Wall Street Traded A Rumor … I’m Buying The Sell Off

On Friday, shares of Broadcom finished down about 6%. AMD finished up about 6%. By the weekend, the explanation everybody kind of had settled on was that AMD was taking Broadcom's Google custom chip business. Google was starting to cut Broadcom out. But, here's the deal. None of that actually appears anywhere in the research note everyone was citing. On today's video, I'm going to walk you through what the semi-analysis report actually said. I actually was able to read it. Why the AMD paragraph doesn't mean what people think it does. Also, there's only one event in the next 60 days that actually matters for Broadcom, and it's not when the company reports their earnings in just a few weeks. And by the end of the video, you'll have a much better understanding of Broadcom's custom silicon business and why I think it's still one of the best-positioned companies during this entire AI build-out. So, let's start with the basics. Google has been designing their own custom data center chips for almost a decade now. The only other hyperscaler that is even close to them is Amazon. Google calls these custom chips TPUs. And while Google has a tremendous amount of resources and experience designing these chips, there's certain intellectual property and more importantly, manufacturing allocation with Taiwan Semiconductor that Google doesn't have. What many investors don't realize is manufacturing chips with Taiwan Semiconductor is I mean, it's a bit of a good old boys network type of thing. Part of it is Taiwanese culture, values, relationships, loyalty, and trust. And the bottom line, if you've been making chips with Taiwan Semiconductor for a long time. You get allocation to manufacture chips that exceeds that of other rivals. To go along with that, Broadcom has patents, it has intellectual property related to designing the chips, the networking that Google just simply doesn't have. So, it's a great partnership and one that has lasted nearly a decade. And that decade-long partnership is exactly what everybody decided was over all of a sudden on Friday. But, that's not actually what the semi-analysis note actually described. First, is that Google is expecting less chips from Broadcom in 2026 due to challenges ramping up chip on wafer substrate supply. Now, chip on wafer substrate or CoWoS, as I think maybe you can say it, is the packaging process at Taiwan Semiconductor. And look, it's complex. We don't need to get into it. You can search Google if you want to dive in. I've certainly have. Not only that, but Nvidia has been very aggressively buying the capacity for this chip on wafer substrate process for Taiwan Semiconductor for a few years now meaning there's just not that much to go around. But, the facts are Broadcom is cutting deliveries to Google because of supply. Broadcom simply can't get enough of the manufacturing process from the one company that can do it. Nobody at Google is canceling anything. Now, the semi-analysis note also reported that Google's TPU volume from Broadcom is also being cut next year, in 2027. That lower output is being driven by greater allocation to other customers. This is because Broadcom is shifting some of its manufacturing allocation it gets from TSMC to other customers like Meta, Anthropic, and OpenAI. So, Broadcom's 2027 Google TPU number came down because Broadcom is moving some of the capacity to other customers. It isn't losing any business. It doesn't have enough capacity to serve all the customers all at once. It would love to, trust me. So, it made a choice about who gets served first. When I began recommending Broadcom to clients back in 2023 when the shares were $80, one of the biggest criticisms was that they only had a single customer, and that was Google. It was actually a big risk. Flash forward a few years, and it's downright comical that the market would sell off Broadcom shares because it was diversifying its customers away from just Google. But, that's exactly what is happening, and it's a good thing, not a bad thing. Now, let's talk about AMD because that news also moved both stocks on Friday. I'll read you the exact excerpt from the SemiAnalysis note, and we'll come back on the other side and discuss what it means. Market chatter suggests Google is working with AMD on a TPU project in the version 10 generation. AMD's involvement would be the first real involvement in a custom AI ASIC project despite having a custom silicon team. AMD has strong IP, especially in advanced packaging. Additional CPU IP could be a big draw for Google given its customers are pushing for TPUs with on-package CPU cores. Now, let's unpack what this note means for AMD and Broadcom. First, SemiAnalysis cited market chatter. That's their words, not mine. In the world of finance, that is about the weakest possible way you can source something. It means somebody heard something. What are we in second grade, guys? Here's what I think it means. Somebody wants to move either AMD or Broadcom stock, so they float some chatter because they know semi-analysis will report on it about doing something, and they won't do any additional fact-checking. Now, in terms of Google using AMD because of advanced packaging and CPU and intellectual property, I actually think that's legitimately interesting. But, keep in mind, Google published on its own blog earlier this year that it integrated ARM-based CPU cores directly into its eighth-generation TPU system to remove what it called, I think it was called like a host bottleneck. Kind of in plain English, the CPU is managing the chip. It was too far away from the other processes, and that distance was costing Google some performance. So, in even kind of plainer terms, Google is already using CPU technology it's licensing from a rival or another company. Now, in the case of ARM, it's not far-fetched that Google wants to diversify away from them and maybe try out an x86 design, which is AMD and Intel's technology. But, here's my favorite part, and it's really the one that is most important for you, the investor. AMD buys Broadcom's Tomahawk networking switches to build its proprietary own AI rack. So, the company that is supposedly displacing Broadcom at Google is actually currently a Broadcom customer on the networking side of things and AMD doesn't have the technology to design Broadcom out anytime soon. So, let's be 100% clear about something. Broadcom has a dominant position in custom silicon and maybe even more importantly in networking. None of that is changing. One more thing worth noting. Google's 10th generation TPU has been reported on since June of this year. And the partners named in manufacturing it are MediaTek, Samsung, obviously Taiwan Semiconductor as well. So, if AMD does end up in the project, it's likely being added to something that already has multiple partners. AMD isn't taking anyone's seat or allocation at the table. Now, here's what I think actually matters to the story and nothing of it has to do with who's designing a chip. Look at who Broadcom signed as customers earlier this year. Google obviously in April extending a long-standing deal. Meta also in April with over a gigawatt to start in three gigawatts committed through 2028. Open AI 10 gigawatts of custom accelerators. Anthropic, which is already placed $21 billion worth of orders across two different quarters. And then finally, maybe the best customer of all, Apple in July needing over $30 billion of custom silicon. Here's the truth. That's the most customer diversity Broadcom has ever had in the history of its business. But here's the part almost nobody is talking about. In June, Broadcom set up financing with Apollo and Blackstone for $35 billion to specifically fund the AI labs. Broadcom is going to backstop these lease payments for about five years. So, Broadcom isn't just selling chips to these AI labs. It's helping arrange the money that they'll need to buy the chips. Nvidia has signed a ton of these deals as well. The industry calls it circular financing, but it is what it is. AMD has too, but here's the key thing to understand about AMD's financing. They actually had to give up big chunks of ownership of AMD to both OpenAI and Meta to secure those deals. Whereas Nvidia and Broadcom are just basically a co-signers on a loan. I'd much rather be Nvidia and Broadcom. Now, which kind of brings me to the event that I'm watching more closely over the next 60 days. Anthropic confidentially filed to go public on June 1st. The Financial Times just reported last week that investors are targeting a valuation around $2 trillion in dollars in an IPO that's coming in October, which would be the largest IPO in history. That means one of Broadcom's custom silicon customers is about to get a boatload of money. So, here's the bottom line. The Anthropic IPO and the commercial success of its models, which is undisputed, is the event that matters for Broadcom over the next 60 days. Not a paragraph in a glorified blog that starts with the words market chatter. Let's be real here, and we'll be real. Supply of AI semiconductors is so tight right now that even second-tier chips several generations old are holding value and can be sold. Custom silicon made in-house by AI labs along with Broadcom is sure as heck not getting tossed to the side in that environment. The reality for both Google and Broadcom is that custom silicon market has expanded so much over the last few years that Broadcom doesn't need to rely on Google nearly as much as they used to. And at the same time, Google has so much demand that they too can't simply rely on Broadcom alone to deliver the supply that Google needs. Both companies, Google and Broadcom, are doing the smart thing, diversifying and de-risking their business. If Semi Analysis writes something like that, it doesn't get any clicks. And if your business is selling newsletters, which is their business, clicks matter a lot more than actual facts do that make you, the investor, money. I realize we live in this headline-driven world where people see a tweet or read an excerpt and they think they know the full truth or they think they know the situation, but investing for the long term hasn't changed with social media. Buying stocks with durable modes, increased demand, customer diversity, and excellent management is a great way to grow your wealth. The way to throw it all away is by trading stocks based on Twitter headlines and posts without digging in and doing what they call due diligence. So, I hope this video helped provide some clarity, and I'm sure when Broadcom earnings report in a few weeks, CEO Hock Tan will get on the call, he'll be asked about this, and he'll have a few chuckles like he always do about the market chatter around the company, and he'll laugh it off and say, "Nope, we've got plenty of business." The bottom line is that Broadcom's business is booming, and I remain confident it's one of the premier semiconductor stocks to own through this entire AI build-out. That was today's video. We'll see you again soon. Good luck with your investments.