Broadcom has greatly outperformed the FAANGs over the past decade yet is rarely discussed as one of the market’s top-performing stocks. The ticker certainly does not participate in a catchy acronym. Half the battle with Broadcom is there are many revenue segments to analyze, some go through harsh cyclical downturns, and it acquires companies hand over fist. To put it plainly, this is not an easy stock to cover; there are no pithy ways to summarize the products and it doesn’t offer growth stock qualities.

Broadcom is reporting the second highest AI revenue on the stock market today, accounting for 16.1% at $1.5 billion, up from the July quarter at $1 billion and 11.3% of revenue. Of the smaller players, peers like AMD have guided for AI revenue of $3.5 billion in 2024, which would account for about 13.6% of the 2024 estimated revenue. Marvel expects AI revenue of $400 million in 2023 or 7% of revenue and is rising to $800 million or 13% of the expected 2024 revenue. Therefore, Broadcom is in second place, but notably, does not have a large lead by percentage of revenue. Juniper at 23.5% of revenue has a higher percentage yet is being acquired by HPE.

Ultimately, our goal is to get this stock lower, but to put into motion now the in-depth research required for a potential entry. As stated in the 2023 Year in Review webinar, Meta was the one that got away given its bottom line fits our criteria, but I consider Broadcom the one sitting in plain sight.

Broadcom has many departments that have been strung together through acquisitions with a vision of consolidating Ethernet, ASICs and now virtualization under one company. The reason this company is sitting in plain sight is because it was a major winner from the mobile era, is partnered with Big Tech as we move in the AI era, and is putting together the pieces to participate in AI strategically by not needing to compete on GPUs.

Broadcom’s Switch Products and Switch Fabric

Broadcom has a dominant market share of switching and routing semiconductors for hyperscalers and is seeking to maintain its market share, most especially as AI changes networking requirements. The Jericho3-AI launch was last April, which is a redesign intended to compete with Nvidia’s InfiniBand.

Broadcom has three switch products. The Tomahawk is the high-bandwidth switch platform, Trident is the platform with more features, and the new Jericho line combines the Jericho switch with routing ASICs. The Jerico3 was redesigned with deep packet buffers. Tomahawk and Trident are used in data centers yet are not optimized for AI workloads, especially when compared to InfiniBand.

Jericho has 160 switch ports dedicated to switch fabric, which allows multiple ASICs to be stitched together to support GPU clusters. The asymmetric split helps the chip overcome network congestion and network failures. According to Broadcom, Jericho3-AI performed 10% better than “alternative network solutions” — which is a clear reference to InfiniBand.

A few specs before we go deeper into how Broadcom’s solutions compare to Nvidia’s. As you’ll note, the specs for InfiniBand are superior with support for 64X 400Gbp or 128 200 GB/s ports compared to Jericho’s 36X 400 GbE and 72X 200GbE network-facing ports.

Jericho has 144X SerDes lanes and 106Gpbs PAM4 supporting 18X 800Gbe, 36X 400 GbE and 72X 200GbE network-facing ports. The Jericho3-AI allows for more than 32,000 GPUs to be linked for a massive AI training system and links directly to GPUs without the need for a server bus.
Tomahawk5 runs at 100 GB/second with PAM4 with aggregate bandwidth of 51.2 Tb/s
Compare this to Nvidia’s Quantum-2 InfiniBand which has support for 64X 400Gbp or 128 200 GB/s ports with 51.2 Tb/s and 66.5 billion packets per second.
Nvidia’s new Spectrum X Platform is an Ethernet solution that delivers 1.6X better networking performance than traditional Ethernet with 256X 200 Gbe ports or 16,000 ports for larger training systems.

On a similar note, and while we are on the topic, Marvell has some skin in the game too by offering a 51.2T switch called Teralynx 10 that offers ultra-low-latency at 80% cost savings. According to Moor Insight Strategy, Marvell is the supplier for AWS for “electro-optics, networking, security, storage, and custom-designed solutions.”

Marvell’s Nova 1.6 Tbps PAM4 electro-optics have eight 200G lanes that “double the networking bandwidth while reducing power and cost per bit by 30%.” According to the press release, “by doubling the bandwidth per lambda, the Nova-based modules reduce the number of lasers and related optical components by 50%.”

Marvell hopes to help data centers transition to 51.2 Tbps networking architectures by offering a platform that needs 32 optical modules instead of 64 optical modules.

Here’s a a quick glance on the rankings for AI networking revenue (approx. revenue)

Nvidia in first place with $2.5B per quarter from InfiniBand
Broadcom in second place with $1.5B from AI, primarily networking
Marvell at $200M per quarter from AI, primarily networking
Juniper Networks reported $321.2 million in the AI enterprise segment, or 23.5% of revenue. Recently, it was announced that Juniper is being acquired by HPE.

Further out in FY2025 and/or CY2025:

Cisco is expecting $1 billion in orders from a recent Nvidia partnership, per the earnings call: “our expectation is the majority of that $1 billion in orders will turn into revenue in our fiscal '25, just to be clear.”
Arista is expecting $750 million by 2025: “We are cautiously optimistic about achieving our AI revenue goal of at least $750 million in AI networking in 2025.”

AI Networking

We’ve covered networking as it relates to data centers, 5G, cloud applications and enterprises when we wrote about Nvidia’s acquisition of Mellanox in 2020 and Marvell’s acquisition of Inphi.

A few years back, we discussed that Nvidia acquired Mellanox for the strategic synergy that InfiniBand and Ethernet can provide in boosting GPU performance. Without proper interconnection, GPU performance could be limited, and so Nvidia strategically wanted to create the best-case scenario of owning both markets for AI accelerators — and their fabric and interconnects.

Mellanox supports Virtual Protocol Interconnect (VPI), which allows the ubiquitous Ethernet to provide bandwidth as cheap as possible, and InfiniBand to deliver higher throughput and fewer bottlenecks during high loads. In 2019, the split in Mellanox’s revenue was about 40% InfiniBand and 60% Ethernet. By leveraging a hybrid of Ethernet and InfiniBand, Mellanox was able to take market share from Ethernet incumbents.

The acquisition went under review in China, with officials believing Mellanox’s market share at the time was about 55% to 60% of the global interconnect market and 80% to 85% of the Chinese market. This illustrates how popular Mellanox was before Nvidia acquired the company.

The outcome of the review was that Nvidia was required to decouple the sale of InfiniBand from the sale of its GPUs to where a customer could buy one but not the other at no penalty. Even if a customer can buy them separately, there are many cases where it’s not practical to do so, such as with the DGX and HGX systems which achieve optimal performance with the A100s/H100s and InfiniBand.

Nvidia has stated that InfiniBand increase the effectiveness of AI infrastructure by 20% to 30%. Remote Direct Memory Access (RDMA) reduces CPU overhead by offloading data movement to network adaptors. In addition, Ethernet has quality of service (QoS) flow control and advanced error handling mechanisms that increase its network efficiency capabilities.

In the last earnings report, Nvidia stated in the opening remarks that “Networking now exceeds a $10 billion annualized revenue run-rate. Strong growth was driven by exceptional demand for InfiniBand, which grew fivefold year-on-year […] Azure uses over 29,000 miles of InfiniBand tabling, enough to circle the globe.”

The software defined fabric is popular for its low latency as its architecture reduces packet loss, high bandwidth and low management costs. The high-speed data transfer has link speeds of 10 to 400 gigabits per second due to its low overhead and efficient transport protocols, which is why InfiniBand is adopted by supercomputers and also AI/Big Data applications with high performance clusters. Due to very low latency, InfiniBand delivers real-time data transfer.

The $10 billion in annualized revenue run-rate reported by Nvidia in the Q3 October quarter represented 500% growth, which is nearly double the growth rate of the overall data center. Per the Q3 earnings call: “Networking now exceeds a $10 billion annualized revenue run-rate. Strong growth was driven by exceptional demand for InfiniBand, which grew fivefold year-on-year [..]” At the time, the data center had a run rate of $60 billion, so networking was 16.6%. In Nvidia’s most recent quarter, networking grew 155%.

According to Del’Oro, a research company out of the UK, AI systems account for less than 10 percent of the total addressable market for network switching, and of that, 90 percent are using Nvidia/Mellanox InifiniBand due to InfiniBand reducing packet loss, which is ideal for AI training workloads.

If right now, you’re thinking: “I thought this analysis was about Broadcom, not Nvidia!?” then that’s a fair question. Given AI networking is heating up across the board (Nvidia’s run rate, Broadcom, Juniper/HPE, Marvell, Cisco potentially, Arista), it’s important we touch on why Infiniband owns 90% of the AI market right now. We are overdue on revisiting Mellanox/InfiniBand as it’s been four years since we covered the acquisition. This also helps frame how Broadcom intends to compete with Ethernet.

Ethernet Vs InfiniBand

I wish I could make networking more conversational, but it’s pretty challenging to do that! Here are some bullet points on how the two compare; I’ve bolded the more important takeaways:

Benefits of Ethernet:

Raw bandwidth is a benefit with Ethernet hitting 51.2 Tb/s two years ago with support for 800 Gb/s port speeds. InfiniBand lags by topping out at 51.2 TB/s with 400 Gb/s port speeds. Although typical server nodes do no need the extra bandwidth, AI clusters come with 400 Gb/s NIC per GPU with some nodes having four to eight GPUs. By 2025, Dell’Oro believes switch ports for AI networks will be operating at 800 Gb/s and will further double to 1600 GB/s by 2027. Dell’Oro believes switch ports for AI networks will be operating at 800 Gb/s and will further double to 1600 GB/s by 2027.
smartNICs and AI-optimized switch ASICs help to reduce packet loss
Large pool of vendors whereas InfiniBand increases dependency on Nvidia. For this reason, AWS and Google Cloud have remained on Ethernet as they prioritize custom silicon.
Ethernet is the incumbent networking standard and most cloud providers have invested heavily here already. With Ethernet, providers don’t have to manage a new network stack.
Ethernet switching has evolved to where a new term has been coined “lossless” Ethernet. Even Nvidia is moving in this direction with their Spectrum X platform, due out this year.

Benefits of InfiniBand:

Outperforms for AI/ML workloads due to low latency and by reducing packet loss. Data packets are sent in a serial approach so multiple channels of data can be sent simultaneously. This is much better for AI/ML than a parallel approach for internal data flow, which creates bottlenecks.
Has 3X to 4X lower latency than traditional Ethernet switches based on ASICs
Highly scalable, can support tens of thousands of nodes per subnet. InfiniBand is also cheaper as it requires fewer connections for reliability.
Its QoS and failover capabilities are a reason it’s adopted for high-performance computing environments.
Reduces CPU resources

So, why is Ethernet Making a Comeback?

Broadcom’s Jericho3-AI has some promising benchmarks that could help shift the dominant market share InfiniBand has in AI networking (or at least prevent a monopoly). These benchmarks showed the Jericho3-AI outperforming InfiniBand by 10%, which is substantial when dealing with AI systems as it’s enough to increase the collective operations of the system.

“Leveraging this unique functionality, the Jericho3-AI fabric provides at least 10 percent shorter job completion times versus alternative networking solutions for key AI benchmarks such as All-to-All. This performance improvement has a multiplicative effect on decreasing the cost of running AI workloads since it implies that expensive AI accelerators are used 10 percent more efficiently. The network, in effect, pays for itself.” — You can read the press release here.

This means bare metal can work more effectively. Per Broadcom: “because it can handle 800Gbps port speed (for PCIe Gen6 servers) and more, it is a better choice [than InfiniBand.” At high price points, all hyperscalers want to see their investments working at maximum clock times. This is achieved by better load balancing and congestion control to improve network latency, whereas InfiniBand reduces port and hop latency inside the switch. Broadcom calls their product differentiation “Perfect Load Balancing” and “Congestion-Free Operation.”

A note on PCIe

The maximum bandwidth supported by PCIe 5.0 is 400Gbps per port. By using 106Gbps PAM4 SerDes, ASICs can be tuned to support 100, 200 and 400 Gbps port speeds. To work around this, and to achieve 800Gbps, chip makers are building NICs directly into the accelerator. According to The Register, the 800Gbps ports built into accelerators may reduce bottlenecks before PCIe 6.0 arrives on the market. The Register, the 800Gbps ports built into accelerators may reduce bottlenecks before PCIe 6.0 arrives on the market. This is what Broadcom is referring to.

Jericho3-AI supports 36 ports at 400 Gbps speed, and this can support Nvidia’s powerful DGX H100, which have 8 ports of 400 Gbps speed. In this case, four node racks are within Jericho’s capabilities. However, the Quantum-2 InfiniBand can handle 64 ports of 400Gbps, and so for Nvidia’s GPUs, it outperforms. Broadcom’s answer to this is that AWS and Google still prefer to not have vendor lock-in with Nvidia and use the Jericho3-AI to make use of their extensive Ethernet systems.

Overview of Custom Silicon (ASICs):

In addition to AI networking, Broadcom participates in the custom silicon market. ASICs are application-specific integrated circuits that are customized to perform a specific function for a specific application, hence the term “application-specific.” This is in contrast to GPUs which are more general-purpose. ASICs are expensive at the onset, yet become cheaper with volume production. We first published this graph in 2019 but the comparison still applies:

For now, custom silicon only makes sense for a company with deep coffers that has immensely popular applications – such as Google, Meta, Amazon. These companies use custom silicon to drive down costs on GPUs for their most popular applications. Across ASICs, the most well-known is Google’s tensor processing unit (TPU).

Google was one of the first to require low-power machine learning workloads for Search, YouTube and Google Maps. The compute intensive workloads were running on Nvidia’s GPUs for both training and inferencing until Google made their own processing unit, TPUs, to perform workloads at a lower cost and higher performance.

Performance between TPUs and GPUs is often debated depending on the current release (A100 versus fourth-generation TPU, for example). In some cases, TPUs have better performance per watt for power-constrained applications. Notably, some of this comes with the territory of being an ASIC, which is designed to do one specific application very well whereas GPUs can be programmed as a more general-purpose accelerator. In this case, the benchmarks where TPUs compete are object detection, image classification, natural language processing and machine translation — all areas where Google’s product portfolio of Search, YouTube, AI assistants, and Google Maps, for example, excels.

Notably, TPUs are used internally at Google to help drive down the costs and capex of its own AI and ML portfolio and they are also available to users of Google’s AI cloud services. For example, eBay adopted TPUs to build a machine learning solution that could recognize millions of product images.

Unless Google releases an internal technology as open-source, it won’t be adopted by the competitors. This is where Nvidia’s neutral position as traditionally a hardware company becomes a positive as it’s universally used by Amazon, Microsoft, Google — — and Alibaba, Baidu, Tencent, IBM and Oracle. Meanwhile, TPUs create vendor lock in (with a direct competitor) which most companies want to avoid. eBay is the exception here as the company needs Google-level object detection and image classification.

Why ASICs are Not a Near-Term Threat to GPUs

AI investors will need to get comfortable hearing about the battle between ASICs and GPUs. This debate has been going on since at least 2018, when Nvidia’s biggest threat was thought to be Google’s custom TPUs. There is some merit to these concerns as the largest customers for Nvidia’s GPUs have enough cash to make custom chips. There’s roughly $35B to $40B per Big Tech company per year that executives will naturally want to optimize to drive down costs.

To program ASICs is difficult, and they are application-specific, which means they cannot be reconfigured. Nvidia is wildly popular because GPUs are easy to program and are the best choice for a wide range of applications. Developers create the moat, which was our original Nvidia thesis. Therefore, I don’t believe there is much risk that Big Tech commercializes AI accelerators.

However, it’s quite plausible that someday Big Tech will reallocate capex toward more ASICs and fewer GPUs to where it will impact Nvidia. For now, demand outstrips supply, and there are long lead times for Nvidia’s GPUs. If a company like Google reallocates to more TPUs, another enterprise will certainly step up to fill those orders.

Broadcom & Google Partnership

It was confirmed last year that Google is a customer of Broadcom for its ASICs (TPUs). This was officially reported when The Information wrote an article stating Google wanted to ditch Broadcom in 2027, which Google has since denied:

“We are productively engaged with Broadcom and multiple other suppliers for the long term. Our work to meet our internal and external cloud needs benefit from our collaboration with Broadcom; they have been an excellent partner, and we see no change in our engagement." -Google’s response to The InformationGoogle’s response to The Information

Prior to this, it was never directly stated that Google was Broadcom’s main ASICs customer. Here is how Broadcom discussed it: “As you know, we supply a major hyperscale customer with custom AI compute engines. We are also supplying several hyperscalers a portfolio of networking technologies as they scale up and scale out their AI clusters within their datacenter.”

The following has also been stated about the Google-Broadcom relationship: “Broadcom supplies wireless chips for Google phones as well as chips for its data center and cloud services. At the same time, Broadcom is one of Google Cloud’s biggest customers for its cloud products. This bidirectional relationship has also forged a special bond." Per the same report, Meta is also working with Broadcom on ASICs, although does not deploy many of these “yet”

Last April, Broadcom migrated its infrastructure from AWS over to Google Cloud. Per the announcement: “Broadcom, a provider of enterprise security solutions, recently worked with Google Cloud Consulting to migrate its infrastructure from Amazon Web Services (AWS), and found the combination of technology and expertise critical for success. “Google's deep technical skills and its data, security and AI offerings have accelerated our transformation towards becoming a software-led company,” said Andy Nallappan, Vice President, CTO and CSO, Broadcom.

VMWare –Software-Defined Networks and Data Centers

In November, Broadcom closed its acquisition of VMware for $69 billion. VMWare is virtualization software that virtualizes compute and data centers. The software creates an abstraction layer, or a “hypervisor” which is the technical term for a computer or server that runs virtual machines called ESX. VMWare was the first company to virtualize x86 machines and was founded in 1998. Operating systems, such as Linux, Windows and MacOS, can share the same, virtualized resources by running on a x86 machine.

Over the past decade, VMware pivoted to offer a software-defined data center. On the most recent earnings call, Broadcom discussed building essentially a private cloud on-premise, which has some advantages, such as lowering capex by pooling memory, security, networking and server resources. “Our strategy going forward is simply to enable global enterprises to run their applications across the other data centers as well as on public clouds by consuming VMware’s higher-value software stack.”

Later it was stated: “We are creating with VMware, the same experience of virtualization of the data center on-prem for those companies, which has workloads, by the way, that are already running VMware products that application that’s already written on VMware Cloud Foundation. This is then giving these enterprises the opportunity to have a hyperscaler on-prem. That’s the plan we’re doing, plain and simple.”

Where software defined networks have seen quite a bit of success is with 5G networks. SDNs separate the control plane on embedded switching systems. This allows the networks to be managed remotely. Products from different suppliers can be used without incompatibility issues. By using open APIs, the 5G market has benefited from a more neutral ecosystem by allowing products from different suppliers. This is because SDNs allow network functions to be programmed by APIs instead of proprietary interfaces.

In 2012, VMWare acquired Nicera to create VMware NSX, virtual networking and security software that virtualizes network components. The NSX products, including NSX-T data center, programmatically creates and manages virtual networks from Layer 2 to Layer 7, which is defined as switching, routing, access control, firewall and QoS. NSX Manager and transport nodes can be assembled in seconds for proof-of-concept deployments, deployments with up to 64 hosts, or large-scale environments. Software-defined networks have a natural synergy with Broadcom, the leader in networking hardware.

A few years ago, VMWare expanded to virtualize containerized workloads for Kubernetes clusters. This product is referred to as Tanzu. This was a necessary evolution to keep cloud native customers. Many Kubernetes clusters use something called a multi-tenancy solution, which is to have non-connected “tenants” use a common pool of resources. This can be hard to implement correctly, and also has limited functionality once it’s set up. Virtualized containers are similar to a single-tenancy solution by having its own API server, controller manager and storage for data. Yet, it’s similar to a multi-tenancy solution by using a common pool of resources. Per the Broadcom earnings call: “And to attract and keep these workloads across the environment, we are investing in a rich catalog of microservices tools. This will be our focus. And the noncore businesses of end-user computing and Carbon Black will be divested.”

All of this sounds good, but virtualization is not a wild success. The software defined data center market at one time was expected to reach $77 billion by 2020 but instead has only reached $28 billion as of 2023. There are many vendors in the space, which creates pricing wars.

And so, it’s speculative as to how the software defined data center or networks would ultimately accelerate in growth based on AI workloads. The anticipated acceleration is mainly from restructuring, rather than product-market fit. This is what management said on the earnings call: “And it just doesn’t stop there because it’s the math and the trajectory. And to answer your question, you’re right, we are accelerating from $12 billion, and we’re probably seeing a double-digit growth for the next three years, just by sheer math of selling that higher value virtualization stack versus the very loose component sales in the past, particularly on compute only.”

There was a solid question about this in the Q&A which I’m quoting in full below.

Harlan Sur:

[…] 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:

Well, as you may be aware, in the last VM Explore in Las Vegas, VMware came out and announced in partnership with NVIDIA, the VMware Private AI Cloud Foundation. Another way of describing it is, the VMware Cloud Foundation Software Stack, the whole VCF stack runs NVIDIA coder, runs the NVIDIA GPU. That is the partnership. So, if you’re an enterprise, it’s a very easy step to get into gen AI analytics because the data center that you as an enterprise own on-prem that runs VCF will by default run the NVIDIA GPU software stack as well.

Another way to put it, it virtualizes the NVIDIA GPU. That’s the VMware software stack as well. So it’s a very strong attraction in our — from our perspective to, in fact, accelerate thinking of a lot of enterprise to adopting the whole VCF site. It’s simply because not only does it virtualize the data centers and make your data on-prem data center much more resilient, easier to manage, lower cost to manage, it has the added benefit, a big attraction this is of being able to right away start running AI workloads

Broadcom’s Financial Overview:

Broadcom consistently delivered a net profit margin exceeding 37% throughout fiscal year 2023. Additionally, Broadcom demonstrates exceptional cash flow generation, with free cash flow exceeding 44% in each quarter of FY2023 and even surpassing 50% in the last three quarters.

Broadcom's acquisition of VMware in November 2023 is intended to bolster the company's position in the AI space, while also strengthening its software business. Secondly, the merger will eventually create recurring revenue streams, which once complete, will be a positive. Furthermore, this acquisition diversifies Broadcom's portfolio, with infrastructure software projected to account for roughly 40% of FY2024 revenue compared to 21% in FY2023. This shift mitigates the impact of cyclical downturns inherent to the semiconductor industry.

The integration process is expected to take a year and will initially have a drag on profit margins due to transition costs and VMware's lower margin profile, cost-cutting measures and merger synergies are anticipated to improve margins in the long term.

When an acquisition is complete, it typically weighs on stock price while investors move to the side lines to see how the teams merge internally and also externally for customers. Unfortunately, the VMware acquisition is not going too well with rumors that customers are disgruntled alongside Broadcom spinning off non-core segments and selling them off to other companies.

That complicates the picture as it requires understanding the precise impact of each segment independently, which is impossible to do given cloud companies tend to cross-sell products. It's also important to note that Broadcom is transitioning VMware clients to a subscription-based business. Per the earnings call: “[…] and we are converting more and more customers step-by-step as they come up for renewal into this higher value stack, and we’re doing it on a subscription basis. So become very focused. So we will kick it off at a much lower rate — because subscription generally brings down revenues, as you know, in software based on revenue recognition. But we see a trajectory of accelerated growth even in 2024 — through 2024. And it just doesn’t stop there because it’s the math and the trajectory. And to answer your question, you’re right, we are accelerating from $12 billion, and we’re probably seeing a double-digit growth for the next three years, just by sheer math of selling that higher value virtualization stack versus the very loose component sales in the past, particularly on compute only.”

When looking at revenue growth, it’s important to strip out VMware’s contribution post-acquisition. Management is firm in the earnings calls that the VMware will accelerate. If this comes to fruition, the Street will likely reward Broadcom as VMware is the primary risk given the synergy of the rather large acquisition ($60 billion) is unproven. However, due to the uncertainty around restructuring the VMware acquisition, time is on our side to try to get AVGO at a lower price.

Broadcom’s Revenue and EPS:

Broadcom’s revenue in the Q4 FY2023 ending Oct grew by 4.1% YoY to $9.3 billion. Revenue was in line with the analyst consensus.
Analysts expect revenue to grow 31.6% YoY to $11.73 billion in the next quarter. Since the company completed the acquisition of VMware on November 22, 2023, these estimates include VMware’s revenue. Headline revenue numbers are expected to accelerate for the next four quarters due to VMware and then will re-acclimate in the January 2025 quarter at 16.4% growth. These quarters are likely to be watched closely due to reasons outlined above, which is that it will be apparent and quite easy to model VMware’s impact.
Organic revenue, excluding the VMware acquisition, represents a 6.1% YoY growth for FY2024 ending in October, down from 7.9% in FY2023 due to the cyclical slowdown in the semiconductor sector. See more discussion on this below.
However, the company will see a higher growth rate in FY2025 at 10.9% YoY for $55.28 billion. The growth rate is helped by an increasing contribution from AI. Management stated in the earnings call that revenue from generative AI will grow from 15% in FY2023 to more than 25% in FY2024. The company’s CEO, Hock Tan, said in the earnings call, “Revenue from generative AI in fiscal ‘23 reached 15% of semiconductor revenue, in line with our expectation. And moving on to fiscal ‘24, we forecast semiconductor solutions revenue to be up mid- to high-single-digit percent year-on-year. We expect revenue from generative AI to represent more than 25% of the semiconductor revenue, consistent with prior guidance, which more than offset the lack of growth from non-AI semiconductor revenue.”
Management chose to not provide quarterly guidance, and to instead provide FY2024 guidance of $50 billion, representing YoY growth of 39.6% at the mid-point. This is a break in style as quarterly guidance is typically given. Here is what was said on the call: “Now on to guidance. As Hock discussed, with the recent closing of our VMware acquisition and the integration process, which will take at least one year, for fiscal 2024, we will provide our outlook for the full year instead of quarterly guidance. Based on current business trends and conditions, our guidance for fiscal year 2024 is for consolidated revenues of $50 billion. Within this, our fiscal year 2024 semiconductor revenue is expected to grow mid- to high-single-digit percent year-on-year. Our fiscal year 2024 infrastructure software segment revenue from continuing operations is expected to be $20 billion, including $8 billion from CA, Symantec Enterprise and Brocade and $12 billion from VMware.”
Management mentioned that VMware’s 11 months expected contribution from the time the acquisition is closed is $12 billion for the FY2024 ending October.

Pictured Above: Revenue includes VMware acquisition. Organic revenue, excluding the VMware acquisition, is expected to be 6.1% YoY growth for FY2024 ending in October, down from 7.9% in FY2023.

Segments

The Semiconductor Solutions segment revenue grew by 3% YoY to $7.3 billion and has witnessed a cyclical slowdown. It is down from 5% growth in the previous quarter and 26% growth in the same quarter last year.

In this segment, networking revenue is the largest by end markets, constituting 42% of Q4 semiconductor revenue.
Networking revenue grew by 23% YoY to $3.1 billion due to strong demand from hyperscalers. Due to AI, management expects FY2024 networking revenue to grow 30% YoY, up from 21% in FY2023.

The company is a beneficiary of generative AI and reported $1.5 billion in Q4 FY2023, for a run rate of $6 billion per year. Citi Analyst, Christopher Danely, said in a research note that the company’s AI revenue will double from $4 billion in FY2023 to $8 billion in FY2024, and he expects the AI business will offset the correction in the semi-business.

Looking further out, Mizuho analyst Vijay Rakesh said “that Broadcom’s AI revenue will likely grow from $8 billion in 2024 to $20 billion in the calendar year 2027, thanks to its custom ASIC AI portfolio.”

The company’s CEO, Hock Tan, said in the earnings call, “This was primarily driven by strong demand from hyperscalers for our custom AI accelerators and as well for our networking switches, routers and NICs, Network Interface Cards, dedicated towards scaling our AI data centers.”

“As you know, even as Ethernet is the standard protocol in front-end networks, hyperscalers are also deploying Ethernet predominantly in their AI networks. In fiscal ‘23, networking revenue grew 21% year-on-year to $10.8 billion. If we exclude the AI accelerators, networking connectivity represented about $8 billion, and this is purely silicon, not systems, not cable nor subsystems. In fiscal 2024, we expect networking revenue to grow 30% year-on-year, driven by accelerating deployment of networking connectivity and expansion of AI accelerators in hyperscalers.”hyperscalers are also deploying Ethernet predominantly in their AI networks. In fiscal ‘23, networking revenue grew 21% year-on-year to $10.8 billion. If we exclude the AI accelerators, networking connectivity represented about $8 billion, and this is purely silicon, not systems, not cable nor subsystems. In fiscal 2024, we expect networking revenue to grow 30% year-on-year, driven by accelerating deployment of networking connectivity and expansion of AI accelerators in hyperscalers.”

The infrastructure software segment grew by 7% YoY to $2.0 billion and accelerated from 5% growth in the July quarter.

Due to the cyclical correction, server storage connectivity Q4 revenue declined by (17%) YoY to $1 billion. For FY2023, it grew by 11%. However, the cyclical weakness is expected to continue, and revenue is expected to decline in the mid-to-high teens for FY2024.

Broadband Q4 revenue also declined by (9%) YoY to $950 million due to the cyclical correction. The management expects the trend to continue, and for FY2024, revenue is expected to be down in the low-to-mid teens. In FY2023, broadband revenue grew by 8% YoY to $4.5 billion.

Wireless revenue declined by (3%) YoY to $2 billion. In FY2023, revenue was down (2%) YoY and the management expects revenue to be stable in FY2024. Lastly, the industrial resale revenue was flat at $236 million.

In the Core software segment, the consolidated renewal rates averaged 119%, up from 117% in the previous quarter. In the strategic accounts, it averaged 130% and was the highest in the last five quarters.

EPS:

EPS came in at $8.25 compared to $7.83 in the same period last year. The adjusted EPS came at $11.06 compared to $10.45 for the same period last year.
Analysts expect adjusted EPS to grow 0.9% YoY to $10.42 in the next quarter.
Prior to acquisition, VMware reported adjusted EPS of $1.83 in its last quarter (July 2023) as a public company.

Margins:

Gross margin for Q4 FY2023 ending Oct was 68.9% compared to 66.4% in the same period last year and 69.5% in the previous quarter.
The operating margin improved 100 bps YoY to 45.6%. The adjusted operating margin improved 20 bps YoY to 61.8%.
The company has a very strong bottom line. The net margin improved by 30 bps YoY to 37.9%. The adjusted net margin improved 90 bps YoY to 51.8% and was flat sequentially.
The adjusted EBITDA margin was 65.1% compared to 64.1% in the same period last year and 65.4% in the previous quarter.
VMWare’s EBITDA margin prior to acquisition in the July quarter was 28.6% compared to Broadcom’s 55.3%. The operating margin of VMware was 16% compared to 43.4% for Broadcom.

The adjusted EBITDA for the FY2023 ending Oct was 64.8%, and the management guide for the FY2024 is 60%, including VMware. The margin drop is mainly due to VMware's current lower margin. However, through cost-cutting initiatives like job cuts, Broadcom will aim to reach a 65% adjusted EBITDA margin for VMware. One focus area is SG&A expense, which constituted around 41% of revenue for VMware compared to around 4% for Broadcom. Hock Tan said in the earnings call, “At steady state, we’ll get to pretty close to 65% on VMware.”

The management also clarified that they are confident of achieving $8.5 billion EBITDA from VMware.

Harlan SurHarlan Sur

And then, just on my first question, are you guys still targeting $8.5 billion of EBITDA in three years on VMware?

Hock TanHock Tan

As Kirsten indicated, as we exit fiscal ‘24, we are practically at a run rate of $8.5 billion EBITDA.

Management said that the integration of VMware will take until the end of the fiscal year and require about $1 billion in transition spending. The CFO, Kirsten Spears, said in the earnings call, “During fiscal ‘24, we expect to incur about $1 billion of spend related to transitioning VMware into the new Broadcom model. This transition spending will be largely completed by the end of the fiscal year as our VMware spending run rate exits fiscal ‘24 at approximately $1.4 billion per quarter, down 40% from a year ago.”

The adjusted EPS for FY2024 is expected to grow 10.8% YoY to $46.83 and a further 19.4% YoY to $55.90 in FY2025. UBS analyst, in a research note, had earlier said that the VMware deal “to be 10% accretive to 2024 EPS and about 17% accretive to 2025.”

Cash Flow and Balance Sheet

Broadcom uses its large cash margin to frequently acquire companies.

Operating cash flow margin for Q4 FY2023 ending Oct was 51.9% compared to 53.2% in the previous quarter.
The free cash flow margin was 50.8% compared to 51.8% in the previous quarter.
The company has cash of $14.2 billion and debt of $39.2 billion. While the short-term debt is $1.6 billion, about $31.3 billion will mature after FY2028. At the end of FY2023, the company took a loan of $30.39 billion to finance the VMware merger and assumed $8.3 billion of VMware debt. While high debt is a concern, most of Broadcom’s debt has long maturities, and the company generates strong cash flows.
The company spent $15.3 billion for FY2023 in cash dividends and share repurchases. It had $7.2 billion remaining in the authorized share repurchase program.

More AI Commentary:

Vivek Arya, Bank of America analyst asked about Broadcom’s participation in the $400 billion AI accelerators market.

Hock Tan

“And we’re seeing this as we all are seeing LLM models continue to change and the face — the shape of generative AI dynamically change more and more, where training and inference are now starting to, in a way, converge and the chip designs are changing. And we are seeing that in the way we design specific custom chips for hyperscalers. That’s interesting. So that’s a very interesting opportunity for us. And as I indicated in my remarks, we see that revenue as part of networking revenue, $4 billion and networking — AI networks and going — doubling almost during 2024. Nothing new. We have said that before. And if anything else, we are reinforcing that particular guidance.”And as I indicated in my remarks, we see that revenue as part of networking revenue, $4 billion and networking — AI networks and going — doubling almost during 2024. Nothing new. We have said that before. And if anything else, we are reinforcing that particular guidance.”

Conclusion:

My takeaway is that the I/O Fund is likely to own Broadcom this year, but our goal is to enter at a lower price. Broadcom is trading above its 3-year median on PE Ratio at 40 PE compared to a 28.5 median. The current sales valuation of 15.5 PS is about double the 5-year median of 7.5 and about double the 3-year median of 8. When we go back to 2014, Broadcom has only traded above a PS Ratio of 10 one time at the height of the 2021 market. However, at a certain price, Broadcom belongs in our AI portfolio. Over the next few years, Big Tech is likely to diversify away from Nvidia’s GPUs, plus Ethernet networking continues to be upgraded for AI purposes, which means Broadcom should be on our radar.

In addition to valuation, for our purposes, we think the timing will be better once the VMware acquisition has settled as there are mixed reports from the customer perspective. Once this blows over and the restructuring is complete, it will be a more optimal time to enter Broadcom, which is richly valued at the moment.

Recommended Reading:

Broadcom: Networking/ASICs Giant and The Second Largest by AI Revenue