The demand for specialized chips used in artificial intelligence (AI) models and applications in data centers has surged significantly over the past couple of years. Nvidia, the leading beneficiary of this trend, has secured a substantial share of the AI semiconductor space by capturing a significant portion of revenue from selling these chips.
In the third quarter of fiscal 2025, Nvidia generated $30.8 billion in sales from data center chips, representing a remarkable 112% increase from the same period last year. Its nearest rival in this segment, Advanced Micro Devices (AMD), is forecasting $5 billion in data center GPU revenue for this year.
However, GPUs are not the only type of chip used for AI model training and inferencing. Broadcom’s rapid growth in demand for its application-specific integrated circuits (ASICs) underscores the increasing need for custom AI processors.
Broadcom’s fiscal 2024 revenue from sales of custom AI accelerators reached $12.2 billion, which is still significantly lower than Nvidia’s projected annual data center revenue of $79.6 billion for this year. Nevertheless, Broadcom’s impressive growth points to a rising demand for alternative chips to tackle AI workloads.
Major cloud companies have been developing their own in-house chips to reduce costs and accelerate AI development, given the high price of Nvidia’s AI chips and lengthy production times. As a result, Broadcom’s ASICs are being used by prominent players such as Google, Meta Platforms, ByteDance, and OpenAI.
Industry experts predict that the market for AI-specific ASICs will grow at an annual rate of 32% through 2030, with some estimates suggesting the market size could reach $621 billion in 2032. Despite Broadcom’s significant growth, Nvidia is still well-positioned to deliver healthy gains in the long run due to its expanding data center opportunity and addressable market.
Source: https://www.fool.com/investing/2024/12/25/what-could-broadcoms-latest-earnings-report-mean-f