4 Stocks Hiding Underneath Every AI Data Center... Buy Now?

[00:00:00] Speaker 1: If you're still buying Nvidia, Microsoft, and the other big AI names, stop. That trade is over. You are never going to make money buying yesterday's winners. Today's best opportunities lie in a layer underneath all of that. A layer that has to exist before any of those companies can do a single thing. It's called photonic infrastructure. Right now, four small companies control it. Today, I'm going to show you exactly who they are. Here's the thing about AI that most investors miss. The bottleneck is not the software. It's not even the chips. It's moving the data, getting it from one part of a data center to another at speeds that don't create lag. Without generating so much heat, the whole building melts. The answer to that problem is photonics, moving data as light through fiber instead of electrons through copper. This technology is not theoretical. It's already deployed in every major AI data center in America right now. The companies building this infrastructure are quietly becoming some of the top performers of this decade. Most of your friends have never even heard of them. Today, I'm going to cover all four. Let's dive in. Now, before we get to the cables and the transceivers and the signal processors, we have to start at the very beginning. Before any AI data center gets built, someone has to make the substrate. This is the raw material that everything else is built on. The company that makes it is AXT Incorporated, ticker AXTI. AXT makes compound semiconductor wafers. Want to get technical? They make iridium phosphide and gallium arsenide substrates. This is the stuff used to manufacture the lasers inside every optical communication system. No AXT wafers, no lasers. No lasers, no photonic cables. No photonic cables in these AI data centers cannot operate at the speeds the hyperscalers need. It all starts and ends there. Now, at the time of this recording, the stock is trading around $63. 12 months ago, it was $1.13. This thing has gone from penny stock territory to $63 a share in under a year. It is the little stock that could, driven entirely by hyperscaler demand for these photonic infrastructure components. And this stock is showing no signs of slowing down. Now, once you have the wafers and the lasers, you need something to process the signal. To take the data moving at the speed of light and make sense of it at the network level. That is Marvell technology, ticker MRVL. Marvell builds digital signal processors, the chips that sit inside optical networking equipment and manage the data flowing through photonic cables. Think of them as the traffic control system for the light. The wafer gets you into the building. Marvell's chip tells the light where to go and how fast to get there. Their technology moves data at speeds that make traditional copper-based systems look like they're running on dial-up. Barclays just upgraded the stock with a $150 price target, specifically citing the growth in their optical business. Now, this stock has been on a tear from $100 to over $130 this month alone, and it's still below Barclays' price target. Now, it's been a volatile couple years for Marvell, but shares are now breaking out to new all-time highs. This is an institutional-grade stock with direct growing exposure to every dollar that hyperscalers are spending on AI data center build-out. Now, quick break here, but if you want to track plays like these alongside my actual trades, check out my Black Ops service. For just $5, you will get an entire year of access, which includes live one-hourly group mentoring sessions, my weekly newsletter, indicators, and a bunch more. And again, right now, it's just $5. There's a link in the description, or you can scan the QR code. Now, we need something to actually put the data onto the light to take an electrical signal from a server and convert it into a photon that can travel through fiber. That's what a transceiver does. And applied optoelectronics, ticker AAOI, is one of the leading manufacturers. Their optical transceivers convert electrical signals into light. That light travels through fiber cables, crossing an entire data center in microseconds, all while carrying massive, massive amounts of data with no degradation and without generating the heat that copper produces. Pull applied optoelectronics out of the equation, and the entire system stops working. You'd have the wafers, you'd have the signal processors, but nothing actually put the data on the fiber. Now, the stock has been one of the most explosive moves in its entire space. It finished March at $75 a share. Two weeks later, it was breaking $150. This is a mid-cap company with lots of room to grow, with massive hyperscale orders that it didn't have a year ago. Rosenblatt initiated coverage with a buy rating last week, citing the company's positioning in optical transceiver technology. The next generation of data center speeds. Now, the last company in the chain may be the most most important of all four. Every single chip that goes into a data center, every photonic chip, every semiconductor, every optical component has to be tested before it ships. That's where air test systems come in, ticker AEHR. Their machines test photonic and semiconductor chips before deployment. They are the last line of defense before that infrastructure goes live. If a chip fails in the field, inside a live data center, the cost is enormous. Testing it before it ships costs a fraction of that. Every chip in every data center has to pass through a system like this. And as the volume of photonic chips being made scales with the colossal AI demand, so does the demand for AIR's testing equipment. The stock is around $70. Last month, it was at $30. This is a $2 billion company that could steal 10x from here. And the business case is incredibly simple. More chips being manufactured means more chips that need to be tested. The demand here is a structural one. Now, one more thought that you may be thinking, and that is this. If photonics is replacing copper cables inside these data centers, that's bad for copper, right? Is this a headwind for this commodity? Is it going to prevent the price from going higher? And I get why you'd ask that. It is the right question, but the answer is no, and here's why. Most people think of copper in a data center as the data cable, but that is a small fraction of the copper in the building. The bulk of the copper in any AI data center is carrying power, not data. Power distribution from the grid to the racks, bus bars, cooling system motors and pumps, electrical grounding, the short distance stuff, GPU to GPU connections between the racks. Now, photonics replaces the backbone data transmission cables, the ones that run between racks, between rows, between buildings. And as of the end of 2025, roughly 85% of new backbone employments already use fiber over copper for these long runs. But here's the thing. Even Nvidia's latest and greatest, their Blackwell NVL 576 system, which is just shown at OFC, still runs copper inside the rack. The GPU to GPU connections remain copper. It only shifts to optical fiber between racks. Google's own engineers said publicly that scale-up interconnects, the densest, most performance-critical links, remain copper-based. Corning's 2026 data center forecast called fiber-fed in-rack connections still in the "first field trial stage." Fiber has taken the long runs. Copper still owns the rack. And on the power side, which again is the majority of data center copper, AI actually makes the problem bigger, not smaller. Every new AI data center draws more power than the last one. More power means more copper than electrical distribution. Light doesn't carry electricity. That problem does not go away. And the numbers back this up. Sprott estimates data centers will need 1.1 million tons of copper per year by 2030. 1.1 million tons. That's what? 2.2 or 2.5 billion pounds. Goldman Sachs projects data center copper demand rising 72% by 2050, eventually reaching 7% of all global copper consumption. The photonic substitution is real. The demand surge from power or infrastructure is bigger, though. So the net effect on copper, in my opinion, is bullish, not bearish. I'm still holding my copper position. I plan to add on as price progresses. Because the photonic buildout is not a threat to my thesis. If anything, it's part of the same AI infrastructure wave driving it higher. And when you put the whole picture together, the AI trade everyone knows is NVIDIA and Microsoft. But the real infrastructure play is the full supply chain. AXT, which makes the wafers that build the lasers, the foundation. Applied optoelectronics builds the transceivers that put data onto light. The connective tissue. Marvell processes the signal at network speed. There's your engine. And air test systems test every chip before it ships. That's the quality gate. Underneath all of it, copper. Still running the power. Still running the rack. Still very much in the game. Now look, do your own due diligence, please. But if you are only looking at the front end of AI, you are missing the companies that make it physically possible. Don't forget to subscribe to the channel. And that $5 Black Ops special is still going on. Click the link in the description, scan the QR code, or go to TradeWithRoss.com to get signed up. I'll see you in the next video.