What truly bottlenecks an AI data center isn't the single winning chip, but the entire photonics industry chain that converts electrical signals to optical signals and moves massive amounts of data out. Brian's core judgment is: when the industry upgrades from 800G to 1.6T, and even marches towards 3.2T, those who often gain the most supernormal profits first are not the hottest star companies in the spotlight, but suppliers like Corning, Amphenol, and Ciena, which all giants must rely on, as well as more upstream materials and testing segments. This podcast comprehensively sorts out the complete photonics supply chain for you, and highlights the golden high-growth stocks you need to keep a firm eye on before Wall Street's main force rushes in belatedly.
Key Insights Summary
Why AI Data Centers Must Shift to Optical Communication
· "Copper cables are hitting a physical limit; every data center must eventually shift to optics. China has already proven to the entire industry that this road can go even further."
· "Once the data transmission distance exceeds about 3 feet, copper cables rapidly lose their advantage, generating more heat and consuming more power; while optics can solve these problems simultaneously."
· "Shifting from electricity to light, that's the core significance of photonic technology."
Why Investment Opportunities Often Lie in the Supply Chain Rather Than Star Companies
· "As long as a new technology is proven viable, the greatest wealth often flows first to companies that all participants must rely on, not the single name that makes the headlines."
· "Glass, lasers, connectors, materials, and testing equipment—none can be missing. This is where the photonics industry chain holds the most value."
What to Watch for in Corning
· "Corning's optical communication revenue grew 36% last quarter, but its corresponding profit grew 93%, indicating that pricing power and economies of scale are simultaneously materializing."
· "Corning has become a core supplier specified by Meta, Amazon, Google, Microsoft, OpenAI, and Nvidia. There is no second competitor in the world with such a client list, and these relationships have already translated into pre-locked multi-year revenues."
· "Meta has committed up to $60 billion, Amazon has also signed multi-billion dollar contracts, and two other hyperscale customers have signed agreements of similar magnitude. Much of this revenue was locked in years before the fiber was even pulled."
What Amphenol and Credo Represent
· "If you want a broader-coverage, relatively less volatile, and fairly valued optical communication interconnect play, Amphenol is a name very worthy of long-term attention."
· "Credo, on the other hand, is a bridge between the old world and the new, squeezing the last drop of life from copper cables within racks while also extending into the optical communication side."
· "Credo's risks are also clear: extremely high customer concentration. If just one major client pauses procurement, the stock could be heavily hit."
Opportunities at the System, Upstream Material, and Testing Layers
· "Ciena's value lies in enabling existing fiber to carry more data without the need to dig and lay new lines."
· "AXT is positioned further upstream, almost a scarce supplier of key wafer materials for optical lasers, but it also faces extremely high risks related to Chinese export licenses."
· "VEO Solutions is more like the 'shovel seller' in the optical communication world. Whether it's fiber links, transceivers, or system equipment, everything must be tested before deployment and monitored afterwards. VEO sells precisely the testing tools all these devices must go through."
Thematic Allocation and ETF Choices
· "If you don't want to pick companies one by one, there are now pure photonics theme ETFs that can cover this mainline with one click."
· "But these funds are newly established, still small in scale with relatively high fees, better suited for a watchlist rather than blind chasing."
China Has Lifted the Fiber Transmission Ceiling Significantly Higher
Brian: Engineers in China recently lit up a single glass fiber strand capable of carrying five times more data than the current world standard elsewhere. This wasn't a newly laid line or large-scale excavation, but activating a fiber thinner than a human hair within already buried cables. What used to take about half an hour to move the entire Library of Congress data can now be done in roughly five minutes.
This technology isn't truly deployed in the US yet, which shows just how fast this field is evolving. AI is generating a data deluge far exceeding current transmission capacity, and US data centers are increasingly lacking this capability. All hyperscale cloud providers will eventually need it, and they won't manufacture the entire system from scratch themselves—they'll buy it.
They buy glass, lasers, and chips that convert electricity to light, and the suppliers truly capable of providing these are only a handful. Having spent years observing the industry from the supply chain procurement side, my experience tells me the pattern of such opportunities has hardly changed. As long as a new technology is proven viable, the greatest wealth often flows first to companies that all participants must rely on, not the single name that makes the headlines.
Why Photonics Has Become the Core Variable Now
Brian: Today, I want to break down the entire photonics industry chain for you, explaining the corresponding publicly traded companies at each layer. Why photonics, and why now?
The answer boils down to a hard constraint. Inside a data center, all chips need to communicate with each other. Over short distances, copper is still the winner; but once you want to transmit data beyond about 3 feet, copper's problems quickly surface. The longer the distance, the more severe the heat and power consumption become.
Light solves almost all these problems at once. It travels farther, generates less heat, and consumes only a fraction of the power of copper solutions. Shifting from electricity to light, that's the core significance of photonic technology.
More importantly, this inflection point is now very clear. Every data center is upgrading from 800G to 1.6T connections, and even 3.2T is on the table. Simultaneously, the fiber mentioned earlier from China has pushed the industry's ceiling even higher.
Breaking it down, the most critical layers include: the glass, fiber, and cables that actually carry the optical signals; the connectors that link everything together; the system equipment responsible for lighting the fiber and transmitting data between buildings and countries; and the underlying materials and testing equipment. I've covered chips, lasers, and silicon photonics themselves in previous videos, so today's focus is on these often-overlooked yet profitable segments.
Within these groups, I'll provide names worth putting on your watchlist. However, I must clarify upfront: many targets have already risen significantly, and genuinely friendly entry points will likely require waiting for pullbacks.
The Glass and Fiber Layer: Why Corning is the One to Watch Closest
Brian: Starting with the most basic glass, the first name is Corning. This is a 175-year-old materials company, the one actually responsible for drawing optical fiber—the kind of "glass thread" mentioned at the video's start. Its global fiber market share is about 20%, making it a very core player.
Where Corning truly pulls ahead is technology. Its latest generation fiber can pack roughly twice the number of cores into the same physical space as standard cables, which is exactly what extremely crowded AI data centers desperately need. Its bend-resistant glass is also difficult for peers to replicate. Combined with the world's largest fiber factory and U.S. domestic supply compliant with "Buy America" rules, these factors together form its real moat.
It's precisely because of this that Corning has simultaneously become a core supplier specified by Meta, Amazon, Google, Microsoft, OpenAI, and Nvidia. There is no second competitor in the world with such a client list, and these relationships aren't just at the 'good cooperation' level; they have already translated into pre-locked multi-year revenues.
Current industry demand for fiber is growing about 22% to 25% annually, but new supply capacity is only about half that growth rate, with lead times stretching beyond 60 weeks. Consequently, hyperscale customers are booking capacity years in advance, even prepaying to lock it in. Meta has committed up to $60 billion, Amazon has also signed multi-billion dollar contracts, and two other hyperscale customers have signed agreements of similar magnitude. Much of this revenue was locked in years before the fiber was even pulled.
What catches my eye most is the profit elasticity. Corning's optical communication revenue grew 36% last quarter, but profit in this segment grew 93%, more than double the revenue growth. This is what pricing power and economies of scale look like together. At the company-wide level, its operating margin has also risen from around 8% two years ago to over 16% now, with management targeting 20% by year-end.
Of course, we must acknowledge reality: this story is no secret. Corning's current PEG is near 3, and its P/S ratio is about 9x, which is not cheap for a materials company. So, if you want the steadiest, least dramatic way to configure the fiber layer, Corning is suitable, but a more reasonable approach is to wait for the next more substantial pullback.
Core Companies in the Interconnect Layer: Amphenol and Credo
Brian: Moving into the connection layer—the one that actually links all components together—the first name is Amphenol. It's a rather low-key giant, making high-speed connectors and cables, including both copper and optical, and its products are now found in almost every newly built AI server rack.
The key to understanding Amphenol is that it is essentially an exceptionally efficient acquisition machine. In January this year, it spent $10.5 billion to acquire CommScope's entire fiber connectivity business, overnight transforming from a connector company into a truly significant fiber player. Now, AI data center business has become the core engine of its entire company, its largest single segment, with organic growth exceeding 80% last quarter.
Its order book also reached a record $9.4 billion, with new orders still coming in faster than shipments. As quarterly revenue jumped from around $4 billion to slightly over $7 billion, its operating margin wasn't dragged down but actually expanded from 22% to nearly 28%.
This is noteworthy. Normally, a $10+ billion acquisition would impose integration pressure for at least a year or two, often causing margins to dip first. But Amphenol's margins went up, indicating its strong capability to rapidly integrate acquired targets into its own high-standard operating system. Consequently, such a large deal didn't become a burden but rather a profit enhancer.
What's even rarer is that its valuation isn't unreasonable. Amphenol's PEG is only about 0.7, with a P/S ratio around 7x. For a company growing this fast, this kind of valuation is uncommon. So, if you want a broader-coverage, relatively less volatile, and fairly valued optical communication interconnect play, Amphenol is a name very worthy of long-term attention.
The second name also in the connection layer is Credo Technology. Its role is more like a bridge between the old world and the new. On one hand, it uses low-power chip technology to maximize the transmission capacity of copper cables within racks; on the other hand, it's also working on optical communication chips and cables, seamlessly transitioning once signals must travel farther.
It recently acquired a silicon photonics company, complementing its end-to-end product stack up to 1.6T, and is already shipping to all top five US hyperscale cloud providers. Its growth has been quite explosive: quarterly revenue surged from $135 million to $437 million in just six quarters, more than tripling.
Another telling metric is its gross margin of about 68%. This number is more typical of a software company than a hardware one. Meanwhile, its operating margin has also nearly doubled with scale, reaching 37%. Management's revenue guidance for the next fiscal year still indicates growth exceeding 80%.
But the risks here are very specific and must be considered seriously. Although it supplies the top five customers, just three of them account for 88% of its revenue. If any single hyperscale customer slows procurement, this stock could be swiftly punished by the market. Additionally, insiders have been steadily selling during the rally, and the company's current P/S ratio is around 35x, essentially pricing in "almost everything continuing smoothly." A PEG near 1 indicates strong growth, but companies like these are more akin to high-conviction plays, better considered only after deep pullbacks.
The Key Player at the System Layer: Ciena
Brian: Moving up further, we reach the system layer—the one that lights up the fiber and moves data between buildings and countries. The most critical name here is Ciena. It's a leader in coherent optics in the West. The company's proprietary WaveLogic technology is the world's first solution to pack 1.6T of data into a single wavelength of light. You can think of this as a "cheat code for capacity expansion without digging," because it enables existing fiber to carry more data without the need to lay new lines.
And this isn't a lab demo. In just two quarters, this single product has secured 49 customers. Simultaneously, its position in key customer relationships is also very strong. Its relevant solutions have been adopted by three of the four major hyperscale cloud and cloud service providers, and cloud customers now contribute nearly half of the company's revenue.
For me, the most crucial data point remains the order backlog. Last quarter, Ciena's backlog increased by approximately $2 billion in 90 days, reaching nearly $7 billion. Almost all of it is scheduled for delivery next year, meaning it has already locked in revenue equivalent to more than a year's worth.
As revenue hit a new record high, growing 40% year-over-year, its operating margin also doubled from under 8% to over 15%. The problem is, the market also knows these good news, so the valuation is already very aggressive. Ciena's forward P/E is currently around 120x, which almost prices in "perfect execution." Therefore, while fundamentals are strong, price discipline is essential.
The Truly Upstream Bottlenecks: AXT and Testing Equipment Company VEO Solutions
Brian: Now we enter my favorite layer: the 'suppliers' suppliers.' The most upstream name in this entire discussion is AXT. The number of companies globally that can truly produce indium phosphide wafers is limited, and this special crystal is precisely the key material that all optical lasers must be built upon. From this perspective, it almost inherently has a moat because this crystal growth capability isn't something you can create overnight.
Now, as all laser manufacturers scramble for supply, AXT's order backlog for such materials has surpassed $100 million, setting a record. But its risks are also among the highest on this list and are very specific. Almost all its manufacturing is in China, and China now requires government approval for each export order. This directly impacted its revenue realization last quarter; even with a record backlog present, it doesn't guarantee smooth shipment.
Additionally, the company raised about $550 million, causing shareholder dilution; insiders also net sold over $70 million worth of stock during the rally. More troubling, it still isn't profitable, although losses are narrowing, and gross margin has improved from 17% to around 30%.
So, AXT's story is real, and the bottleneck is real, but if you look at its P/S ratio approaching 66x, you'll understand it's more like a high-volatility, high-risk "lottery ticket" for a small position, not a core holding suitable for a heavy allocation.
The final name in this layer is VEO Solutions. It is more like the 'shovel seller' in the optical communication world. Whether it's fiber links, transceivers, or system equipment, everything must be tested before deployment and monitored afterwards. VEO sells precisely the testing tools all these devices must go through.
The beauty of this position is its insensitivity to specific winners. You don't need to bet on which laser company ultimately wins or which transceiver becomes mainstream because, whoever wins, they'll eventually need to test on its equipment. This is a type of business position I personally like a lot.
For years, VEO's revenue was largely stagnant, stuck around $285 million per quarter. It wasn't until AI infrastructure build-out truly accelerated that its network testing business began to explode, rising over 54%, as data centers rushed to validate massive amounts of new optical communication equipment. Now its quarterly revenue has surpassed $400 million, and operating margins have returned to double digits.
Beyond this, the company has a quieter but high-margin second business producing anti-counterfeit coatings printed on global banknotes, providing additional high-margin support. Note that insiders, including the CEO, have also been consistently selling. Its current PEG is about 1.4, not too extreme, but like others on the list, it has seen significant gains, and the more reasonable approach remains to let it cool off first.
If You Don't Want to Pick Stocks Individually, There's a New Pure-Photonics ETF
Brian: Many might ask, why didn't I focus on Coherent, Lumentum, Marvell, Broadcom, and the actual chip-making foundries today?
These companies are certainly worth putting on a watchlist. They are in the laser and silicon photonics segments at the heart of this supply chain and are a very important part of the overall theme. It's just that I've previously done comprehensive breakdowns of them, so today I wanted to focus on supplementing the parts others often skip: glass, connectors, system equipment, and their upstream suppliers.
If you don't want to pick companies one by one, there's also a way to configure the entire theme with one click. A pure photonics theme ETF has recently appeared on the market, ticker FOTO, the Tuttle Capital Pure Play Photonics Fund.
One thing I like about it is its "pure theme" screening. It excludes large conglomerates where photonics is only a small part, concentrating holdings on genuine optics companies. Its largest holdings are precisely those laser and transceiver companies I mentioned earlier, like Lumentum, Coherent, Fabrinet, also including IPG Photonics and Inphi. The entire fund has only 15 names, with the top 10 holdings accounting for nearly 90%, making its style very concentrated and explicitly betting on "pure photonics."
Of course, full disclosure is necessary. This fund is only a few weeks old, with no real long-term performance record yet. Its AUM is about $140 million, with an expense ratio of 0.75%. So, you should definitely do your own thorough homework. But if you just want to put the "copper-to-light" theme on your radar initially, FOTO is indeed worth watching.
Summary: Copper's Limit Has Been Reached; True Beneficiaries Will Spread Along the Entire Light Chain
Brian: To conclude. Copper is hitting a physical limit, and every data center on Earth must migrate to light. China has already proven to the entire industry that this road can go even further.
Capital is already flowing into this chain. The real question is no longer whether this will happen, but which batch of companies can capture the most incremental gains during this accelerated transition.
If you combine the laser and silicon photonics companies I discussed before with the glass, connectors, systems, and upstream suppliers from today's video, you essentially have a fairly complete map of the photonics industry chain from start to finish.
