Author: Zhou Yuan / Wall Street Watch

Recently, Broadcom's performance report released on December 12 for the 2024 fiscal year showcased the strength of its AI Technology and products, shocking the Global market.

Due to the market acceptance of Broadcom's ASIC (Application-Specific Integrated Circuit), NVIDIA, a company dealing with General-Purpose Computing Chips, saw its stock drop nearly 7%, hitting a new low since October 18.

What exactly does Broadcom do?

Broadcom's Business mainly covers two major areas: Semiconductor Technology and Infrastructure Software, showcasing a strong and diversified characteristic.

In the 2024 fiscal year, Broadcom's net income reached $51.6 billion, with gross margin at 76.5%, making its profitability and operational efficiency quite remarkable.

Among them, the revenue from Semiconductor Business was $30.1 billion, and the revenue from Infrastructure Software Business was $21.5 billion, representing a year-on-year growth of 196%. In the AI field, Broadcom's performance was particularly outstanding, with AI business revenue reaching $12.2 billion, a year-on-year increase of 220%, significantly improving its share in the Semiconductor Business, fully demonstrating its competitive strength in the AI era.

Broadcom's overall technical strength includes network chip technology, AI custom chip technology, and interconnect technology. Like NVIDIA, Broadcom also has specialized Software and ecological layouts.

In the era of high-performance computing and AI, network chips are a key pillar of Broadcom's business.

As the world's largest network chip manufacturer, Broadcom's Ethernet product line provides various solutions for AI clusters. For example, the Tomahawk 5 Ethernet Switch Chip in the Endpoint Scheduled architecture has achieved an 80-fold bandwidth increase since 2010—from 640Gbps to 51.2Tbps in 2022, while energy consumption has been reduced by over 90%, efficiently supporting data scheduling for small-scale AI clusters.

In the Switch Scheduled architecture, the Ramon and Jericho3 - AI Chips can achieve multi-path interconnection, connecting up to 32,000 GPUs, with each AI accelerator card providing 800Gbps of data bandwidth, effectively enhancing network performance.

In addition, Broadcom's Thor series network cards are continuously innovating: Thor 2 is the industry's first 5nm process 400 gigabit Ethernet NIC device, supporting 16 PCI Express 5.0 channels, capable of directly driving 5-meter Copper cables, and supporting RoCE v2 RDMA technology. Broadcom is also developing even faster network card chips to continue leading the development of network chip technology.

In the field of AI custom chip technology, Broadcom is the world's largest AI custom chip service provider, holding about 70% of the $17.5 billion serviceable market.

Since entering the customized AI accelerator market in 2014, Broadcom has closely collaborated with clients to develop multiple generations of XPU accelerators.

Broadcom plans to launch the next-generation XPU based on 3nm technology in the second half of 2025 to accelerate high-performance computing development.

In terms of IP investment, Broadcom is extremely forward-looking, having invested approximately 3 billion dollars in the differentiation and innovation of XPU IP, with a wide range of IP portfolio including SerDes IP, AI-optimized NICs IP, cache IP, CPO IP software API, etc., providing critical support for XPU.

At the same time, Broadcom has also made significant breakthroughs in packaging technology, with its 3.5D eXtreme Dimension System-level Packaging (XDSiP) platform technology integrating a large number of chips and HBM stacks, significantly advantageous in interconnect density and power efficiency, effectively enhancing the efficiency of large-scale AI computing and reducing power consumption.

Interconnect technology and related products are also areas of technological advantage for Broadcom.

For instance, in the PCIe + Retimer technology, Broadcom has a strong presence in the PCIe technology field.

What is PCIe.

PCIe (Peripheral Component Interconnect Express) is a high-speed peripheral component interconnect standard, a high-speed serial computer expansion bus standard used to connect the main board of a computer to various peripheral devices.

Since the launch of the first generation PCIe in 2003, Broadcom has cumulatively shipped over 1 billion PCIe ports. In 2018, Retimer technology was introduced to solve long-distance signal transmission problems, becoming a standard in subsequent PCIe developments. In March 2024, Broadcom launched the world's first 5nm PCIe Gen 5.0/CXL2.0 and PCIe Gen 6.0/CXL3.1 retimers, providing important assurance for high-speed interconnect.

In the field of optical interconnect, Broadcom has long been in a leading position, especially with fruitful achievements in optical interconnect technologies such as VCSEL and EML high-linear continuous wave lasers. These technologies play a key role in high-speed interconnect for AI and ML systems, with speeds iterating up to 200Gbps.

Currently, Broadcom ships over 50 million high-speed optical communication lasers each year.

Broadcom launched the industry's first 51.2Tbps CPO Ethernet Switch, Bailly, which significantly reduces running power consumption of optical interconnects compared to pluggable transceivers and improves silicon area efficiency, while its products like the 5nm Sian 2 also possess strong electrical and optical interface capabilities.

Additionally, the 5nm Peregrine and 3nm Condor SerDes technologies developed by Broadcom not only support Copper connections but also natively support CPO, applicable on multiple product platforms. This technology has bandwidth capabilities that surpass 100GB, while the top domestic similar products only have a bandwidth of 2GB.

Broadcom has also achieved remarkable success in infrastructure Software, with revenue accounting for as much as 41.6% of total revenue.

Through a series of strategic acquisitions, such as the acquisition of software giant VMware in 2023, Broadcom has built a relatively complete Software business system, further strengthening its competitiveness in the Semiconductors technology and infrastructure Software market, providing customers with more comprehensive solutions, and forming an efficient collaborative network and software ecosystem.

However, to be realistic, the ASIC ecosystem integrity and usability of Broadcom are not as good as NVIDIA's CUDA; the Software environment is relatively simple, mainly designed for specific purposes and algorithms, with a high level of programming difficulty, and the development tools and Software libraries are not rich enough, requiring developers to spend more time debugging and developing, limiting its promotion in a wider range of application scenarios.

In contrast, NVIDIA's GPU products iterate quickly, keeping up with technological development trends and market demands, with new products launched or performance improved every year.

In comparison, once Broadcom's ASIC chips are designed and manufactured, it becomes very difficult to make changes; the design and manufacturing cycle is relatively long, making it less suitable for the rapidly changing AI market with frequent technological iterations, which may lead to it falling behind NVIDIA in terms of technological advancement.

The inability to adjust technology and manufacturing processes on demand is the main reason why customized chips are more expensive than general-purpose chips.

In addition, customized chips have another issue: customer base is limited.

Broadcom primarily focuses on providing customized chip services for large technology companies, resulting in a relatively narrow customer base; NVIDIA's GPU products are highly versatile and can meet the needs of businesses of various sizes and applications, reaching a wide customer base from small startups to large enterprises.

If any of Broadcom's major clients undergo a technological shift or business adjustment, Broadcom's performance will also be significantly negatively impacted.

From this perspective, NVIDIA has a greater advantage in market coverage and customer diversity, enabling it to better respond to market fluctuations and changes in customer demand.

A lack of full-stack solutions is also Broadcom's shortcoming.

Broadcom mainly focuses on chip design and packaging, lacking a full-stack solution like NVIDIA: when customers use Broadcom chips, they need to solve issues related to network devices, software, and other supporting facilities themselves, increasing the cost and complexity of customer usage.

Perhaps the market has fully recognized Broadcom's shortcomings in the face of NVIDIA; thus, on December 23, after the release of the fiscal year 2024 performance report on December 12, NVIDIA regained lost stock price ground, increasing by 1.70% overall; while Broadcom's stock price fell 6.84% from its high of December 14 to $232.5 on December 23, although its total market value still remains above $1 trillion, at $1.09 trillion.

In terms of future prospects, Broadcom is not only a leader in the Semiconductors Industry but also a key force driving the development of AI Technology.

From network connection to AI computing power, from Semiconductor Hardware to Software infrastructure, Broadcom's technological puzzle is complete and powerful, providing a solid foundation for the development of the AI industry.

Of course, Broadcom also faces many challenges, such as increasingly fierce market competition, requiring continuous investment in research and development to maintain its CSI Leading Technology Index; at the same time, the rapid development of AI Technology means Broadcom needs to continuously innovate to adapt to the ever-changing market demand.