Broadcom’s impressive F2Q24 performance, driven by strong AI sector sales and strategic acquisitions, highlights the company’s significant growth potential in the AI semiconductor market. The increasing demand for AI networking and ASIC-based AI accelerator semiconductors positions Broadcom well for future growth, especially as big tech companies seek cost-effective and performance-optimized solutions.
Broadcom’s Strong Second Quarter: AI Boosts Profits
Broadcom, a major player in the custom chip (ASIC) market, has been making significant strides in the AI semiconductor sector, closely following Nvidia. As of 2023, Broadcom holds a 7.2% global market share in semiconductors, placing it fourth worldwide. In the ASIC segment, Broadcom is the leader with a 35% market share.
On June 12th, Broadcom announced its fiscal second quarter (F2Q24) results, surpassing market expectations. The company reported a remarkable revenue of $124.9 billion, a 43% increase year-over-year (YoY), exceeding the consensus estimate of $120.1 billion. Operating profit for the quarter was $71.4 billion, up 36% YoY, with an operating profit margin (OPM) of 57%, slightly above the expected 56%. This exceptional performance is primarily due to increased sales in the AI sector and the impact of VMWare acquisitions. AI sector sales alone reached $31 billion, a 34.8% increase quarter-over-quarter (QoQ), while VMWare sales were $21 billion, up 28.6% QoQ. Combined, these segments accounted for 46% of Broadcom’s total revenue, up from 37% in the previous quarter.
As a result of these strong results, Broadcom has raised its fiscal year 2024 (FY24) revenue guidance from $50 billion to $51 billion, driven by an expected increase in AI sector revenue from $10 billion to $11 billion.
Rising Demand for AI Networking Chips
With the race for GPU acquisition heating up among big tech companies, the demand for networking and ASIC (Application-Specific Integrated Circuit) semiconductors is expected to grow. Broadcom offers a range of AI networking semiconductors, such as the Tomahawk 5 Ethernet switch and the Jericho3-AI fabric chip. These products are essential for building large-scale clusters for AI machine learning, enabling connections between GPUs within and across data centers.
Currently, Broadcom’s AI system deployment is divided approximately into 2/3 for AI accelerators and 1/3 for networking infrastructure. However, this balance is expected to shift, with networking anticipated to account for up to 40% by the end of the year. This trend highlights the growing importance of networking semiconductors in the AI ecosystem.
AI accelerator chips are gaining more attention than AI networking semiconductors due to their significant growth potential. While Nvidia’s general-purpose GPUs currently dominate the AI accelerator market, the landscape is expected to change with competition from big tech companies’ proprietary AI accelerator chips, based on ASIC or FPGA technologies.
In the early stages of the AI market, where AI workloads are still stabilizing, Nvidia’s versatile and high-spec GPUs hold a significant share. However, as AI workloads become more stable and widespread, cost efficiency will become a key factor. This shift will likely drive substantial growth in customized AI accelerators optimized for performance and cost-effectiveness, tailored to each big tech company’s AI systems and applications. By 2027, ASIC-based AI accelerator chips are projected to account for 30% of the overall accelerator semiconductor market.
Broadcom is leading this trend, particularly in ASIC-based AI accelerators. Since 2016, the company has been co-designing custom AI accelerators like the Tensor Processing Unit (TPU) with Google and the Meta Intelligence Accelerator (MITA) with Meta since 2020. Recently, Broadcom has added new customers, with market speculation pointing to Amazon, Apple, or ByteDance as potential clients.
If big tech companies continue to invest in AI strategies that focus on diversifying the accelerator chip supply chain and reducing total cost of ownership (TCO), Broadcom could secure additional clients in the future, further strengthening its position in the AI accelerator market.
Can Broadcom Replace Nvidia?
Following the announcement of its strong F2Q24 results and a 10-to-1 stock split, Broadcom’s stock surged by 25%, positioning it as a potential alternative to Nvidia. However, recent price adjustments have mirrored Nvidia’s volatility, raising questions about the sustainability of Broadcom’s stock as an Nvidia substitute, especially during periods of declining Nvidia stock prices.
Currently, Broadcom’s enterprise value-to-EBITDA (EV/EBITDA) multiple stands at 22x, significantly higher than the sector average of 14x over the past five years. Additionally, the company’s long-term debt and short-term borrowings are relatively high, at $73.4 billion and $2.4 billion, respectively.
Despite these challenges, Broadcom’s AI-related revenue is projected to increase from $11 billion this year to $16 billion next year and $20 billion by fiscal year 2026 (FY26), according to Melius Research. However, even with the recent upwardly revised market consensus, the overall revenue growth rate for next year is expected to be 16%, falling short of the industry average. Over the next five years, the average annual revenue growth rate is anticipated to be around 15%.
Moreover, Broadcom’s revenue is heavily reliant on a few major customers, posing a significant threat to revenue stability if it lags behind competitors like Marvell Technology. This risk factor is comparable to Nvidia, which serves nearly all AI-related companies.