Summary
Broadcom Inc.'s AI chip advancements and optimistic earnings guidance have propelled its market cap to over $1 trillion, challenging Nvidia's dominance.
Broadcom's ASIC AI chips offer potential cost and efficiency advantages over traditional GPUs, attracting significant interest from major tech companies.
Despite promising developments, Broadcom's stock appears overvalued post-rally, with a P/E of 36 and overbought technical indicators.
The future of AI hardware remains uncertain, with competition from Nvidia, AMD, and other players, making it prudent to hold rather than chase AVGO.
I am James Foord, an economist with a global approach to investing. I lead the investing group The Pragmatic Investor where the focus is on building a robust and truly diversified portfolios that will continually preserve and increase wealth.
vm
Thesis Summary
Broadcom Inc. (NASDAQ:AVGO) joined the trillion-dollar club after investors were impressed following the last earnings.
It seems that AVGO could make a real dent in the AI chip market through ASIC AI chips, which could present an advantage to traditional GPUs.
The market seems to have awarded Broadcom the title of runner-up in the AI market, and perhaps even sees it as a contender to Nvidia’s (NVDA) throne.
I last rated AVGO a sell, as I believe that the hype around AI was overstated. In light of the recent earnings, I can’t deny that there’s some reason to believe AVGO’s AI chips could, in fact, compete with Nvidia.
The stock still seems overvalued after the recent rally, but I am upgrading to a Hold.
Broadcom Pleases Investors
Broadcom’s earnings came in slightly above expectations, but it was really the optimistic guidance in the numbers and the positive sentiment expressed in the earnings call that catapulted the stock to over $1 trillion market cap.
AVGO in charts (SA)
As we can see, Semiconductor Solutions increased the most QoQ, showing a very encouraging change in trend. While Infrastructure Software was flat, it’s still up near 200% YoY, and EBITDA increased over 50% YoY.
AVGO Earnings Snapshot (SA)
For Q1, AVGO expects revenues of $14.6 Billion, with adjusted EBITDA at a very impressive 66% margin.
This increase in margin and growth will no doubt be led by increasing sales of Broadcom’s ASIC AI Chip, which it calls XPUs.
In 2027, we believe each of them plans to deploy 1 million XPU clusters across a single fabric. We expect this to represent an AI revenue Serviceable Addressable Market, or SAM, for XPUs and network in the range of $60 billion to $90 billion in fiscal 2027 alone
Source: Earnings Call.
$90 billion in potential revenues by 2027 is substantial, even in the fast-growing world of AI. Broadcom has also secured notable partnerships, including Apple (AAPL), as it was recently announced that they will be working together on a custom AI chip.
We could be seeing a very significant shift in the AI market, which would explain why we are seeing AVGO rise while Nvidia's stock languishes.
ASICs vs. GPU
It’s not the first time we see a stock pop with just the mention of “AI revenues.” However, this time it feels a bit different.
Broadcom rallied over 20% following their earnings and is up almost 10% as I write this on Monday the 16th of November. Meanwhile, Nvidia’s stock is down again.
Where others have failed, it seems like AVGO could now succeed, thanks to its “novel” approach.
Rather than competing head-to-head with NVIDIA, Broadcom is offering an alternative to its customers, the XPU, which is not a GPU, but more closely resembles an application-specific integrated circuit.
What’s the advantage of an ASIC?
As the name implies, an ASIC chip is designed for specific purposes, and this means that it can potentially carry out the same task as a general-purpose GPU more efficiently, meaning faster and at a lower cost.
For example, Edge’s Sohu chip created specifically for inference theoretically outperforms the Blackwell.
By focusing solely on transformer inference, the Sohu chip achieves remarkable efficiency and speed, processing over 500,000 tokens per second for models like Llama 70B - which they claim to be an order of magnitude faster and cheaper than even Nvidia’s next-generation Blackwell (B200) GPUs.
Source: Medium.
This is what Broadcom is attempting to do with the XPU and its latest line of offerings. Rather than selling “out-of-the-box” GPUs, it can work directly with customers to build specific solutions.
Why would big players in the AI space want to do this?
Potentially lower cost
Higher efficiency
Less dependence on Nvidia
More customization and control.
It’s Happened Before
Quite interestingly, we’ve already seen this movie play out before. A few years ago, we saw a shift in the crypto-mining industry from GPU to specialized chips, ASICs.
A lot of the reasons that applied back then to crypto, could apply today to AI.
ASIC VS GPU (ASIC marketplace)
The key thing to understand here is that, as AI applications and systems become more widespread and understood, it may begin to pay off to build custom chips. These offer more versatility and specialization capabilities.
Could this time be different?
However, investors should not be so hasty when it comes to determining an AI winner. Yes, AVGO has shown a lot of promise in the last quarter, and the fact that it is already likely working with a lot of the big tech companies on “alternatives” to Nvidia’s GPUs, is encouraging news.
But that doesn’t mean ASICs will be the future, and it also doesn’t mean Broadcom will be the clear winner.
The Ecosystem Matters
Based on the evidence laid out above, one could come to believe that Nvidia’s days are numbered, but that’s not necessarily true.
Even with ASICs AI chips, there are still plenty of reasons to favor Nvidia chips.
GPUs that can be used for many things can be more useful and less risky to buy
Nvidia has an established track record
Nvidia has CUDA and an established ecosystem.
This last point is perhaps the most important one. AI isn’t just about the hardware, it’s also about the software, and CUDA has been regarded for a long time as a key moat and differentiating factor.
More importantly, Nvidia GPUs and CUDA have become the standard, so that’s what people know and build on. There are definitely huge network externalities that we can’t overlook.
The Future of AI Is Uncertain
On top of this, the future of AI is still very uncertain. It could be GPUs or custom chips, but it could also be TPUs, or Neural Processing Units (NPUs).
It’s already hard enough to predict what the future of AI will look like, and even harder to predict who the winners will be. AVGO has shown an encouraging trend in their latest earnings, but competition from smaller players, and even established players like Advanced Micro Devices (AMD) and Nvidia itself, should not be overlooked.
Even if GPUs were to become obsolete, would we expect Nvidia to just sit and watch this happen with arms crossed?
Is AVGO A Buy Right Now?
Following the recent rally, Broadcom is certainly not cheap.
AVGO and Peers valuation (SA)
With a P/E of 36, it's equally expensive as Advanced Micro Devices. However, AVGO is quite expensive if we look at metrics like forward PEG. I pointed this out in my last article, and my view hasn’t really changed, but this may change if AVGO can deliver on its promise of more AI growth and higher margins.
EPS estimates (SA)
So far, this is not really reflected in the EPS estimates, which show that AVGO’s EPS will grow at around 20% CAGR over the coming years.
Furthermore, the stock looks quite overbought after the recent rally.
AVGO TA (Trendspider)
We can see a strong bearish divergence forming on the RSI, which has failed to make higher highs since July 2023.
Final Thoughts
I don’t think the market is necessarily wrong here. I am quite encouraged by AVGO’s latest earnings and think there are plenty of ways in which the company can capture more of the AI market. There are definitely some very logical reasons why big tech companies might favor Broadcom’s specialized AI chips in the near future.
However, following this massive rally, I don’t think it’s prudent to chase this stock. There’s still a lot we don’t know, and though AVGO shows potential, there’s a lot of competition out there.
Comments