AVGO在估值上具有吸引力（前瞻市盈率21倍），但受客户集中风险限制，适合小仓位配置。

In today's video, we will talk about Broadcom, whose trading symbol is AVGO. This is a very popular stock, especially in the field of artificial intelligence, because it is the company that designs custom chips for companies like Google, Meta, Anthropic, and OpenAI. Therefore, it is a very important company that has experienced tremendous growth over the past few years .

However, Broadcom's stock has fallen by about 25% from its all-time high , and declined another 4% the day after it announced its earnings. I was a little late with this because I spent the whole day getting acquainted with Broadcom, its various businesses, its competitive advantages, its economic strongholds, and why investors are selling this stock so heavily despite what appears to be a great earnings report.

What I want to do is start by taking a look at their latest earnings report, the highlights, and the transcript, and then I've prepared a presentation to really delve into Broadcom's business so we can all understand it better.

So, let's now take a look at the key earnings results. Well, in this first snapshot, we can see that revenues reached $29.6 billion, an increase of 86% year-over-year. So, Broadcom has nearly doubled its revenue year-over-year, even though it now generates $30 billion in quarterly revenue.

They also achieved operating cash flow of $14.2 billion in a single quarter with capital expenditures of only $500 million. Therefore, they achieved $13.7 billion in free cash flow with a free cash flow margin of 46%. These are truly impressive figures.

They also gave us their guidance for the next quarter, and they expect revenues to reach $34.8 billion, a 93 % year-over-year increase. Therefore, Broadcom expects its revenues to continue accelerating in the next quarter.

Then, further down, they stated that their custom AI chip revenue grew by 221% year-over-year, and is expected to grow by 236% year-over-year in the next quarter. So, Broadcom's AI chip revenue growth rates also continue to accelerate, tripling, literally three times on an annual basis.

Therefore, Broadcom benefits greatly from large cloud companies building their own custom chips, such as Google's Tensor Processing Units (TPUs) . The following screenshot shows us a quick table of some of their other key metrics and growth rates. So, once again, revenue grew by 86%, operating income rose by 92%, profits rose by 95%, earnings per share rose by 96% , operating cash flow rose by 98%, free cash flow rose by 95%, semiconductor solutions revenue rose by 127%, and infrastructure software revenue rose by 29% across the board.

These are amazing growth rates, and the growth is not limited to revenues, but net profits are also growing tremendously. Once again, free cash flow nearly doubled on an annual basis. So, I want to quickly move on to Stock Unlock to show you what Broadcom's growth rates look like because this is truly incredible.

Based on the past twelve months, Broadcom's revenues have absolutely exploded . If we move to a quarterly basis, look at how much revenue growth rates have accelerated over the past five quarters. Revenues are growing exponentially at Broadcom as more giant companies and leading AI models seek to design and build their own custom chips. They are turning to Broadcom to do that. So, Broadcom is experiencing strong positive winds driving its business.

This is exactly the same story with the company's free cash flow. In just the last two quarters, free cash flow has nearly doubled from $8 billion quarterly to $14 billion now. Looking at the past twelve months, you can see that they have now generated $39 billion in free cash flow, and that free cash flow has begun to accelerate noticeably.

Okay, now let's return to Broadcom's earnings report, and these screenshots are taken from the company's conference call transcript. Now , I'm not going to read the whole text here, but if you want to pause the video to see the points I've highlighted, please do so , and I actually recommend doing so. But the takeaway from this snapshot is that Broadcom is providing custom chips to Google, Anthropic, OpenAI, and Meta, and expects continued accelerated growth from these businesses. And the revenues that Broadcom sees are expanding very rapidly.

In this next snippet of text, Broadcom says it expects 10 gigawatts of additional computing in fiscal year 2028. Anthropic is on track to become Broadcom's largest customer, and OpenAI will deploy 5 gigawatts in fiscal year 2028 to become the company's second-largest customer . In my opinion, this is exactly what the market didn't like about this earnings report.

Moving on to the next screenshot, Broadcom provides us with its guidance. They expect AI revenue to double in 2027 and then double again in 2028. They believe they can generate $230 billion in AI revenue in their fiscal year 2028. They also expect to exceed $30 in earnings per share by 2028. Again, they believe they can exceed that figure. $30 in earnings per share over the next two years.

And remember, Broadcom's stock price today is $360. So, if Broadcom manages to achieve that, today's price-to-earnings ratio, based on 2028 figures, is only 12. For a company of such high quality, the world's number one designer of custom chips, and which is seeing its revenues grow almost twice year-on-year, trading at around 12 times its 2028 earnings is remarkable.

It is clear that the market does not fully trust these guidelines, because if it did , I don't think the stock would be selling for $360 today. Again, this is one of the highest quality companies in the world, at the forefront of the AI revolution and the capital expenditure cycle, as we call it, and is selling for 12 times its 2028 earnings, and they believe they are already capable of exceeding $30 in earnings per share.

I mean, for example, if Broadcom stock simply trades at 20 times earnings in 2028 and actually achieves its $30 earnings per share target , its price will reach $600, which represents a return of about 66% over the next two years, or a compound annual growth rate of 29% compared to the stock price today. Again, this is just a stock trading at 20 times earnings and achieving their $30 per share target .

So, if Broadcom can achieve that, the stock looks surprisingly cheap here, and could deliver very strong returns over the next two years. Therefore, it is quite clear to me that the market does not trust these guidelines, and I will explain why.

Returning to the transcript of the conference call, one analyst compiled all the growth and concluded that 60% of AI revenue for 2027 would come from OpenAI and Anthropic. Then 75% of AI revenues for 2028 will also come from them .

What this means is that Broadcom's AI revenue growth and a very large portion of its revenue in this area come from OpenAI and Anthropic, and the market is not impressed with this or does not trust these guidelines because OpenAI is not a profitable company.

So, if a large portion of their revenue and projections come from a business that is not profitable to this day, how much confidence can you actually place in those numbers ? This is what the market is questioning, and this is why the market is skeptical about these forecasts and is not evaluating Broadcom based on what is stated in the text here .

And I think that's kind of what's happening to Broadcom here. Now, let's move on to the presentation I've prepared about Broadcom. By looking at the competitive features and the various revenue streams of this business, we'll be able to determine whether the stock represents an opportunity now or if it looks more risky and the market is right in its position.

So, let's start reviewing the presentation. In this first slide, we see Broadcom's various business units and the distribution of their revenues as of their most recent quarter. The primary source of revenue is the AI Semiconductor Unit, which comprises two actual sub-units . The first are custom accelerators or XPUs, which are chips that Google, Meta and OpenAI are building with Broadcom. The other unit is artificial intelligence networks. This unit creates artificial intelligence clusters containing tens of thousands of chips and networking equipment to help systems communicate with each other at extremely high speeds .

Then they have semiconductors that are not related to artificial intelligence. This unit consists of many different products. Things like wireless communication, such as Wi-Fi and radio, broadband, server storage, enterprise networks, and industrial chips that go into cars and machines.

Then we have infrastructure software , which may surprise some, but Broadcom is one of the world’s largest enterprise software companies, and it is completely separate from its chip business. It is a group of software companies that Broadcom has acquired over the years , and it is said to have a gross profit margin of 94%. These software companies operate in the fields of private cloud, mainframe computing, cybersecurity, and fiber optic networks .

The point here is that all of this is extremely important software related to cloud networks, and again, to cybersecurity. Then we can see the diversification of Broadcom's revenues, with AI accelerators accounting for 41% of revenues, AI networks 15%, non-AI semiconductors 14%, and infrastructure software 30% of revenues in the last quarter.

So, Broadcom's revenues are actually very diversified, and don't just come from giant corporations or custom chip building. This is the source of much of its accelerated growth, but Broadcom actually has a very solid and profitable business.

For example, companies like AMD, Nvidia, Marvell, and Broadcom all send their blueprints to TSM to actually manufacture the chip . Moving up the chain, we have Broadcom and Marvel, which are chip design companies. They work with companies such as Google, Meta, OpenAI and Anthropic to design chip blueprints that can then be sent to TSMC for manufacturing.

Broadcom is the leader in chip design by a wide margin, with Marvell coming in second by a long way. Then at the top, we have the chip buyers. So, it works like this. Chip buyers turn to Broadcom for their design. Then Broadcom goes to TSMC to manufacture it, and TSMC goes to ASML to get the machines needed to actually do it. This is how this series works.

My next logical question is: What is Broadcom's competitive advantage, and who can actually compete with this business?

Broadcom's main competitors are Marvel, MediaTek, and Allchip, and even Google and Meta are building their own design teams in-house. So, my question was: Can these competitors replicate what Broadcom is doing? The answer is that Marvel and Taiwanese companies can indeed design advanced silicon chips. So, yes, some of their competitors can replicate what Broadcom is already doing.

But what is difficult to replicate is the level of implementation. Broadcom has the fastest time to market without the need for redesign, i.e., without failed chips. In a world plagued by production bottlenecks, this is a tremendous advantage.

Then one of my other questions was: Can Google use competitors to get the same services? The answer is yes. Google already uses MediaTek in parallel with Broadcom to build Tensor Processing Units (TPUs). So, in terms of simple design, there are other companies that can do what Broadcom does.

So, my next question was, what motivates companies to stick with Broadcom when there are similar companies that can do the same job? Well, the first reason is the costs of moving. Each processing unit is a deep, shared engineering relationship that spans years. Changing the program in the middle or even changing it completely means losing a lot of gains and time, which ultimately means money. Therefore, it is not really in the company's best interest to move from Broadcom.

Then there is the factor of time required to reach the market. Broadcom gets chips to market faster, which also gives customers a competitive advantage. There are no redesign processes , which means it also has a cost advantage as there is less waste and less damaged chips that need to be remanufactured.

Then we have the fact that Broadcom has a comprehensive solution. Customers can turn to other companies, but no company has the full range of what Broadcom does or owns. Adhering to Broadcom simplifies supply chains.

Finally, we have the bundle system ( packages). Broadcom also owns networking equipment and infrastructure as part of its offerings. Other suppliers do not have this ability to assemble. Therefore, switching from Broadcom means replacing with multiple suppliers, not just one , which is clearly less efficient.

So, my summary is that Broadcom does not have a monopoly on its chip designs, and companies can replicate what it does in this area. So, Broadcom's competitive advantage comes from its cost efficiency, size, market leadership and brand position, and being a one-stop destination for services.

Now, let's move on to the risks I was able to identify based on what other investors are saying and why they are avoiding the stock.

So, the first is simply high multiples. The company is currently trading at approximately 40 times its free cash flow and earnings for the past 12 months. This is a high multiplier for any business, and it means that it is based on expectations of extremely rapid and sustained growth.

Then we have customer focus, and in my opinion, this is the most important factor. Investors are concerned about the concentration of Broadcom's revenue guidance. Once again, it was noted that 60% of AI revenue for fiscal year 2027 and 75% of AI revenue for fiscal year 2028 will come from OpenAI and Anthropic. This also means that approximately 58% of the company’s revenue guidance for fiscal year 2028 will come solely from its leading AI labs, meaning that more than half of the company’s total revenue will come from OpenAI and Anthropic just two years from now. Investors really don't like this .

Then we have implementation risks. The directives are somewhat outside Broadcom's control, as chip supplies are constrained by how quickly land and power can be secured, and how quickly data centers can ultimately be operational . If any substantial slowdown occurs, it could affect the guidance and Broadcom could end up not meeting it.

Then we have the AI capital expenditure cycle , which is the point that pessimists have been talking about for a long time, that this expenditure cycle will eventually end . Cloud computing giants will eventually reduce their spending . Once that happens, the foundations of the AI business will deteriorate, growth will slow , and complications will shrink considerably. Investors do not want to remain in these businesses when that

Then we have circular funding, because in the Broadcom call, they were asked about funding concerns for OpenAI and Anthropic. Simply put, Broadcom said it believes these companies will go public and receive funding from third parties, including Broadcom, meaning Broadcom could end up funding its customers, which is not to the liking of many investors.

Essentially, everything you see on your screen right now is making investors wary of the long-term direction and profit potential of this business. In other words, the guidance does not have much confidence in investors, and for this reason, the stock is not priced appropriately compared to the guidance, at least in my opinion.

In my opinion, looking at the metrics of the past 12 months for a company that is growing at 90% annually gives you the wrong picture. You are starting from the wrong point. You need to look at future metrics when a company is growing at this rate. On a forward-looking basis, Broadcom shares are trading at approximately 21 times earnings and free cash flow, making them appear fairly priced relative to their quality and growth.

Regarding customer concentration, Broadcom said it still suffers from supply constraints, not demand. This means that if something affects OpenAI's ability to pay, the demand gap will be filled by another client such as Meta or Google. Demand currently exceeds supply for all Broadcom customers, and this situation appears likely to continue for a long time.

The answer to this is that the risks have already been taken into account within the guidelines. In the conference call, the CEO said they had lowered their estimates from the 30 gigawatts of initial demand they had previously cited, because they do not expect to deploy it in full in a timely manner. The current revenue forecast came after these figures were revised downwards and after discussing what customers see as realistic regarding the data centers coming online.

Moving towards the end of the AI capital expenditure cycle , a large part of the new demand comes from inference, which is proportional to the use and adoption of AI, rather than from training new models. Companies are now focusing on reducing the cost of using artificial intelligence to increase adoption rates, and this approach is already paying off . As I mentioned in the last video, the consumption of AI tokens has increased 25-fold year-on-year, and doubled twice month-on-month. The conclusion is the main driving force behind it. In other words, if the adoption and use of artificial intelligence continues to grow, funding should continue.

Moving on to circular financing, Broadcom is not the primary lender, and supplier financing is actually a normal business practice . There are usually several guarantees, warranties, or even shares associated with doing so as well.

Finally, we have the credibility of long-term forecasts. The answer here lies in Broadcom's long history of reservations. Broadcom has exceeded its own expectations and raised them for ten consecutive quarters. These projections are also based on multi- year agreements and not just hopes.

First, Broadcom is clearly the market leader and the biggest beneficiary of custom chip design. I believe it will continue to grow and will remain in high demand. Secondly, I believe the AI capital expenditure cycle will continue until at least the end of this decade, with custom chip design also receiving a larger share of capital expenditure. Broadcom doesn't look expensive based on future earnings projections. In fact, I think it looks very cheap if its predictions can actually be met. Again, it is trading at a forward price-to-earnings ratio of 21 times.

Fourth, the company has a history of being conservative in its forecasts, which leaves room for better performance and exceeding those expectations. The CEO suggested that they are repeating this approach again with their $30 earnings per share forecast for 2028.

Fifth, I believe there are risks associated with OpenAI and Anthropic, but I also believe that other clients will fill the gap if these companies fail to meet their obligations.

Finally , I believe that Broadcom will continue to grow rapidly. It is a diversified business and is transforming into one of the most important companies in the world . With a forward price-to-earnings ratio of 21, I think the stock looks attractive and I'm considering starting an investment position here.

To give you more of my perspective on why I think Broadcom might look interesting, let's take a look at its historical forward price-to-earnings ratio, which currently stands at around 22.17. During the release day in 2025, the stock reached its lowest point at a forward price-to-earnings ratio of around 20, which is slightly lower than its current price, but still within the same price multiple . Back in November 2023, the stock was trading at a forward price-to-earnings ratio of around 22, which is again the same multiple at which it is trading today.

Now, let's look at Broadcom's performance since November 2023. The price was around $94 at that time, and the stock has risen by about 276% since then. Looking at the returns since April 2025, the stock has risen by 144%. These returns in share price come entirely from the massive growth in the company's profits. Because these were the last times the stock traded at around 22 times forward earnings. Therefore, all of Broadcom's stock price growth has been a result of the tremendous growth in its business fundamentals.

What this tells me as a potential investor is that if Broadcom can increase its earnings, and double them another two years later, the stock may look very cheap now and another buying opportunity.

What really worries me, and is the primary concern I found, is that 58% of its revenue projections for 2028 are expected to come from my companies OpenAI and Anthropic. This is a very high customer focus, and it really means that these two companies have to deliver the required performance and meet their spending commitments to Broadcom in order for the latter to see that growth.

I also have a question about what Broadcom will be like by 2030? If OpenAI’s IPO does not happen and its funding stops, how sustainable will its cash flows be until 2030? This is another question that keeps going through my mind, and I don't have a real answer to it.

So, if I were to invest in Broadcom, for me, given that I don't have a strong long-term view of cash flows , I probably wouldn't make it a large share of my portfolio, but I would add it to gain exposure to the semiconductor sector.

I think it's a very distinguished and high-quality company , but its revenue focus and long-term outlook seem a little unclear to me.