Gaetano

Gaetano

@Gaetano2026

Oct 31, 2025, 20:13 UTC

Instead of betting on a single winner in the AI hardware race, I've constructed a diversified portfolio to capture value across the entire ecosystem. Here's the breakdown of my thesis on $MRVL, $ALMU, $POET, $LITE, and $COHR. The AI revolution is a massive challenge in physics and engineering. AI models require so much data that the copper wires in data centers are hitting a wall. The future of AI will be built on light. The companies that master the hardware for this light-based communication, known as photonics, are positioned for incredible growth. My investment thesis is built around a carefully chosen basket of five photonics and semiconductor companies: Marvell $MRVL, Lumentum $LITE, Coherent $COHR, POET Technologies $POET, and Aeluma $ALMU. This is a strategic portfolio designed to capture value across the entire industry. It balances established market leaders with fast-growing innovators. This portfolio gives exposure to the immediate AI build-out, the networking brains that manage it, the coming changes in manufacturing, and the new technologies that will unlock entirely new markets. The Anchors: $LITE & $COHR – Dominating the Present Lumentum and Coherent are the established giants of the photonics world. They are the most direct beneficiaries of the billions of dollars being spent right now by cloud companies to build AI data centers. $LITE: Lumentum is a powerhouse in high-performance optical parts. Its key strength is its advanced laser technology, which is critical for high-speed data transmission. Lumentum has secured a vital position in the AI ecosystem through its deep partnership with Nvidia. It is a primary supplier of the ultra-high-power lasers needed for Nvidia's next-generation systems. These systems package optics and chips together for maximum efficiency. This relationship makes Lumentum a core part of the supply chain for new AI hardware. The company is even expanding its US manufacturing to meet this specific demand. $COHR: Coherent has a different strategy. It controls its entire supply chain. It operates on a "materials-to-systems" model, meaning it makes everything from the basic engineered materials to the finished laser systems. This gives the company huge resilience against supply chain problems, a major advantage today. Like Lumentum, Coherent is a key partner for Nvidia and has strong relationships with other tech giants like Apple. Its business is also more diverse, serving industrial and electronics markets, which provides a more balanced risk profile. Holding both $LITE and $COHR provides stable exposure to the current AI boom through two different but complementary strategies. The Networking Brains: $MRVL – The System-Level Leader If Lumentum and Coherent make the individual high-performance parts, Marvell designs the super-fast "brains" that manage and direct the flow of data. Marvell is a leader in custom silicon and networking hardware that is essential for modern AI data centers. Custom Silicon and DSPs: Marvell works directly with the world's largest cloud companies to design custom chips optimized for their specific AI needs. A key product is the PAM4 DSP. Think of a DSP as a critical, high-speed translator that converts electrical signals from a computer chip into light signals for fiber optic cables, and back again. Marvell's DSPs are the industry standard for enabling the 800G and 1.6T speeds required by AI. High-Speed Switches: Marvell's Teralynx switches are another core product. These are the traffic cops of the data center, directing immense amounts of data between thousands of AI processors with incredibly low delay. This low latency is crucial for making massive AI clusters operate as one giant computer. Advanced Optics: Through its acquisition of Inphi, Marvell became a leader in silicon photonics. This technology allows them to build optical components directly onto silicon wafers, just like computer chips. This leads to a strategy called Co-Packaged Optics (CPO), where the optical connections are placed right next to the main AI processor. This dramatically saves power and increases speed, a key goal for future AI systems. Marvell provides the intelligence and orchestration layer for the AI data center. It is a crucial link between the raw components and the complete, functioning AI factory. The Manufacturing Disruptor: $POET – Redefining the Future The old way of building photonics devices involved carefully and expensively aligning tiny parts by hand. This process is too slow and costly for the massive scale that AI requires. POET Technologies was created to solve this problem. POET's main innovation is its Optical Interposer. This is a silicon platform that allows electronic and photonic parts to be assembled at the wafer level using standard automated equipment, much like regular computer chips. This "semiconductorization of photonics" promises to bring the scale and low cost of the electronics industry to the world of light. POET sells its integrated "optical engines" to companies that build the final products. This strategy was validated by a huge partnership with Foxconn Interconnect Technology. This is more than just a customer win. It is like giving a powerful new weapon to a global manufacturing giant. It allows Foxconn to enter the high-performance networking market and challenge the established players directly. The Deep Tech Catalyst: $ALMU – The Foundational AI Play Aeluma is the boldest bet in the basket. It represents a wager on a fundamental manufacturing breakthrough. While Aeluma is focused on creating new mass markets for sensors in cars and phones, its long-term importance to the AI data center is strategic and profound. Aeluma's AI strategy is focused on its development of quantum dot lasers for silicon photonics. A huge challenge in this field is putting the light source onto the chip. Silicon is great for guiding light, but it cannot create light efficiently. The current solution is to use a separate, off-chip laser, which is bulky and wastes power. Aeluma is pioneering a way to build these light sources directly onto large, 300mm silicon wafers. By using a special process to deposit tiny "quantum dot" materials directly onto the silicon chip, Aeluma aims to create on-chip lasers that are far more power-efficient and compact. This is a game-changing technology for next-generation AI systems where every bit of power and space matters. This technology has already won contracts from NASA and is being developed with AIM Photonics, a key US research consortium. This positions Aeluma not as a competitor in the data center today, but as a potential foundational technology supplier to the entire industry tomorrow. Overlap and Competitive Dynamics: A Multi-Layered Battleground The relationships between these five companies are complex. They compete, they partner, and they enable each other. Here are SOME of the overlap: Direct Competition (Giant vs. Giant): The clearest conflict is between Lumentum and Coherent. They are direct rivals, fighting for the same large contracts to supply lasers and networking parts to customers like Nvidia and major cloud providers. System vs. Components (MRVL vs. LITE/COHR): This is mostly a customer-supplier relationship. Marvell's switches and custom chips need the high-performance lasers and components that Lumentum and Coherent provide. However, as Marvell develops its own advanced optical solutions like CPO, it could start to integrate functions that the component makers currently sell separately, creating future competition. Indirect Competition (Innovator vs. Giants): POET presents a different kind of threat to the incumbents. It does not sell finished products that compete with Lumentum and Coherent. Instead, it sells its unique manufacturing platform to new challengers like Foxconn, who will then compete with the giants. This is a battle of business models. Foundational Partnership (The Aeluma Angle): Aeluma operates on a more fundamental level. In the long run, Aeluma is not a direct competitor to the others in the AI space. It is a potential supplier. If its on-chip quantum dot lasers prove to be the best solution, a future Marvell chip or a POET-enabled module could be powered by Aeluma's laser technology. Aeluma's success could lift the capabilities of the entire industry. This basket is built to succeed in multiple future scenarios. If the established giants continue to dominate, $LITE and $COHR will lead the returns. If the industry shifts to custom silicon and system-level solutions, $MRVL is perfectly positioned. If the industry moves to a more flexible manufacturing model, $POET is the key enabler. And if the next great leap forward comes from a breakthrough in core technology, $ALMU holds the potential for explosive growth. This is our strategy for investing in the essential hardware that will power the next decade of artificial intelligence. I have a lot more to say on this basket, so please understand that this is limited to the scope of an X post and I will follow up with more over the new few days. If you enjoyed this report, follow @crux_capital_ for more! Not financial advice or a recommendation

Gaetano

@Gaetano2026

Mar 23, 2026, 11:00 UTC

$LITE is one of my largest holdings After taking significant profit off the table when it was $380 (I know, in hindsight, not the best idea) I ended up adding back after earnings in the 400's and have been adding since. One thing I always am is dynamic. I am very happy to challenge my thesis and question where my knowledge gaps lie. Lumentum's earnings report opened my eyes to their long-term, sustained potential and just how strong their position in the market is. Honestly I second guessed Lumentum pre-earnings. But earnings are unpredictable. Especially when you are dealing with companies in a sector with so much downside if growth doesn't impress investors. While I am always happy to protect my capital, I am also happy to size up when I build more conviction.

Gaetano

@Gaetano2026

Jul 15, 2026, 17:31 UTC

@SayNoToTrading Lets ride!!! Added all of those today as well

Gaetano

@Gaetano2026

Oct 26, 2026, 22:14 UTC

@ConnorJBates_ Only because they (maybe) made my stocks go up more than others today😂

Gaetano

@Gaetano2026

Oct 27, 2026, 04:36 UTC

4 Adds today to the portfolio Now sitting at ~21% cash More diverse than I have ever been but still lots of exposure to AI if momentum wants to come back in Enough diversity to weather the storm if it doesnt t.co/VFurUXBi4E