The “Copper Wall.”
That is the invisible barrier the entire tech industry is about to crash into.
For fifty years, we’ve relied on a simple bargain: we shrink transistors, pack more of them onto a silicon chip, and get faster computers. This is the heart of Moore’s Law. But we’ve hit a snag that no amount of engineering genius can solve with traditional methods.
Heat and resistance.
Modern data centers—the “brains” powering the AI revolution—run on electricity. Electrons move through copper wires to transfer data. But when you try to push billions of calculations per second through microscopic copper threads, you get friction. Friction creates heat. Resistance slows things down.
Right now, AI supercomputers are spending nearly as much energy moving data as they are processing it. We are literally burning billions of dollars just to push electrons through a wire. It’s like trying to drain a swimming pool through a drinking straw.
Enter photonics: replacing electricity with light.
Instead of sending electrons through a copper wire, photonics technology uses lasers to shoot photons (light particles) through microscopic cables.
Why does this matter?
- Speed: Light is the fastest thing in the universe.
- Cooler Operations: Light creates virtually no heat as it travels. No resistance means no wasted energy.
- Bandwidth: You can send thousands of times more data over a single beam of light than you can over a copper wire.
This might sound like science fiction, but the backbone of the internet (fiber optics) has used light for decades to connect cities. But now, we are shrinking that technology down to the size of a microchip. We are moving from connecting continents with light to connecting computer chips with light.
The “Nervous System” of AI
The key thesis behind photonics stocks: The first AI wave was the race to build the best “brain” (the GPU). But the bottleneck isn’t the brain anymore; it’s the nervous system.
If AI is the gold rush, photonics is the railroad. The companies that manufacture the lasers, the optical sensors, and the “digital plumbing” that allows chips to talk to each other at light speed are sitting on a swelling utility-like demand curve.
Optical Components & Transceivers
This segment covers the “blue-collar” machinery of the photonics revolution. If the GPU is the brain, these are the synapses.
Every time you hear about a “cluster” of 100,000 GPUs, you are really hearing about a cabling nightmare. To connect those chips, data centers use transceivers—hot-swappable modules that plug into servers to convert electrical signals (from the chip) into optical signals (for the fiber cable) and back again.
We are currently undergoing an upgrade supercycle. The industry is ripping out old 400G modules and desperately trying to secure supply of 800G and 1.6 Terabit (1.6T) modules to keep pace with Nvidia’s Blackwell and Rubin architectures. This supply squeeze is where these photonics stocks live.
Lumentum Holdings (NASDAQ: LITE)
Lumentum is not trying to win the war for the best transceiver; they are selling the gunpowder to everyone fighting it. Their moat is their mastery of the EML (Electro-absorption Modulated Laser).
Here’s the key dynamic: High-speed optical modules (800G and above) generally cannot use cheaper, lower-quality lasers. They require EMLs to maintain signal integrity at blistering speeds. Lumentum is the heavyweight champion of manufacturing these chips at scale.
Even if a competitor wins a contract to build the transceiver module for a hyperscaler like Google or Meta, there is a high probability they still have to buy the laser chip inside that module from Lumentum.
The 2026 Catalyst: The “DeepSeek Moment” Fallout. In the first half of 2025, the industry pivoted aggressively from “scale-up” (bigger individual chips) to “scale-out” (massive clusters of smaller chips). This shift plays directly into Lumentum’s hand because it requires exponentially more laser interconnects. They are currently the only supplier shipping 200G-per-lane EMLs at volume, which are the critical component for the next-gen 1.6T pluggables that Nvidia’s NVL72 racks demand.
Coherent Corp. (NYSE: COHR)
If Lumentum is the specialized laser engineer, Coherent is the industrial factory. With roughly 25% market share, they are the volume leader. When a hyperscaler like Microsoft needs to deploy 500,000 units of a specific transceiver yesterday, Coherent is often the first phone call because they have the raw manufacturing capacity to deliver.
Coherent is a vertically integrated beast. They make the wafers, the lasers, and the finished transceiver modules. This gives them significant control over their margins and supply chain security.
The 2026 Catalyst: 1.6T Dominance. Coherent has aggressively positioned itself as the first-mover for 1.6T transceivers. While the market is currently digesting 800G, AI frontier models will begin to require 1.6T speeds to minimize training time. Coherent’s ability to ramp this specific product line will be the primary driver of their stock performance over the next 12 months.
Applied Optoelectronics (NASDAQ: AAOI)
Applied Optoelectronics is the high-risk, high-reward turnaround play. AAOI is significantly smaller than LITE or COHR, but they have a unique leverage point: Vertical Integration + Domestic Production.
Unlike many competitors who outsource assembly to Southeast Asia, AAOI manufactures its own lasers and assembles modules in Sugar Land, Texas (and Taiwan). In a geopolitical climate where “supply chain sovereignty” is becoming a buzzword for US Big Tech, having a domestic supply of 400G and 800G transceivers is a strategic asset.
The 2026 Catalyst: The “Mystery” Hyperscaler. Throughout late 2024 and 2025, AAOI has been ramping up production for a big contract with a “large hyperscaler” (widely believed to be Microsoft). The thesis here is a turnaround story. AAOI has historically struggled with execution, but if they successfully fulfill this ramp without yield issues, they prove they can play in the big leagues. The stock is a play on execution, not just sector growth.
Silicon Photonics & Co-Packaged Optics
This is the “end game” of the thesis.
If the previous photonics stocks targeted the “plumbing” (cables and transceivers), these are about the “organs.” Silicon photonics is the alchemical process of printing lasers and optical circuits directly onto silicon wafers, just like we print transistors.
The ultimate goal here is Co-Packaged Optics (CPO). Instead of having a switch chip here and an optical module over there (connected by a copper trace that burns power), CPO glues the optics directly onto the same package as the chip. It eliminates the copper entirely. It’s the only known way to get past the 100 Terabit-per-second speed barrier without melting the data center.
There are several photonics stocks fighting for this real estate.
Broadcom (NASDAQ: AVGO)
Broadcom is not asking the market to adopt CPO; they are forcing it. As the undisputed ruler of switch silicon (their Tomahawk chips run nearly every major hyperscale network), Broadcom has the unique power to integrate photonics whether the industry is ready or not.
Their “Bailly” 51.2T CPO switch is the spear tip. They have already logged over 1 million flap-free hours at Meta (Facebook), effectively killing the “it’s too risky” argument.
The 2026 Catalyst: The “OpenAI” Custom Chip. While the market focuses on their switches, the real 2026 firework is the custom AI accelerator they are designing for OpenAI. Rumors confirm this chip will ship in 2026. If Broadcom integrates their silicon photonics IP directly into this custom compute chip (not just the switch), they effectively lock in the most valuable AI real estate in the world. You own Broadcom because they are the only ones who can act as a “one-stop shop” for the entire AI transport layer.
Marvell Technology (NASDAQ: MRVL)
If Broadcom is the “Apple” of this world (closed ecosystem, proprietary), Marvell is the “Android.” They are the preferred partner for everyone who hates being locked into Broadcom.
Marvell’s dominance lies in the DSP (Digital Signal Processor)—the brain that translates digital data into analog light signals. As speeds hit 1.6 Terabits, the physics gets messy, and you need a world-class DSP to clean up the signal. Marvell’s “Nova” 1.6T DSPs are the gold standard.
The 2026 Catalyst: The “3D” Silicon Photonics Engine. Marvell is currently ramping its “3D Silicon Photonics” engine, which stacks components vertically to save space. As Amazon and Google look to build their own custom silicon to rival Nvidia, they are turning to Marvell to provide the connectivity IP. Marvell’s connectivity revenue is growing faster than its compute revenue—a signal that they could win the “nervous system” war even if they don’t win the “brain” war.
MACOM Technology Solutions (NASDAQ: MTSI)
While Broadcom and Marvell fight over the digital chips (DSPs), MACOM owns the difficult analog physics. They make the actual lasers and drivers that create the light.
MACOM is the champion of a rogue movement called LPO (Linear Drive Pluggable Optics). The idea is bold: “DSPs consume too much power. Let’s get rid of them.” MACOM produces the high-performance analog chips that allow data centers to rip out the power-hungry DSPs entirely for short distances.
The 2026 Catalyst: The “DSP-Free” Revolution. In 2026, as AI clusters get denser, power becomes the only metric that matters. Hyperscalers are desperate to cut 5-10 watts per module. MACOM’s linear drive technology allows them to do exactly that. If Microsoft or Meta announce a large-scale “Linear Drive” deployment, MTSI stock will re-rate overnight. They are the only ones with the analog pedigree to make this work at scale.
Credo Technology Group (NASDAQ: CRDO)
Credo is the smallest name in this segment, but they have a cult following among network engineers. They built their reputation on AECs (Active Electrical Cables)—smart copper cables that fixed the reliability issues inside the rack.
But now, they are crossing the Rubicon into photonics. They realized that reliability is the #1 pain point for AI training (if one link fails, the whole training run crashes). They are bringing their “ZeroFlap” reliability ethos to the optical market.
The 2026 Catalyst: The “BlueBird” Optical DSP. Credo has launched “BlueBird,” their entry into the optical DSP market to challenge Marvell. The thesis is simple: Hyperscalers love Credo’s AECs because they just work. If Credo can transfer that trust to their optical chips, they can steal market share from Marvell. Watch for Credo to announce a “Tier 1” hyperscaler win for BlueBird in 2026—that is the validation event the market is waiting for.
Photonics Foundries & OSAT
This is where the rubber meets the road. You can have the most brilliant chip design in the world (Broadcom/Nvidia), but someone has to physically manufacture the wafer and then assemble that fragile glass-and-silicon sandwich into a working module.
These photonics stocks span the Foundries (who print the chips) and the OSATs (Outsourced Semiconductor Assembly and Test – who package them). This is often the most overlooked part of the supply chain because it isn’t “sexy,” but it’s the choke point. If these factories stumble, the photonics upgrade cycle pauses.
Fabrinet (NYSE: FN)
Fabrinet is not a traditional “contract manufacturer” like Foxconn; they are a high-precision optical specialist. Nvidia relies heavily on Fabrinet to assemble its complex optical interconnects. As the industry moves toward “scale-out” architectures (massive clusters of GPUs connected by light), the complexity of packaging these optical modules skyrockets.
You cannot just solder these things; you have to align microscopic lasers with glass fibers to within a fraction of a micron. Fabrinet is one of the few companies on earth with the yield rates to do this profitably at mass scale.
The 2026 Catalyst: The “CPO” Assembly Ramp. While everyone talks about the chips in Co-Packaged Optics, the assembly is the nightmare. Fabrinet has positioned itself as the primary OSAT partner for the first wave of CPO deployments. The market treats them as a proxy for transceiver demand, but the real re-rating event will come when they confirm volume production for Nvidia’s next-gen optical switch systems in mid-2026.
GlobalFoundries (NASDAQ: GFS)
GlobalFoundries has made a massive bet that it can corner the market on manufacturing silicon photonics chips. While TSMC is busy with the ultra-complex 3nm logic chips (the “brains”), GlobalFoundries is aggressively targeting the “feature-rich” nodes where photonics live.
Their “GF Fotonix” platform is effectively the industry standard for companies that don’t own their own fabs. In a massive strategic move just days ago (November 2025), GlobalFoundries acquired Singapore’s Advanced Micro Foundry (AMF). This is not a bolt-on; it is a dominance play. AMF was a hidden gem in the industry, and by buying them, GlobalFoundries has effectively bought the capacity to be the largest pure-play silicon photonics foundry in the world.
The 2026 Catalyst: The “Qualcomm” Effect. Watch for the expansion of their partnership with major fabless chipmakers like Qualcomm or Broadcom who need a reliable US-friendly foundry. The AMF acquisition allows GFS to offer a “China-free” supply chain for optical chips—a selling point that will become mandatory for US Hyperscalers in 2026.
Tower Semiconductor (NASDAQ: TSEM)
Tower Semiconductor is the artisan to GlobalFoundries’ factory. They specialize in Silicon Germanium (SiGe), a material that is absolutely critical for high-speed analog circuits.
Here is the physics problem: As we move to 1.6 Terabits, pure silicon struggles to switch lasers on and off fast enough. Silicon Germanium handles these speeds effortlessly. Tower has a unique “open foundry” model where they let customers mix and match Silicon Photonics with high-speed SiGe on the same platform.
The 2026 Catalyst: The “Nov 2025” CPO Breakthrough. Tower just announced a major breakthrough in their CPO manufacturing process that allows for “wafer-scale” bonding of lasers. This sounds technical, but it means they can now stack the laser directly on top of the control chip at the wafer level, drastically reducing the cost of 1.6T components. As the market digests this news over the next two quarters, watch for Tower to announce a “Tier 1” design win that validates this technology for mass deployment in AI clusters.
Photonics Disruptors & Moonshots
This is the section for the “venture capital” portion of your portfolio. If the previous names were the industrial giants and the steady toolmakers, these are the laboratory bets. These companies are not trying to optimize the current standard; they are trying to rewrite the laws of physics governing the sector.
If their technology fails to scale, they risk going to zero. But if they succeed—if their material science becomes the new industry standard—these are the photonics stocks with 10x or 20x potential. They take aim at the “unsolvable” problems of heat and size that silicon alone cannot fix.
POET Technologies (NASDAQ: POET)
For decades, the photonics industry had a fatal flaw: it was too bespoke. Connecting a laser to a chip required expensive, active alignment—essentially paying a PhD to glue things together under a microscope. It was unscalable.
POET’s “Optical Interposer” is the fix. Think of it as a universal motherboard for light. It allows companies to “flip-chip” lasers and sensors onto a silicon wafer passively, without that expensive alignment. It turns a boutique artisan process into a high-speed assembly line.
The 2026 Catalyst: The “1.6T” Volume Ramp. After years of “pilot programs,” the dam finally broke in late 2025. POET secured a critical production order for their 1.6T optical engines—the specific engines needed for the next generation of AI clusters. With a fresh $75M war chest and a partnership with NTT to disrupt mobile networks, POET is graduating from science project to factory. The bet here is simple: supply shortages for conventional 1.6T modules will force hyperscalers to adopt POET’s scalable engine to keep their AI roadmap on track.
Lightwave Logic (NASDAQ: LWLG)
Silicon is hitting a wall. You can only switch a silicon modulator on and off so fast before it degrades. Lightwave Logic says the answer isn’t better silicon; it’s Polymers.
They have synthesized a proprietary organic polymer (“Perkinamine”) that switches faster than silicon and uses a fraction of the power. For years, the knock on this stock was reliability—”polymers melt, silicon lasts forever.” But in mid-2025, they passed the holy grail of reliability testing (Telcordia), proving their organic material can survive the brutal heat of a data center for years.
The 2026 Catalyst: The “Tier 1” Licensing Deal. Lightwave Logic is a licensing play, not a manufacturer. They aim to be the “ARM Holdings” of photonics. The market is currently pricing in a massive licensing deal with a major foundry (likely one of the giants mentioned in the previous section). The moment a “Tier 1” partner confirms they are putting Perkinamine into their standard library, this stock graduates from a speculative chemistry play to a royalty-generating machine.
Aeluma, Inc. (NASDAQ: ALMU)
Aeluma is doing something that shouldn’t be possible: they are growing high-performance compound semiconductors (the exotic stuff usually found in military lasers) directly on top of cheap, massive 12-inch silicon wafers.
This is “Heterogeneous Integration.” Usually, you have to choose between high performance (Indium Phosphide) or high scale (Silicon). Aeluma merges them. This makes them uniquely positioned not just for AI, but for the automotive LiDAR market, where you need high-powered lasers that are cheap enough to put in a family sedan.
The 2026 Catalyst: The “AIM Photonics” Validation. Aeluma has embedded itself deeply into the US defense and research ecosystem, specifically through its membership with AIM Photonics. The catalyst to watch in 2026 is the commercial spinoff of this defense work. As they move from “government grants” to “commercial LiDAR contracts,” the valuation math changes entirely. Watch for a joint development agreement with an automotive Tier 1 supplier—that is the signal that their “hybrid” chips are ready for the road.
Breaking Through the Copper Wall
The “Copper Wall” is a physical reality that data center operators already pay billions of dollars to mitigate. The transition to photonics is the only known path to increasing bandwidth density while managing the thermal constraints of AI infrastructure.
This watchlist provides a tiered approach to this transition. Optical Components (Lumentum, Coherent, AOI) offer immediate exposure to the current capex cycle. Silicon Photonics (Broadcom, Marvell, MACOM, Credo) captures the architectural battle for the interface. Foundries (Fabrinet, GlobalFoundries, Tower) offer an industrial floor to the thesis. Finally, the Disruptors (POET, Lightwave, Aeluma) provide high-beta exposure to the material science breakthroughs that could define the post-2026 landscape.