven demand is accelerating customers' capacity expansion plans and expects €43–45 billion of sales in 2026.

So there is a powerful feedback loop emerging:

AI demand

→ more advanced chips

→ more semiconductor manufacturing investment

→ greater demand for advanced lithography, metrology and inspection

→ more sophisticated optical systems

→ further advances in photonics

From using photonics to build chips to putting photonics next to the chip: This is where the story becomes particularly interesting. As AI accelerators become more powerful, moving data electrically between processors, memory and networking equipment increasingly becomes a bottleneck in terms of bandwidth, power and signal integrity.

This is driving technologies such as:

SEMI is increasingly treating these technologies as part of the broader semiconductor ecosystem. In 2026, for example, SEMI's 3D & Systems Summit specifically included co-packaged optics alongside chiplets, hybrid bonding and heterogeneous integration in its semiconductor technology program.

The implication is profound: Photonics is no longer only a tool used to manufacture the semiconductor. It is increasingly becoming part of the semiconductor system itself.

Photonics Intel assessment: The semiconductor industry may therefore be one of the most important long-term accelerators for photonics. The opportunity isn't limited to the optical systems that print or measure the chip. The next wave is about using photonics to overcome the limitations of the chip's own electronic architecture.

That creates a very interesting convergence: Photonics is helping manufacture the AI chips, measure their performance, and increasingly move more data around them. And as AI systems continue to scale, the semiconductor industry could become an increasingly powerful engine for photonics innovation.

8. Space May Become an Increasingly Important Photonics Market

Space creates a different set of challenges. Weight and power are extremely expensive. Alignment is difficult. Maintenance is essentially impossible. Radiation and temperature cycling place additional demands on components. At the same time, optical technologies offer significant advantages for:

NASA's 2026 state-of-the-art communications assessment notes that laser communication terminals can offer smaller footprints and lower power than RF systems, although their pointing requirements are significantly more demanding.

The European space market is also expanding across Earth observation, GNSS, secure communications and space situational awareness, creating additional demand for advanced optical and sensing technologies. Space therefore represents another environment in which high-performance photonics has to become smaller, more efficient and more reliable.

A striking example: SpaceX's Starlink laser network: One of the clearest demonstrations of the transition toward compact, high-performance optical systems in space is SpaceX's Starlink constellation.

Rather than relying entirely on radio links between satellites and ground stations, Starlink satellites use laser-based optical inter-satellite links to communicate with neighbouring satellites. This creates a mesh network in orbit, allowing data to travel between satellites before reaching a ground station. NASA's documentation of SpaceX's commercial optical-relay program describes the architecture as a Starlink constellation connected by optical inter-satellite links, with the company developing compact laser terminals for spacecraft integration. The significance is not simply that SpaceX uses lasers instead of radio.  The laser terminal has to work on a satellite

A space optical communication terminal has to combine several difficult functions into a relatively compact package:

Laser source: generates the optical carrier

Beam steering / pointing:  directs the extremely narrow beam toward another spacecraft

Optics: Expands, focuses and receives the optical signal

Detector: Converts the incoming light back into electrical data

Control electronics: Continuously maintain the optical link

And all of this has to operate while the satellite is moving at roughly orbital velocity and the relative geometry between spacecraft is constantly changing.  This makes pointing and tracking one of the fundamental engineering challenges. NASA notes that the advantage of optical communications comes with significantly tighter pointing requirements than RF systems.

Why use lasers? The payoff is substantial. Infrared optical communications can carry more data in a single link than conventional RF systems, while optical terminals can be smaller, lighter and more power-efficient. NASA specifically highlights these advantages for spacecraft laser communications.

For Starlink, there is another major advantage: The satellites don't always have to communicate directly with a ground station. By connecting satellites directly with laser inter-satellite links, Starlink can route traffic through space rather than always forcing it through the nearest available ground gateway. SpaceX reported that improvements to its laser mesh produced a 30–40 ms reduction in latency in Asia and Africa, demonstrating that the optical network is becoming a significant part of the constellation's routing architecture. The latency advantage of Starlink's optical mesh does not come from light traveling faster than radio waves. It comes from geometry.

NASA reported in July 2026 that SpaceX had more than 25,000 laser terminals operating in orbit across the Starlink constellation. NASA also stated that the same Starlink mini-laser technology is being adapted for spacecraft missions such as Artemis III.

That changes the story considerably. We're no longer talking about: "Can optical communication work in space?" We're talking about: "Can thousands of compact optical communication terminals be manufactured, launched and operated as part of a commercial space network?" SpaceX appears to have demonstrated that this is possible at constellation scale.

The Starlink example illustrates why miniaturization matters in space and how contributes to our daily internet experience directly. Once an optical communication terminal becomes sufficiently compact, power-efficient and manufacturable, it stops being a specialized scientific instrument and becomes a standard subsystem that can be installed on thousands of spacecraft.

This is one of the striking examples of the new transition for photonics. The value of the technology is no longer determined only by the performance of an individual optical terminal, but by whether the terminal can be made small enough, efficient enough, reliable enough and inexpensive enough to deploy at scale with the correct architecture.

And this is exactly consistent with NASA's current commercial-space communications strategy. In 2026, NASA awarded SpaceX $69.95 million to demonstrate high-rate science-data relay services using the Starlink optical network, with the goal of eventually allowing spacecraft in low Earth orbit to use the commercial optical network for high-rate communications.

What Is Still Limiting the Growth?

The outlook for photonics is extremely strong. But there is a fundamental difference between demonstrating an optical technology and turning it into a product that can be manufactured millions of times, operate reliably for years, and compete economically with an electronic alternative. This distinction may become one of the defining challenges for photonics over the next decade.

The industry is moving from relatively discrete optical systems toward highly integrated combinations of lasers, detectors, waveguides, electronics, mechanics, thermal management and software. As that happens, the limiting factor increasingly shifts from the performance of an individual optical component to the performance of the entire system. Recent work on heterogeneous integration makes this point explicitly: successful systems require co-design across photonics, electronics, packaging, thermal management, manufacturing and reliability.

In other words:

The next generation of photonics will be won at the system level, not necessarily at the component level. There are several reasons why.

9.1 Cost: The technology has to become economically useful: A laboratory can demonstrate an optical phased array with excellent beam steering. A research group can demonstrate a very sensitive detector. A startup can build an extremely efficient laser.

But commercial systems have additional requirements:

This becomes particularly important as photonics moves toward applications requiring thousands or millions of units. The same issue appears in autonomous systems. A LiDAR system can have excellent performance, but if the cost of the complete sensor remains too high, its adoption may be limited even when the underlying technology works.

This is why I think one of the most important metrics for emerging photonics should be: Performance per euro — not simply performance.

9.2 Manufacturing: Can the laboratory become a factory?

This may be the biggest hidden challenge. Many advanced photonic technologies involve structures that are extremely sensitive to:

A research prototype can be individually optimized. A production line cannot do that for every device. The industry therefore has to move from:

"Can we make one?"

to:

"Can we make 100,000 identical devices?" That requires semiconductor-style manufacturing discipline. Recent work on advanced nanophotonic manufacturing identifies fabrication variability, optical loss accumulation, heterogeneous material stacking and manufacturability as major bottlenecks as photonic systems become more complex. This is particularly important for integrated photonics.

A tiny variation in a waveguide or resonator may be negligible in a simple device but become significant when hundreds or thousands of photonic elements are connected together. The manufacturing challenge therefore becomes:

Precision + repeatability + yield + throughput

And those four requirements do not automatically come together.

9.3 Packaging: The part nobody sees. Packaging is often treated as an afterthought.

For advanced photonics, it isn't. A practical photonic system may need to bring together:

Laser => Photonic chip => Detector => Electronic processor => Fiber / free-space optical interface => Thermal system => Mechanical structure

These pieces may be manufactured using completely different processes and materials.

Getting them to work together requires extremely precise:

This becomes particularly difficult when lasers and photonic integrated circuits are combined with high-speed electronics. Semiconductor Engineering notes that photonics manufacturing increasingly depends on controlling optical loss, material interfaces, waveguide dimensions, coupling, laser integration, fiber alignment, packaging and thermal effects as photonics moves toward higher-volume semiconductor production.  The irony is that the photonic chip itself can become incredibly small while the packaging required to make it useful remains surprisingly difficult.

9.4 Thermal management: Photonics does not escape the laws of thermodynamics

Miniaturization creates another problem. Putting more functionality into a smaller volume increases power density. This is particularly important when photonics is integrated with electronics.

Consider a future package containing:

Each subsystem has different temperature requirements. Lasers may have strong temperature dependence. Detectors can be temperature sensitive. Electronic processors generate substantial heat. Optical components can experience changes in refractive index or alignment as temperature changes.

The 2026 literature on heterogeneous 3D integration highlights exactly these problems: non-uniform power densities, hotspots, thermal-expansion mismatch and limited heat-removal paths can degrade both performance and reliability so critically that the unit become unfeasible to realize. And this becomes a particularly interesting problem for AI hardware.

The industry wants to bring optics closer to the processor because electrical interconnects become increasingly costly in bandwidth and energy. But bringing photonics closer also puts the photonic components in the same thermal environment as extremely high-power electronics. So we get a fundamental engineering trade-off:

The closer photonics gets to the processor, the better the interconnect can become — but the harder thermal management becomes. That is one of the reasons co-packaged optics is technically fascinating.

9.5 Calibration: Integration makes everything more sensitive. Traditional optical instruments can often be individually aligned and calibrated. Highly integrated systems have less freedom. Imagine a photonic system with hundreds of optical channels.

A small:

can alter the system response. This means that calibration can become a substantial part of the product. And calibration itself has consequences:

More calibration → more manufacturing time → higher cost → lower throughput

The obvious solution is to make the system self-calibrating.

That requires:

This is another reason why advanced photonics increasingly resembles electronics:

The optical system is becoming a controlled system rather than a collection of passive components.

9.6 Power: Smaller does not automatically mean lower power Miniaturization is often presented as:

smaller → better

But that is not always true. A smaller optical system may require:

The optical hardware may become dramatically smaller while the total system power does not decrease proportionally.

This matters enormously in:

For these platforms, the real metric is not simply: How small is the sensor? It is: How much useful information can the sensor produce per watt and per kilogram? That is a much more demanding metric.

9.7 Data: More sensing can create a bigger problem. Another paradox is emerging. Better sensors generate more data. A high-resolution camera produces more pixels. A LiDAR generates 3D point clouds. An event camera can produce extremely high event rates. A hyperspectral sensor produces information across many wavelengths. A multi-modal sensing system may combine all of them. But the machine then has to:

capture → process → store → transmit → interpret

that information. This is particularly important for autonomous systems. The goal isn't actually to maximize the amount of data. The goal is to maximize useful information per unit of power, bandwidth and computing. This is why sensing and computing are becoming increasingly intertwined.

The sensor may increasingly need to perform some preprocessing locally, extracting features before sending the information to a central processor. That creates another convergence:

Photonics + electronics + AI rather than simply: optical sensor → electronic computer

9.8 Reliability: The hardest test is time. A laboratory demonstration answers: Does it work? A product has to answer: Does it continue working? For years. This becomes especially difficult when heterogeneous materials are combined.

Different materials expand differently as temperature changes. Repeated thermal cycling can generate stress at interfaces and contribute to delamination, interconnect failure or other degradation mechanisms. The 2026 research literature on heterogeneous integration identifies thermal-expansion mismatch and repeated thermal cycling as significant reliability concerns.

For space and defence systems the requirements become even more demanding because of:

The electronics industry has decades of experience building reliability into silicon manufacturing. Photonics is increasingly having to develop the equivalent reliability engineering culture for integrated optical systems.

The IEEE's 2026 reliability symposium for electronics and photonics packaging, for example, explicitly focuses on failure mechanisms, accelerated testing, reliability modelling and lifetime assessment for photonic and heterogeneous packages.

9.9 The deeper problem: Everything is now connected. This is perhaps the most important point. In an old optical system, you could often optimize the: lens, then the detector, then the electronics, and integrate them afterward. Increasingly, that approach doesn't work at all in the current interconnected scheme.

Suppose you improve the laser.

That changes: power → temperature → packaging → detector performance → calibration

Suppose you reduce the optical size.

That changes:

aperture → sensitivity → signal-to-noise → algorithms

Suppose you move the photonics package next to the processor.

That changes:

bandwidth → thermal management → mechanical design → reliability

Everything starts interacting with each other so intricately that optimizing only one part of the problem does not necessarily yield the most optimum system.  A wholistic approach is becoming increasingly critical and that is also why heterogeneous integration research increasingly emphasizes co-design, bringing together systems architecture, semiconductor technology, packaging, photonics, thermal engineering and reliability rather than optimizing each subsystem independently.

The shift from component engineering to systems engineering: This leads to what I think is the most important conclusion of this entire section.

The historical photonics question was often: Can we build a better optical component?

The next question is: Can we build an optical system that can be manufactured, packaged, cooled, calibrated, powered and operated reliably at scale?

That is a much harder problem. And it fundamentally changes the competitive landscape. The companies that win may not necessarily have the best individual laser, best detector or best optical design. They may be the companies capable of successfully integrating all of them and optimizing the system-level performance. That is why I would argue that systems engineering may become one of the most important competitive capabilities in photonics over the next decade.

Photonics Intel Assessment

The biggest risk for photonics is not that the underlying physics will stop working. The bigger risk is that integration becomes too expensive, too difficult or too unreliable to scale. Conversely, the biggest breakthroughs may not come from discovering a new optical effect.

They may come from solving seemingly mundane but increasingly challenging engineering problems driven by market trends explained above:

In other words:

The future of photonics may depend as much on manufacturing engineering as on optical physics.

10. Who Is Best Positioned?

I don't think the next decade will necessarily belong to the company with the best individual laser, detector or lens. The strongest position may belong to companies that can combine multiple technologies into an effective system.

The more important competitive advantage may belong to companies that can combine technologies across several layers of the stack:

Photonics → optics → semiconductors → computation → AI → complete systems.

This is already happening. AI infrastructure is pushing optical interconnects toward co-packaged optics; semiconductor manufacturers are integrating photonics into advanced packages; and defence companies are combining EO/IR sensors, navigation, AI and data fusion into complete sensing systems.

Photonics companies: lasers → detectors → integrated photonics →Fiber optics

Optical companies: Lenses → mirrors → beam steering → imaging systems

Semiconductor companies:  Detectors → processors → photonic integration

AI companies: Perception → classification → sensor fusion → Interpretation

System companies:  Complete sensing platforms

Five companies illustrate this transition particularly well.

1. NVIDIA: Controlling the system architecture

Strategic position: AI + computing + networking + photonics

NVIDIA is perhaps the most interesting company in this list because it didn't start as a photonics company. Its advantage is that it controls the AI computing architecture into which photonics is increasingly being incorporated. NVIDIA is now integrating silicon photonics directly into its networking platforms through co-packaged optics (CPO). Its current Quantum-X and Spectrum-X photonic switches put the optical components directly alongside the switching ASIC rather than relying exclusively on conventional pluggable transceivers. NVIDIA says its CPO architecture can provide substantially better power efficiency and lower latency, and it is targeting AI factories with millions of GPUs.

This is a crucial distinction. NVIDIA isn't simply asking: How can we build a better optical transceiver?

It is asking: How should photonics be incorporated into the architecture of an AI factory?

That gives NVIDIA enormous strategic leverage. And NVIDIA is going deeper into the supply chain. In March 2026, it announced a $2 billion investment in Coherent and a multiyear agreement covering advanced lasers and optical networking products.

Why I would watch NVIDIA. If the industry moves from: electronic processor + external optical transceiver toward: processor + photonic engine + advanced package then the company designing the computing architecture has enormous influence over which photonic technologies become standard.

Photonics Intel view:  Highest system-level leverage.

2. Coherent : Controlling a large part of the photonic stack.

Strategic position: materials → lasers → photonic devices → optical engines → systems

Coherent is almost the opposite of NVIDIA. Where NVIDIA comes from computing, Coherent comes from photonics and has built an unusually broad technology portfolio. Its capabilities span technologies including: high-power lasers, InP lasers, VCSELs, silicon photonics, modulators, optical components, photonic integrated circuits, optical transceivers, advanced packaging.

At OFC 2026, Coherent demonstrated several different CPO architectures, including silicon photonics, high-power InP lasers, VCSELs and InP-on-silicon modulators.  That's important because we don't yet know which photonic architecture will dominate.

There could be several winners:

Silicon photonics, InP, VCSEL, External laser source, CPO, Near-packaged optics

A company capable of supplying multiple architectures has an advantage because it doesn't need to bet everything on one technical approach. And NVIDIA's $2 billion investment in Coherent in 2026 is a strong indication of the strategic importance of that capability.

Why I would watch Coherent: Coherent is positioned relatively close to the physical building blocks of photonics, while also moving upward toward integrated systems.

Photonics Intel view: One of the strongest vertically integrated photonics positions.

3. Lumentum: The laser bottleneck.

Strategic position: lasers + photonic components + optical infrastructure

Lumentum occupies a particularly interesting position because many of the most advanced optical architectures still depend on a small number of extremely important components—especially lasers.

The company supplies technologies including: InP lasers, EMLs, VCSELs, optical components, external laser sources, optical circuit switches, optical networking technologies.

Lumentum's own 2026 assessment of the AI-optics market highlights the convergence of 1.6T optics, 200G/lane technologies, CPO, external laser sources, VCSELs and optical circuit switching.  This is important because the industry is discovering that the bottleneck isn't necessarily the optical concept—it can be the ability to manufacture enough of the critical optical components.

And the strategic importance is reflected in NVIDIA's decision to invest $2 billion in Lumentum and secure future supply of advanced laser technology.  

Why I would watch Lumentum: If optical interconnects become fundamental infrastructure for AI, then high-performance lasers become something closer to semiconductor infrastructure than a niche photonics component.

Photonics Intel view: Critical component/supply-chain leverage.

4. TSMC : Making photonics part of the semiconductor package. Strategic position: semiconductor manufacturing + advanced packaging + photonics TSMC represents perhaps the most important part of our original argument: Photonics and electronics are becoming one system.

TSMC's Compact Universal Photonic Engine (COUPE) is designed to integrate photonic and electronic dies using its advanced packaging technologies. In its 2026 technology roadmap, TSMC said COUPE would reach a major milestone with true co-packaged optics beginning production in 2026. The company says integrating the optical engine directly into the package can provide 2× power efficiency and 10× lower latency versus a pluggable implementation.

This is potentially a very important transition.

Historically: Optics helped manufacture semiconductor chips.

Today: Optics manufactures, measures and inspects the chips.

The emerging architecture is: Photonics becomes part of the semiconductor package itself.

TSMC therefore sits at one of the most important interfaces in the ecosystem:

Photonics ↔ semiconductor manufacturing ↔ advanced packaging ↔ AI computing

And that interface could become extremely valuable as AI systems continue demanding greater bandwidth and lower energy consumption.

Photonics Intel view: Possibly the most important manufacturing/integration position.

5. Safran: What the integrated EO system looks like.

Strategic position: optics + EO/IR + lasers + navigation + AI + complete defense systems

Safran is important for a different reason. The previous four companies illustrate the photonics/computing ecosystem. Safran illustrates what happens when those technologies are brought together into an operational sensing system.

Safran's defense portfolio spans: electro-optical sensors, thermal imaging, laser systems, stabilized gimbals, navigation, inertial systems ,AI, data fusion, complete ISR systems

A particularly interesting 2026 example is Land OmniGuard, which combines Safran's optronics, resilient navigation and embedded AI. Its architecture integrates multiple existing technologies into a passive detection-and-tracking chain.

Safran is also moving aggressively into small-UAV EO/IR systems. In June 2026, Safran and Theon announced a joint venture to develop small EO/IR gimbals below 8 kg for ISR and targeting applications, combining Safran's stabilized-gimbal and EO-system expertise with Theon's sensor technology.

This is exactly the kind of integration you are describing.

It isn't: camera + laser + thermal sensor

as separate products. It becomes: sensor + laser + navigation + stabilization + processing + AI → operational capability

Why I would watch Safran: The company illustrates the system-level endpoint of the photonics ecosystem.

Photonics Intel view: Strong example of integrated EO/IR + AI + navigation becoming one system.

Photonics Intel Assessment:

What we believe is real: The expansion of electro-optics into autonomous systems, robotics, defence, space and semiconductor manufacturing is not simply a short-term technology cycle. The underlying driver is fundamental: Machines increasingly need to perceive and interpret the physical world. Photonics is exceptionally well suited to providing that perception.

What may be overhyped. Not every new photonic technology will become commercially important. A smaller sensor is not automatically a better sensor. A solid-state architecture is not automatically superior to a mechanical one. An AI-enabled camera does not automatically produce better information. And a laboratory demonstration does not automatically translate into a manufacturable product.

The real winners will be determined by system-level performance, cost, reliability and manufacturability.c

What we would watch

Over the next several years, I would pay particular attention to:

1. Photonic integration

2. Sensor fusion

3. Miniaturization

4. Edge processing

5. Manufacturing

Key Takeaway: The most important development in photonics may not be a single breakthrough. It is the gradual transformation of optics from a specialized component into a core information technology. Not simply as: Technology that helps machines see. But increasingly as: Technology that helps machines understand the world.