Why the NVIDIA Phase is Already Crowded
If you’re looking to invest in AI stocks, forget about Nvidia. That was the first wave of the AI trade — the obvious one: semiconductors. Nvidia has been the clear market leader from the beginning. That’s why the stock is up 2,000% in three years. That one’s played out in 2025.
We’ve seen the pick-and-shovel plays. It was nuclear, natural gas, and a bunch of clean energy stocks. They made huge gains as investors placed their bets on what would power these colossal data centers. And while energy is crucial for widescale deployment of AI, there is another problem slowing development even more.
Now, I haven’t heard anyone else talking about this, but it is a concentrated bet on one of the most profound hardware transformations occurring in computing right now.
The Overlooked Bottleneck Slowing AI Scaling
So today, what I’d like to do is share seven stocks that will benefit from the transition. I’m going to give you the names, details, and ticker symbols of each one.
How an AI Data Center Rack Really Works
To fully comprehend this opportunity, we first need to understand how an AI data center rack works.
Put simply, the storage server stores massive amounts of data that feeds the AI compute nodes — that’s the GPUs that everybody buys from Nvidia. This is where all the calculations are performed. The power and cooling units at the bottom are pretty self-explanatory.
But our focus here, which is often overlooked, is the top-of-rack switch.
Why the Top-of-Rack Switch is the Key
The top-of-rack switch manages data traffic between the storage server and the GPUs, as well as between the server racks. With AI companies planning one-gigawatt data centers, even with the best AI training GPUs, each data center will have more than 10,000 server racks.
That is why communication between and within these racks is crucial to maintaining efficiency and scaling effectively.
The Copper Wall Explained in Plain English
Which brings us to the copper wall.
The copper wall is the physical limit beyond which copper cables cannot transmit data fast enough over a given distance. As signal frequency increases and more data is pushed through, the current density concentrates near the surface rather than flowing uniformly. This leads to higher resistance and signal loss.
In simple terms, the faster the data transfer, the shorter the distance it can travel.
In the early 2010s, a standard passive copper cable could reliably transmit 10 gigabits per second up to seven meters. With exploding AI workflows, speeds have increased to 112 gigabits per second, where traditional copper can only reach about two meters.
Why 224G Makes Passive Copper Fail
The upcoming generation of servers will operate at 224 gigabits per second, at which point the reach of passive copper drops to less than one meter. That one-meter limit is the copper wall.
A standard data rack is roughly two meters tall. If a switch at the top of the rack cannot reliably send a signal to a server at the bottom using passive copper, the traditional architecture fails.
To scale AI clusters from thousands to hundreds of thousands of GPUs, the industry has no choice but to move beyond copper.
The Shift to Optical Interconnects and Fiber
The solution is optical interconnects for short-reach applications that were previously dominated by cheap passive copper.
Optical interconnects convert electrical signals into pulses of light, send them through fiber optic cables, and allow massive amounts of data to travel with minimal loss and far greater distance. Data moves faster and farther.
Even fiber has limits, but GPUs are improving at a breakneck pace.
Blackwell, 800G ports, and What Comes Next
Nvidia’s latest Blackwell cluster requires switches that support 800-gigabyte-per-second ports, which depend on underlying silicon that can support 224-gigabit speeds. The next generation of Nvidia chips is expected to double that speed.
As with any transition, this won’t happen all at once. It will be a gradual replacement of existing components.
The 3- to 4-year Roadmap: AEC → LPO → CPO
For the rest of 2025 and into 2026, passive copper cables at 224 gigabits per second are limited to less than one meter. The simple solution is active electrical cables, or AECs. These use a chip to boost the signal midway, extending reach to two to three meters.
By 2027 and 2028, as speeds move toward 1.6 terabits per second, even AECs become too thick and inefficient. The solution becomes linear drive optics, or LPOs, which use the GPU itself for signal processing.
Beyond that, starting around 2029, the final step is co-packaged optics, where the signal is converted to optical directly on the chip. This achieves near-perfect transmission with minimal losses but requires a complete redesign of server manufacturing.
Now, if you haven’t already, sign up for my War Room trading service. It is just five bucks for the entire year. You’ll get live weekly mentoring sessions with me, proprietary indicators, bonus reports, and a bunch more. It’s the best $5 you’ll spend this year.
Big Money Moves in Optical Interconnects
Major players are already positioning for this shift. Marvell Technology recently announced the acquisition of Celestial AI for $3.25 billion. Celestial is a leader in optical interconnect platforms.
Nvidia is also working to fuse compute, NVLink, NVSwitch, and networking silicon to keep traffic within its ecosystem.
Corning, best known for Gorilla Glass, is gaining momentum in optical communications. Analysts expect roughly 39% year-over-year growth in Corning’s optical communications segment due to data center demand.
The 7-Stock Basket and How it’s Built
The best way to capture the upside is by building a basket of optics-focused companies. I’ve divided them into three segments.
Segment 1: Analog and Optical Core
This is where I would focus today.
- MACOM Technology Solutions (MTSI)
MACOM’s high-performance analog chips could become critical as the industry transitions to LPO. Their growth is driven by data center demand, and they benefit whether the industry sticks with pluggables or moves to custom optical engines. - Fabrinet (FN)
Fabrinet takes no technology risk. They manufacture for the winners. They are a primary manufacturer for Nvidia’s optical interconnects and a key partner for Lumentum and Coherent. - Coherent (COHR)
Coherent manufactures its own indium phosphide lasers and modulators, giving it a margin advantage as speeds ramp to 1.6 terabits.
Segment 2: Infrastructure Architects
- Arista Networks (ANET)
LPO cannot work without switches designed to support it. Arista has been a vocal supporter of LPO-capable switching and is actively validating these modules. - Marvell Technology (MRVL)
Marvell dominates the DSP market but is aggressively pivoting to co-packaged optics. If DSPs persist, Marvell wins. If CPO arrives, Marvell’s silicon photonics platform wins.
Segment 3: Moonshots
These are high-risk, high-reward plays.
- POET Technologies (POET)
POET’s optical interposer platform enables chip assembly using standard automated equipment. The company has pivoted from R&D to commercialization and raised $150 million to scale production. - Lightwave Logic (LWLG)
A pure play on material science. Lightwave’s electro-optic polymers allow faster switching at lower power, bypassing current silicon limitations.
The Biggest Risks: Capex and Geopolitics
The biggest risk is a slowdown in AI capital expenditures. Even a modest cut in spending could result in sharp declines for optical stocks due to leverage.
Geopolitics is another risk. Optical assembly is concentrated in Thailand and China, and renewed trade tensions could pressure margins.
The Real Thesis: Why This Shift is Inevitable
From my perspective, the transition from copper to optics isn’t a question of if, but when. These stocks are a bet on that inevitability.
If you haven’t already, click the link below to sign up for my my War Room trading service.and get an entire year of stocks picks like this for just $5 today
