A Massive CPU Shortage Is Here — 3 Stocks Positioned to Win Big

Welcome back to the channel everyone. Over the last year, everyone on Wall Street has been completely obsessed with graphics processing units, GPUs, uh the standard AI hardware accelerators. But while the entire retail crowd was staring at one side of the data center, a massive structural pivot quietly happened on the other side. A top semiconductor executive just stepped up to an earnings call and completely rewrote the playbook for the next 5 years. But before we get started, if you want to know how the smart money is actually front running this hidden tech boom, make sure to hit that subscribe button and turn on notifications right now. >> Yeah, Rachel, during AMD's recent quarter 1 earnings conference call, chair and CEO Dr. Lisa Sue completely altered the market's long-term expectations. For months, the Bears have argued that your standard CPUs were becoming obsolete. Legacy tech in an AI dominated world. >> Well, Lisa Sue completely blew that narrative out of the water. She announced that the total addressable market for server CPUs is growing at 35% annually through 2030. Now, that's up sharply from the 18% target that management shared just last November. This means they've officially doubled their long-term market forecast from $60 billion all the way to over $120 billion by 2030. Uh so today we're going to be breaking down why this shift shift is happening as well as the three stocks that we're looking at buying right now to capture this major inflection point. Jose, let's start with your first pick. >> Yeah, so obviously I have to start with the source of the shock wave itself, Advanced Micro Devices, ticker AMT. Look, the market completely misunderstood why Lisa Sue doubled the server CPU outlook to over 120 billion on that earnings call. [snorts] It all comes down to a massive transition from basic large language models to agentic AI systems. Training model is a heavily reliant on GPUs. But when you deploy autonomous AI agents at a global scale systems that spawn subtask, execute code, manage data pipelines, and search the web, the primarily bottleneck shifts to orchestration compute. Agents require massive sequential computing power which runs on CPUs. In the recent quarter 1 print, AMD put an absolute masterclass. Total revenue jumped 38% year-over-year to $10.3 billion, delivering a non-GAAP earnings per share of $1.37, which easily beat the street. Their data center segment spearheaded the growth skyrocketing 57% year-over-year to 5.8 billion, driven heavily by their epic server CPU sales. More importantly, their free cash flow more than tripled to a record $2.6 billion. With their upcoming sixth generation architectural ramp, AMD is targeting dominant revenue market share in server CPUs. They aren't just the alternative to Nvidia on the accelerator side. They are the absolute tech players in the compute agentic side right now. Yeah, Jose, the the thesis for Agentic AI, it's a phenomenal point and if millions of AI agents are running subtasks simultaneously, that data center general compute layer has to expand and it has to expand exponentially. I want to take a look at the operational margins and the execution risks here. So AMD's data center growth is impressive. There's definitely no denying that. Now, overall net income for Q1 actually fell sequentially quarteron quarter. And I mean, you look at this, AMD is investing billions of dollars into shifting production lines towards advanced packaging. They're securing foundry capacity for the next generation chips. So there are a lot of practical explanations for that sequential decline. We've also seen headwinds in their their gaming, their embedded segments, of course, of late. But if they reach a scenario where they're burning through cash just to keep up with the R&D requirements for both high-end GPUs and this newly expanded CPU market, do you think they can scale their earnings fast enough to justify their current valuation or are you concerned they might stretch themselves maybe too thin as they're fighting this sort of two-front war against Intel and Nvidia? >> Yeah. Yeah, that that's a crucial point there uh Rachel regarding kind of that segment drag, but that's exactly why the CPU TAM expansion is such a massive catalyst. Server CPUs typically carry much higher mature gross margins than early stage lowy yield AI accelerator RAMs. Now, management has already projected that their data center mix scales to become the absolute majority of the business. Corporate gross margins will consistently expand towards the 56% range over the next few quarters. They aren't stretching themselves thin. In my opinion, they're using their high margin epic CPU cash cow to entirely self-fund their AI hardware roadmap. Hypers scale cloud providers like Microsoft Azure and Amazon AWS are aggressively buying x86 servers right now to handle the stepfold increase in daily agentic processing traffic. AMD has the supply capabilities to step into that supply vacuum. While Intel, unfortunately, still struggles with its widely reported fab delays and ser processor shortages. The market is pricing AMD like a cyclical chip stock, but it's actually a structural data center monopoly for now. >> Yeah, I I think that's right. I mean, and the cloud data center visibility is also clearly there. Uh le let's move to your second pick though because this one takes the exact same architectural shift but it's applying it away from the cloud and directly onto local devices. >> Yeah. So stock number two is going to be Qualcomm ticker QCOM. Everyone associates them Rachel with the cyclical smartphone market but Lisa Sue's announcement completely shifts the calculations here. If the circuit if the server market is compounding at a 35% annual to across over $120 billion, the major question is how infrastructure adapts to handle these massive new workloads. Qualcomm is quietly positioned itself as kind of this ultimate dark horse here. To understand why though, you have to look at a quick 101 of their business model. They operate in two segments. QCT which is semiconductor products at roughly 86% of revenue and then they have QTL which is their technology licensing business at roughly 14%. With QCT handsets is the largest at $6 billion last quarter powering premium Android smartphones globally and holding greater than 70% share in Samsung's Galaxy's flagship line. Then you have IoT covering industrial edge devices, consumer electronics, and augmented reality and virtual reality, which is generating 1.7 billion last quarter. But the real explosive growth here is how they're taking their massive mobile cash flow and attacking the automotive and the data center market. >> Yeah, I mean the the diversification approach sounds great on paper. Obviously, they're entering the enterprise uh data center space. That means fighting a multiffront war against really entrenched titans. So if we're talking about custom silicone, high-speed infrastructure for hyperscalers, companies like Broadcom and Marll absolutely dominate that landscape right now. You know, Qualcomm is a mobile native trying to break into enterprise cloud networks and hyperscalers require years of trusted hardware validation and Qualcomm doesn't really have that legacy track record in server stacks. I I wonder what your thoughts are on that. Yeah, and I I think that lack of record uh of track record, Rachel, is what's giving this uh massive margin of safety for Qualcomm. Right now, looking at forward earnings is trading at roughly 23. Uh so so not a crazy crazily overpriced stock. Qualcomm has fast-tracked their entire process by pairing their Orion CPU architecture and purpose-built AI inference accelerators with their critical connectivity IP they've gained through their recent acquisition of Alpha Wave. Now, this isn't a speculative bet for 2030 at the same time, right? CFO just confirmed on their recent earnings quarter 2 fiscal year 2026 that initial custom silicon shipments for a leading hyperscaler are locked in to begin in the upcoming December quarter of calendar 2026. The CFO confirmed this custom silicon engagement is expected to be accredited at the operating margin level from day one. Brocom has already [snorts] demonstrated the economics of custom AI silicon at scale and the new entrant with Qualcomm's CPU uh entering the same market is a credible threat that some might believe the street is barely pricing in. >> So if they even if they can execute on the custom silicon side and data centers, what about the rest of the business? I mean for example handsets are highly cyclical and breaking into new industrial verticals. It it takes a lot of capital, right? Completely agree Rachel and that is where the ultimate proof concept comes in. Their automotive business just crossed the 5 billion annualized run rate for the first time. 5 years ago automotive was barely a footnote in their financial model. Today is the fastest growing QCT product line hitting 1.3 billion last quarter and growing at 38% year-over-year. The company expects to exit fiscal year 2026 at a run rate above 6 billion and in the upcoming quarter 3 fiscal year 2026 automotive year-over-year growth is expected to accelerate further to approximately 50%. This is driven by their fourth generation Snapdragon digital chassis and integrated system handling the digital cockpit connectivity adas and automated driving. And by year end, their fifth generation Snapdragon is expected to ship, delivering three times higher CPU throughput and 12 times higher NPU performant than the current generation. Content per vehicle is set to jump sharply at the very moment ADAS adoption is inflecting. >> But why does uh you know vehicle NBU and CPU performance matter for a data center boom? I think that's maybe the biggest question I have. Because a gentic AI doesn't only require massive data sensors. Right Rachel? When AI agents runs in a car, the orchestration layer, scheduling task, fusing sensor data, and managing multi-step tool use runs locally on the CPU. Qualcomm has spent years building the world's highest performing CPU across the mobile platform and the automotive platform. And the arrival of a gentic workload converts that investment into a direct competitive advantage. I mean CEO Cristiano recently noted on the quarter 2 fiscal year 2026 call that the emergence of a gentic AI workloads with open claw as an early example are fundamentally changing users experience and reshaping their road map in every platform they develop. Today's install base was not built for these capabilities, representing a massive upgrade opportunity and expansion of their addressable market. And while they built this out, they have their QTL licensing business holding foundational patents on 3G, 4G, and 5G wireless technology. It collects royalty from nearly every smartphone sold globally, pulling a massive 72% EBIT margins last quarter. It is one of the most durable high margin revenue streams in the entire semiconductor industry, giving them a structural safety net to fund this R&D engine. Qualcomm is using its mobile cash flow to capture both the local endpoint of the aentic boom and the back end of server explosion. Speaking of which, Rachel, I know you brought a third stock that attacks this infrastructure expansion from an entirely different angle. What do you have? Well, you talked about AMD and Qualcomm, so I had to bring in Taiwan Semiconductor Manufacturing, ticker TSMC. Uh, look, we can sit here and debate whether, you know, AMD is going to take server share from Intel or whether Qualcomm can outpace traditional client architectures, but there is an undeniable reality. Regardless of who wins the chip design wars, everyone has to go to TSMC to actually get their silicon printed. So, they are the absolute vanguard of the entire advanced computing universe. Uh when Lisa Sue states on an earnings call that the server CPU market is doubling to over $120 billion because processing requirements are scaling faster than anticipated, that actually translates directly into a multi-year demand pipeline for TSMC's advanced foundry nodes. AMD's fifth and sixth generation Epic processors, Qualcomm Snapdragon platforms, Nvidia's accelerators. These are all manufactured on TSMC's 5nanometer, 3 nanometer, and upcoming 2nometer lines. So TSMC possesses absolute and really unrivaled still pricing power. And this is really because they're the only foundry on Earth capable of manufacturing these ultra dense architectures that commercial yields. And so they're able to pass the rising raw material and advanced packaging costs directly onto their customers. I mean this is a business that uh historically maintains gross margins well above 50%. >> Yeah, Rachel and nobody can deny that TSMC is a spectacular, beautifully managed business. They are the literal foundation of modern civilization. But when we talk about TSMC, we have to address the massive concentration risk under the hood. They have an extreme geographic concentration in Taiwan, leaving them exposed to immense geopolitical volatility. Because they are the sole advanced foundry for the entire world, they are forced to invest staggering amounts of capital expenditure every year just to build out clean rooms and buy expensive fab equipment. If we see a cyclical overupp of chips in the trailing half of the decade or if their new fat plants in Arizona and Europe face prolonged operational delays and higher than expected labor costs, won't those massive fixed depreciation charges absolutely crush their operating margins? >> So, you know, geopolitical anxiety is, I think, one of the key reasons why we've seen TSMC traded what I would call a discount relative to its its long-term growth story and power. I think that as long-term investors that can be a really uh intriguing entry window for the stock. But but let's look at the operational reality of that capex risk. So TSMC doesn't build multi-billion dollar fabrication facilities on a whim or a guess. They they build them based on locked in long-term capital commitments from hypers scale customers. These are Apple, you know, Nvidia, AMD, of course, which we've talked about today. And so their customers are really funding these capacity roadmaps in advance to secure their allocation slots. Now the the geographical risk you TSMC is actively executing a massive diversification strategy. They are ramping up their advanced packaging and production nodes across Arizona, Japan, Germany. Even if those international fabs carry higher initial construction costs, again going back to TSMC's pricing power, they really can charge a premium uh for that geographically diversified onshore silicon. And you know for me as an investor I I would say that the biggest takeaway is you can't scale Agentic AI or build 120 billion dollar server CPU market or run local edge devices without paying the TSMC tax. >> Agree Rachel and this is the end of the episode. I mean the AI revolution is expanding far beyond the initial hardware layers that the mainstream media has been covering for the last two years. As a gentic AI and localized inference scale up, the infrastructure demands are forcing a massive shifts of high performing general compute. We want to know what you think. Is Lisu right about the massive server CPU revival or is the market overestimating the compute demands of autonomous AI agents? Drop your thoughts in the comments below. Reread every single one of them. If you found this semiconductor breakdown valuable, smash the like button, subscribe to the channel for more research deep dives, and we will see you all in the very next video.