저자: 미상 · 외국인
요약 보기
엔비디아 젠슨 황의 Computex 기조연설을 앞두고, AI 데이터센터의 랙 간 연결(구리·AEC·광) 병목이 다음 투자 테마라고 주장. Credo(크레도)를 실적 발표 전 유력 종목으로 제시하고, Broadcom도 연결성 대장주로 언급. 다만 Credo는 고객 집중 리스크와 높은 내재변동성, 과열 밸류에이션을 인정하며 실적 확인 후 눌림목 진입을 권고. 대량 블록 거래 흐름을 기관 매집 근거로 제시.
다듬은 글
Welcome back. If you are looking for some pre-market action, tonight is your night. Because all eyes right now are on the massive setup that's coming out of Taipei. And let me pull it up and show you what I'm talking about cuz just in a few hours, Jensen Huang is going to take the stage. Not even a few hours, in about 1 hour at 11:00 p.m.
Eastern. He's going to take the stage and deliver his highly anticipated Computex keynote address at NVIDIA GTC Taipei. During this address, we all expect that Jensen will unveil the next structural milestone for their next-generation AI five-layer cake ecosystem. The financial media after this and all day tomorrow is going to spend time talking about chip architectures and floating-point operations.
But on this channel, for me and for what I want to share with all of you guys, I'm going to listen to this speech and I'm going to try to identify the bottlenecks or the physical constraints in the market that Jensen identifies. Because NVIDIA's next challenge in my opinion is no longer compute alone.
It's about scaling connectivity across increasingly dense AI systems. And as these AI compute clusters grow, the industry is hitting a major wall. It's not just about how fast a GPU can process data or how quickly Micron's HBM can feed it. One of the most overlooked bottlenecks right now in the market is the physical rack connection layer.
And I'm And in my opinion, that's exactly why the single most important signal you need to listen for or watch for during Jensen Huang's keynote speech tonight isn't just a faster GPU architecture. It's the system-level design. Because NVIDIA right NVIDIA no longer sells individual processors. NVIDIA now is building entire AI factories.
And let me show you what I'm talking about. You probably remember it from his last speech overall, but you can see on the scale here just the complexity. When you start to think about it at this scale, like NVIDIA is doing, the challenge shifts from computer alone to moving data efficiently between thousands of processors.
When you bundle thousands of next-generation Vera Rubin Vera Rubin chips together into massive architectures like the Rubin NVL 576, standard passive copper cables become increasingly difficult to manage as rack densities rise. This will create This does create brutal airflow, power, and signal integrity challenges inside modern AI systems.
And now, I can tell you that the critics will say that the highest-end architectures bypass this prob- bypass the problems with copper entirely because they use optical fiber. And that's absolutely true for long distances between clusters, but going full optical inside an individual rack is still way too cost-prohibitive for massive day data center build-outs.
For many near-range rack deployments, hyperscalers are increasingly relying on active electrical cables, or AECs. These serve as a practical bridge between traditional passive copper and a full optical architecture. Heading into tonight's address for Jensen, I think one thing we have to listen to are the bottlenecks, but it's also becoming increasingly clear that the next phase of the AI trade is all about owning companies and finding stocks that allow these chips to communicate effectively together.
So, that's where we're going to dive in tonight. We're going to talk about a few stocks that we can look out for and why they help solve this problem in the market. If you guys haven't done it already though, . Because when Jensen lays out Nvidia's next-generation AI infrastructure roadmap tonight, we all need to listen closely to how he describes the fully integrated system.
Every system right now, every system has your weak link, and that's why I think tomorrow there's a company that's dropping its earnings after the market close, and these earnings are way more important than most real most investors realize right now. That company is Credo. And you can see on your screen that Credo will drop its earnings at the close of market tomorrow.
They're going to be reporting their financial results for the fourth quarter and for fiscal year ending May 2026. Many of you guys know Credo, and here's how they're been performing. On Friday, they had a pretty strong day, 6% growth of almost a $14 gain. In the after hours of the premarket, we're picking up another 3 and 1/2% right now.
Credo is not a generic chip maker. They specialize in high-speed connectivity solutions. Specifically, the specialized digital signal processors or DSPs that go inside those active electrical cables. An active electrical cable is essentially a copper wire with a brain built into both ends. Credo's DSP technology allows hyperscalers to extend copper connectivity while reducing many of the cabling power and airflow challenges that emerge as rack densities increase.
Now, to understand the true market opportunity here, you have to look at a Credo and how they position themselves against their direct rivals. So, two of those rivals that I want to talk about are Astera Labs and Marvell Technologies. So, Astera, they dominate the smart read timer chip market, cleaning up data signals directly on the server motherboard.
Credo, on the other hand, they they own the actual physical highway highway carrying the data out of the box. Marvell looms in the background or just looms in the market as a diversified giant dominating the longer range optical connections between server clusters. If Astera cleans up the signals inside the chassis, Credo owns the cables that links links it to the racks.
And if you want to look at this this asset objectively, you have to acknowledge the bear case You can because bearish critics will point out the fact that Credo carries an immense customer concentration risk. And let me show you that from a Motley Fool article right here. I'll pull that up and you can see this is a huge part of the bear case.
It says Credo Tech has one problem and here's why I'm buying anyway buying why I'm buying anyway. It says customer concentration risk and you can see right here is a headline risk. Roughly 80% of Credo's revenue hyperscalers. So there's no question that creates a bit of a risk and that's what the bears have been talking about overall because historically relying heavy heavily on just a couple of top hyperscalers like Microsoft or Amazon for a massive chunk of their active electrical cable revenue has been a risk with this company and it's a completely valid criticism.
If one of those tech giants slows down, it's a if it slows down specific rack deployment even in a single quarter Credo's revenue will be significantly impacted. However, Credo's print tomorrow one of the things we need to look for and read for when they drop their print tomorrow afternoon and tell they will tell us if their trailing revenue which has exploded past a billion dollars of the last 12 months.
If that trailing revenue is finally diversifying across a broader market. That Credo isn't the only player that's capitalizing on this system level scaling. If Credo represents the highly high velocity small cap pure play on physical cables, Broadcom is the unshakable institutional anchor of this entire connectivity thesis.
And once you take pull up Broadcom, we can all know how they've been operating in the market and how they've been performing in the market, but you can see Broadcom significantly it's trading up towards its 52-week high. $20 gain on Friday almost a 5% after hours right now picking up another $12 another 3% as as well.
So Broadcom like I said, they are the kind of the institutional anchor of this whole entire connectivity thesis. Broadcom just doesn't make components. They They own the underlying network fabric of the modern AI data center. They are the undisputed king of the high-end Ethernet switching silicon and custom AI ASICs.
When hyperscalers build out massive multi-rack factories, Broadcom's Tomahawk and Jericho switching platforms are what it what actually route the astronomical amounts of data between the server clusters. Broadcom also reports their quarterly earnings this week. They'll report on June 3rd, so Wednesday, June 3rd, and they'll drop their next earnings into the market.
And they represent the overall, I think, the scale-grade beneficiary of this exact architectural shift that we're seeing in the market. If you find small-cap concentration in Credo too volatile for your portfolio, Broadcom is the blue-chip monopoly you definitely want to consider within your portfolios.
Now, you ask yourself, what is the smart money doing ahead of the bell tomorrow? When you look at the institutional dark pool volume, we are seeing massive, highly calculated positioning in these connectivity plays, both in Credo and Broadcom. And some traders traders will argue that Nvidia already owns this space with their proprietary InfiniBand switches.
But remember the structural difference. InfiniBand is a networking protocol that manages the brain of the cluster. Credo's uh AECs, they are the physical muscles connecting the individual GPUs to the switch ports inside the tightest spaces in the rack. Broadcom, meanwhile, they provide the heavy-duty Ethernet switches that link those clusters together at the macroscopic scale.
And because of this, Credo, they act as a pure-play fast-moving small-cap indicator for this entire connectivity trade. However, we have to look at the valuation trap here. Let me show you this piece as well. There's a couple things I want to walk through just with regards to Criteo to give you the full kind of bear and uh bull case overall.
But you can see if you if you want to understand why the stakes are so high for Criteo right now. There's a few things I want to call out on this on the Market Chameleon data here. The stock rallied hard last week and it closed, you can see at 263, giving it a massive right now, almost a 41 billion market cap.
You can see it up here on the top of the screen where you see the market cap. Now, I want you to look right underneath that as well because this is the valuation trap that I want to talk about. The market is forcing this asset to trade at an astronomical uh trailing PE of 122 and even a further and eye-watering price to sales multiple of over 38.
This means that Wall Street has priced this stock to absolute perfection ahead of its upcoming earnings print. The options market knows that a violent collision is coming, which is why the 30-day implied volatility has spiked to a maximum 100% elevated risk, a level the stock hasn't exceeded all year and you can see that right here I'm circling on my on my screen.
But don't take my word for it as well. We need to look at how the smart money is actually moving behind the scenes for the stock cuz when we pull up the daily block trade screeners or we pull up the block trades for Market Chameleon, which I want to show you right now, you clearly see the institutional footprint is undeniable for the stock.
Look at the exchange column on the right. It is a literal wall of dark pool data. All morning and all afternoon, institutions were systematically dropping rhythmic identical blocks worth 1 to 1.7 million dollars at a time, keeping their accumulation entirely hidden off the public exchanges so they wouldn't trigger a retail front running frenzy.
And right in the the seconds of trading, right at the 4:00 a market on close or the closing cross, we saw a massive monster, a massive you can see it right here almost a $290 million uh trade. 1.2 almost 1.3 million shares overall right at the 23603. So, that that trade over there was a billion dollar institutional block that slammed at the closing cross in a single print.
The real kicker here though is what happened after the bell. So, let me pull that up. You can see the number of block trades that we saw right after the closing cross. Cuz once retail traders logged off for the weekend, the private networks lit up. Throughout the after hour session, the whales continued continued crossing positions off exchange including a standalone $11 million block and an $8.2 million block.
Notice how nearly every single one of these massive prints was executed at the exact fixed price of 23603. When you see millions of shares quietly exchanging hands in the dark pools after the market closed anchored at single price point like we see there with 260 23603, it tells you everything you need to know.
The institutions aren't guessing on the AI infrastructure cycle. They are aggressively locking in their stakes before the catalyst drops. So, now that we've gone through both sides of the coin, here's your direct tactical playbook for tomorrow's trading session. First, if you can listen to Jensen's speech tonight, do it or pick it up first thing in the morning before the market opens.
Cuz we have to look to see how the market positions itself ahead of the the opening bell based on Jensen's comments on system level architecture. If the early capital centers on inter-rack copper containment versus optical scaling, the money is going to flow straight into the hardware connectivity plays.
And those plays would be Broadcom, you know, Astera Labs, Marvell, and of course Credo. But as I showed you, Credo is trading with massive implied volatility ahead of tomorrow's earnings print. So because of that historical valuation volatility, I would not recommend blindly gambling and jumping in on Credo before the the morning's open.
Instead, watch the key institutional support levels, and if their earnings report confirms accelerating hyper scaling hyper scale demand hyper scalar demand, excuse me, and a broadening customer base, Credo represents a powerful structural hold and a great opportunity potentially for an entry point if we see any bit of pullback.
Cuz remember, a strong print for Credo tomorrow is also an immediate leading indicator for Broadcom heading into their print or their earnings later this week on Wednesday. The biggest takeaway for me from this video is Jensen Huang's keynote speech tonight is no longer about just the raw metrics of a new a new GPU chip.
It will be a reminder that the AI race is becoming a systems race. Compute, memory, networking, power, cooling, and connectivity all have to scale together. That's why investors can no longer focus exclusively on Nvidia. The next winners may come from the bottlenecks surrounding the GPU. The trade has matured into buying the companies that make the chips actually work together.
Without advanced connectivity, the world's fastest processors are nothing more than expensive paperweights. Guys, thanks for watching the video tonight. Enjoy the speech. I hope you get a chance to watch it. I'm going to be watching it here in about 30 minutes, but I'll see you tomorrow after the closing bell. Have a great night.
Comments 0