📄
AVGO Q4 2023 Earnings Call
December 7, 2023 at 12:00 AM
🔍
Company
AVGO
Quarter
Q4 2023
Date
December 7, 2023 at 12:00 AM
Speakers
2
Word Count
~5,849
Transcript Content
Operator
Our first question will come from Vivek Arya with Bank of America.
Speaker
Hock, so yesterday, one of your peers suggested that the market for AI accelerators could be as large as $400 billion. So kind of three related questions. What do you think about that number? And then number two, how does Broadcom participate in that, just beyond your large kind of ASIC project on the compute offload side? And then what does this larger AI accelerator market imply for your Ethernet networking business? I assume that they are correlated, but what is the right way to think about what is presumably a much larger market for accelerators and how it impacts Broadcom’s growth prospect? Hock Tan (CEO): Thank you for those interesting questions. Starting with the first part, we’re seeing a market that continues to grow and accelerate. It's also clear that it's very dynamic as the architectures of large language models and software models continue to evolve rapidly. We are observing changes in the requirements for compute silicon, which is both interesting and fascinating for us. This presents a significant opportunity. If a customer has a substantial business model and the resources to support it, it may make sense for them to design AI compute engines that incorporate both memory and the compute engine itself, customized for their specific application needs and LLM model. We are witnessing continuous changes in LLM models and the evolution of generative AI, where training and inference are beginning to converge and chip designs are adapting accordingly. This is evident in how we develop specific custom chips for hyperscalers, which is a compelling opportunity for us. As I mentioned earlier, we project that our networking revenue, which includes AI networks, will reach $4 billion and nearly double throughout 2024. We have communicated this before, and we are reinforcing that guidance. Networking is particularly noteworthy as it is growing in line with our AI accelerators and compute engines. We see this growth continuing, especially with the continuous training of very large language models with diverse and extensive parameters. There is no slowdown in the expansion of AI networks; in fact, we are observing an average doubling in the size of those networks across the board. To answer your question, I fully agree with AMD's assessment that demand appears to be accelerating rather than remaining stable or slowing down.
Operator
Thank you. One moment for our next question. And that will come from the line of Ross Seymore with Deutsche Bank.
Speaker
Just a quick clarification on the question. The clarification is, is the fiscal year guidance going to be the new protocol, or is that just this quarter? And then the real question, Hock, is on the VMware side of things, Kirsten talked about it potentially accelerating off of that $12 billion base. Can you just talk about the linearity of it maybe throughout the year, or more importantly, how is it going to be accelerating as people start to look at what the VMware Street estimates were before. We know we have to take out the two divested operations. But what are the drivers of acceleration? And how should we consider the magnitude of that as we look forward? Hock Tan (CEO): Yes. We are currently in an exciting situation as we transition into the next chapter of VMware. Our focus is on VMware Cloud Foundation, which is the complete software stack that virtualizes data centers on-premises and not entirely in a cloud environment. We are progressively converting more customers to this higher value stack on a subscription basis as they come up for renewal. While subscription models typically lead to a decrease in revenues due to revenue recognition practices, we anticipate a trajectory of accelerated growth through 2024. This growth is supported by the mathematics of our strategy, and to address your question, we are indeed accelerating from a $12 billion base and expect to see double-digit growth over the next three years, driven by the sale of this higher value virtualization stack compared to the less focused sales approach we had previously, especially in compute-only sales.
Speaker
And on the fiscal year side, is this a one quarter thing, or is this the new way you guys are going to be doing it? Hock Tan (CEO): That's a good question. To provide some insight, in 2024, since we are experiencing an accelerating trend, we believe it is more suitable to shift to annual guidance for that year. We will continue to report our results quarterly and will update our annual guidance for 2024 each time we share the quarterly results.
Operator
Thank you. One moment for our next question. And that will come from the line of Harlan Sur with JPMorgan.
Speaker
One quick housekeeping item question. So, even off the lower revenue base starting in fiscal ‘24 for VMware, is the team still targeting $8.5 billion-plus of EBITDA in three years? And then for my main question, Hock. As you mentioned, right, one of the fastest growing workloads in accelerated compute is accelerated compute and generative AI. And all of these workloads are increasing at an exponential rate. You talked about the benefits to your silicon franchise. But given the significant performance requirements of these workloads, right, training, inference, it appears that more of the near-term adoption of running these workloads is on bare metal, GPU, TPU, accelerated servers. So, how is the team exploiting a software-defined data center solutions via either cloud foundations or Tanzu to try to help customers focus on AI sort of drive better utilization, better economics, faster deployments on this very fast growing part of the market? Hock Tan (CEO): At the recent VM Explore in Las Vegas, VMware announced the VMware Private AI Cloud Foundation in partnership with NVIDIA. Essentially, the VMware Cloud Foundation Software Stack, which operates on NVIDIA coders and GPUs, forms the basis of this partnership. For enterprises, transitioning to generative AI analytics becomes straightforward since the on-premises data centers running VCF will inherently support the NVIDIA GPU software stack. This means it can effectively virtualize the NVIDIA GPU alongside the VMware software stack. This creates a compelling incentive for many enterprises to consider adopting the entire VCF setup. Not only does it enhance the resilience and manageability of on-premises data centers while reducing management costs, but it also allows for the immediate execution of AI workloads.
Speaker
And then, just on my first question, are you guys still targeting $8.5 billion of EBITDA in three years on VMware? Hock Tan (CEO): As Kirsten indicated, as we exit fiscal ‘24, we are practically at a run rate of $8.5 billion EBITDA.
Operator
One moment for our next question. And that will come from the line of Stacy Rasgon with Bernstein.
Speaker
Kirsten, considering your expectation of a 60% EBITDA margin for the Company this year, how should we interpret the initial and final EBITDA margin rates in relation to that yearly target? Additionally, I believe you mentioned that VMware's operating expenses would decrease by 40% by the end of the year compared to the start. I'm somewhat surprised it's not a larger reduction, possibly due to reinvestments. Is $1.4 billion per quarter the correct exit rate for VMware's operating expenses, and should we base our estimates on that? Kirsten Spears (CFO): That’s VMware spending. So that’s total spending. Hock Tan (CEO): Yes, it will be. But let me tell you, Stacy, you're missing the biggest point. As I indicated to an earlier question by Vivek, our revenue during this process, even over 12 months or 4 quarters, is on a growth trajectory due to the way the math works. As we sell more and recognize revenue on a ratable basis, our quarterly revenue is on a growth trajectory. That will continue and will extend beyond 2024. However, 2024 on its own won’t see a rapid increase in revenue. What you need to visualize is a revenue trajectory in 2024 that is expanding while operating expenses are declining due to reduced transition expenses. That's why we are telling you that by the end of fiscal 2024, we will essentially reach the guidance we provided when we announced this deal.
Speaker
So what’s the total company EBITDA margin, say, exiting the year then, just to level set? Hock Tan (CEO): Well, that’s a total we will get pretty close to where we are supposed to, before we started this whole exercise.
Speaker
What was that 65%? I can’t remember. Hock Tan (CEO): It’s somewhere between 60% and 65%. How does that sound, Stacy?
Speaker
I mean you did say 65% on it. I think I recall you saying you were going to run VMware at 65%. So I guess, 65% is the right exit rate? Hock Tan (CEO): At steady state, you’re right. At steady state, we’ll get to pretty close to 65% on VMware.
Operator
Thank you. One moment for our next question. And that will come from the line of Timothy Arcuri with UBS.
Speaker
Hock, in the language for the approvals from China, they noted some restrictive conditions and there were some protections around some sensitive information from your competitors. Can you detail what these are? And does this change your view on the synergies you can drive, either cost or more importantly, revenue? Thanks. Hock Tan (CEO): No. I think those conditions are pretty well laid out on the website of the relevant authorities. I frankly don’t think that it’s very appropriate for me to sit here and repeat all those conditions again. It’s right on the website, and that’s what it is.
Speaker
Okay. And it doesn’t make you think any differently about the synergies that you can drive from the business? Hock Tan (CEO): No.
Operator
Thank you. One moment for our next question. That will come from the line of Christopher Rolland with Susquehanna.
Speaker
Thank you for the question. Congratulations on the quarter and on closing the VMware deal. It seems that the cost of capital has increased since the announcement of the deal. Now that this is finalized, does this impact your approach to capital allocation moving forward? It appears that you have increased the dividend and restarted share repurchases now that the deal is closed. Will you be prioritizing repurchases more, or is Broadcom still focused on acquisitions as you work to reduce debt? Kirsten Spears (CFO): We are continuing our share repurchase program as promised, so we are definitely buying back shares. As Broadcom, we will quickly reduce our debt and keep all options in mind. Hock Tan (CEO): Chris, to elaborate on that, we financed part of the VMware acquisition with cash because it took longer than expected. This flexibility allows us to provide dividends to the new shareholders from VMware while also fulfilling our commitment to the shareholders to buy back $7 million of shares for the remainder of calendar 2023.
Speaker
Great. Thanks. And maybe a quick follow-up. Thoughts on just why you didn’t offer next quarter in favor of the full year? And if you had any thoughts on the shape of revenue for next year, whether it’s back half loaded significantly or pretty linear? Hock Tan (CEO): We recently acquired a significant company and are in the process of restructuring its business model to a subscription-based model, as mentioned by Kirsten. This change points to a clear growth trajectory for revenue driven by the transition to subscriptions and the introduction of higher-value products. Given these factors, it makes more sense for us to provide a full year forecast rather than a three-month outlook. Transition-related spending could fluctuate, and revenue may accelerate or slow down; thus, setting a three-month timeline for guidance might not be in the best interest of our shareholders. However, we feel much more confident in our ability to reach our targets by the end of the year.
Operator
And one moment for our next question. And that will come from the line of Toshiya Hari with Goldman Sachs.
Speaker
Hock, I had a question on the semiconductor business and specifically on the non-AI side of things for both networking and server storage connectivity. As you noted, you’re obviously going through a cyclical correction. Historically, you’ve had a pretty good understanding of where customer inventory is. And when we simply look at their balance sheets for the public companies, inventory is pretty elevated, particularly on the networking side. What is your interpretation of where inventory is for your products? And how should we think about the timing and pace of recovery as you look into 2024? Thank you. Hock Tan (CEO): On our books, you can see inventory for our products is pretty good, right, especially compared to our peers, and that’s because we keep it tight. Out in customers, and we don’t sell through channels, we don’t sell much through channels. We usually do a lot of it direct to our logic customers. We feel they are in good shape, relatively speaking. We are still in good shape. Now, if you ask me, maybe server storage, that could be a little excessive, but not broadband and certainly not in networking. So overall, on our products, we still feel rather good about it. And the best indication is the level of our own inventory on our own books. But what we do see is customers are perhaps much more cautious about buying more stuff, not just because they have too much of my inventory, I think because they have too much of everybody else’s inventory out there. We tend to see some caution in the way they choose to buy. Having said that, we’re still keeping to our lead times.
Speaker
And Hock, any comment on sort of the timing or the shape of the recovery in ‘24? Hock Tan (CEO): If only I know. I mean, I’ll be speculating to say second half of ‘24 things will start looking better compared to the first half of ‘24.
Operator
One moment for our next question. And that will come from the line of Karl Ackerman with BNP Paribas.
Speaker
Hock, I was hoping you may discuss the reason for divesting EUC and Carbon Black. And maybe more importantly, as you think about the growth rate off this $12 billion, Kirsten, could you discuss the opportunity you see in front of you as DRAM memory pooling brought in from the adoption of CXL within data centers that would seem to be a very big opportunity for VMware? Thank you. Hock Tan (CEO): What was the first question again? Sorry. Okay. Why did we choose to sell End-User Computing and Carbon Black? They are good assets, no doubt about it. They are stable and sustaining. The reason for their sale aligns with our strategy. We concentrate on acquisitions that deliver the most value to our business model and avoid distractions from noncore focus. For us, VMware is centered on core areas like data centers, core networks, and core computing. Therefore, we plan to invest in and direct our sales and R&D efforts toward those core areas of VMware Cloud Foundation. While End-User Computing and Carbon Black are good assets, we prefer to divest them now. We will find suitable new owners since there are many interested parties willing to take these assets. We will be careful about where we place these assets because many customers of these two segments are also customers of VMware Cloud Foundation.
Operator
Thank you. And one moment for our next question. And that will come from the line of Matt Ramsay with TD Cowen.
Speaker
Hock, I want to start by mentioning that I come from a semiconductor background rather than software. I'm curious about the strategy to transition VMware's customer base to subscription models and how this compares to your experiences with CA and Symantec. Do you think the conversion process will take a similar amount of time and achieve the same level of success, or are there notable differences in aspects such as the customer base, the duration of the long tail beyond the Fortune 1000, or the types of technologies involved? Are there any key similarities or differences in the approach that we should consider regarding the pace of this business conversion? Thank you. Hock Tan (CEO): No. These are very different assets. I'm not saying one is necessarily better than the other, just different. In CA, particularly where we have a presence in mainframe and some distributors, we focus heavily on our core customer base, which represented about 70% of our overall revenue at that time. We prioritize supporting these customers and provide them with excellent support along with growing feature requests in the area. We also focus on Symantec, which serves a small group of core customers. A significant aspect of CA's technology, especially in mainframe, is that it handles many legacy applications that remain very relevant today. Customers prefer to operate these applications using mainframe tools because modernizing or changing them does not make sense for their specific reasons. On the other hand, VMware offers a product that is both contemporary and forward-looking. It allows the creation of a virtualized cloud environment within a company’s own data center, making it suitable for any global business. Transitioning to a public cloud is fully virtualized and highly resilient when utilizing a software-defined environment. With VMware, we are providing a similar virtualization experience for on-premises data centers tailored for companies that already have workloads running on VMware products, specifically those built on VMware Cloud Foundation. This approach gives enterprises the chance to implement a hyperscaler within their own premises. That’s our straightforward plan.
Operator
Thank you. One moment for our next question. And that will come from the line of William Stein with Truist.
Speaker
Hock, in the past, I think we’ve all been aware that there’s one major customer on the accelerated compute side. I suspect that he’s broadened and deepened perhaps and hoping you can give us some characterization of that, maybe the number of customers or projects, how diversified it is at this point, that would really help. Thank you. Hock Tan (CEO): Yes, it has. This aligns with what I mentioned earlier about how merchant silicon will succeed. However, with the rapid evolution of AI, particularly in large language models and generative AI, it's clear that in hardware, one size does not fit all. Different models require tailored solutions. For some of the hyperscalers that have the necessary resources and scale to develop customized hardware that aligns with their foundational and application models, we are starting to observe the impact of that. Beyond this, I'd prefer not to share further details at this time as we are under a non-disclosure agreement.
Operator
Thank you. One moment for our next question. And that will come from the line of Harsh Kumar with Piper Sandler.
Speaker
Congratulations on closing the VMware deal, Hock. I know you've been working on that for a while. Your management has had a lot of time to evaluate this deal during the closing process. I'm curious about what you have found most satisfying so far and what you consider will be the biggest challenge in the integration over the next 12 months. Hock Tan (CEO): Over the past 18 months, we have been working towards closing this deal since the announcement. You are correct that we had a significant opportunity, and I want to express my appreciation for the supportive management team at VMware that has engaged with us exceptionally well. This has involved a lot of planning and contemplation. It has also afforded me the chance to connect personally with at least 150 CIOs globally from VMware's largest customers over the last 12 to 16 months. One clear takeaway is that the VMware core product, particularly the VMware Cloud Foundation software stack, which facilitates the virtualization of not just computing but also storage, networking, and the orchestration and management layer across the entire stack, is highly sought after by CIOs and heads of infrastructure at many large companies. They want the capability to deploy solutions that manage their diverse data centers, which currently feature a mix of virtualization, compute, and various hardware vendors, under a single abstraction layer. This approach not only reduces hardware expenses but also leads to significant cost savings in management. This illustrates the value of the technology VMware offers. The products are in place, and our focus is on execution, as highlighted in the numbers Kirsten has presented regarding the first year following the completion of this acquisition.
Operator
And that concludes today’s question-and-answer session. I would now like to turn the call back over to Ms. Ji Yoo for any closing remarks. Ji Yoo (Head of Investor Relations): Thank you, operator. In closing, Broadcom currently plans to report its earnings for the first quarter of fiscal ‘24 after close of market on Thursday, March 7, 2024. A public webcast of Broadcom’s earnings conference call will follow at 2:00 p.m. Pacific. That will conclude our earnings call today. Thank you all for joining. Operator, you may end the call.
Operator
Thank you all for participating. This concludes today’s program. You may now disconnect.
Need more earnings data?
Explore more earnings information and financial data for AVGO
Transcript sourced from Financial Modeling Prep (FMP)