In the rapidly evolving landscape of artificial intelligence (AI), it’s vital to recognize that Nvidia is not the sole player catering to the expansive needs of major AI investors.
As we look ahead to 2025, the tech industry is gearing up for a substantial investment wave. The four largest technology corporations globally are poised to allocate a staggering $320 billion in capital expenditures (capex) for this year. A significant portion of this budget will be directed toward the construction and modernization of AI data centers. These facilities are crucial for training and deploying generative AI technologies, which are becoming increasingly integral to various applications across industries.
Nvidia (NASDAQ: NVDA) is undoubtedly set to benefit immensely from this influx of capital. The majority of major tech firms have reiterated the critical nature of their partnerships with the leading graphics processing unit (GPU) designer. However, the insights shared by the leadership of Meta Platforms and Alphabet, Google’s parent company, indicate a promising outlook for another semiconductor manufacturer.
Broadcom (AVGO -1.17%) stands to gain significantly from the anticipated increase in tech investments in 2025 and beyond. This article will delve into the comments from Meta and Google regarding their partnerships with Broadcom and the implications for its investors.
Image source: Getty Images.
Understanding the Impact of Custom Chips on AI Efficiency
Broadcom manufactures a diverse array of semiconductors, but its role in powering AI data centers is particularly crucial. The company produces chips that facilitate the operation of network switches, which are essential for ensuring that data is transmitted swiftly and efficiently between servers. In essence, Broadcom’s technology enhances the performance of Nvidia’s graphics processing units within Meta’s and Google’s data centers, maximizing their operational effectiveness.
Moreover, Broadcom is involved in the development of proprietary chips utilized by Meta and Google. For instance, the company is responsible for the technology behind Meta’s custom AI chip known as the Meta Training and Inference Accelerator (MTIA), which is designed to optimize various AI processes.
On the other hand, Google has been collaborating with Broadcom since 2015 to create its Tensor Processing Unit (TPU). During the fourth-quarter earnings calls, both Meta and Google expressed optimism about their custom chip initiatives. Meta’s Chief Financial Officer, Susan Li, stated that they began deploying MTIA for enhancing the performance of their ranking and recommendation systems for ads and organic content in 2024. She further anticipated an increase in MTIA’s adoption for these applications throughout 2025.
Meta also envisions a future where its custom chips will gradually replace Nvidia’s general-purpose GPUs in its data centers. Li remarked on their plans to expand MTIA’s capabilities to support core AI training workloads and various generative AI applications in the coming years.
Alphabet’s CEO, Sundar Pichai, highlighted the strong demand for the latest TPUs during their fourth-quarter update, noting that they had launched their sixth-generation TPU to the public in December. This indicates that 2025 could be a pivotal year for the adoption of TPUs, especially given that developers had limited time to integrate these new chips before the quarter ended.
Custom AI accelerator chips are increasingly recognized for their efficiency compared to general-purpose GPUs when handling specific AI workloads. As major tech companies expand their data center operations, achieving even marginal improvements can significantly impact profitability. Consequently, the importance of leveraging custom chips and maintaining control over the technology stack is growing, particularly as capital expenditure budgets continue to rise.
Broadcom’s Long-Term Vision for Custom AI Accelerators
Broadcom’s partnerships extend beyond Alphabet and Meta; ByteDance, the parent company of TikTok, ranks among its top three customers. During the fourth-quarter earnings call, Broadcom’s management revealed that they had secured partnerships with two additional clients developing next-gen chips through their expertise, with speculation pointing toward Apple and OpenAI.
Apple has shown a strong preference for utilizing Broadcom’s chip technology, having previously employed Google’s older TPUs to train its Apple Intelligence system last year. If Apple pursues an initiative to gain more control over its AI training and development processes, it could quickly become one of Broadcom’s largest clients.
The market for custom AI accelerators is projected to reach between $60 billion and $90 billion within the next three years. According to last year’s estimates, Broadcom commanded approximately 70% market share in custom AI accelerators, generating around $12 billion in sales from this segment alone.
To provide context, Broadcom’s total sales across its diverse operations for 2024 were approximately $51.6 billion, indicating that the growth from AI chips could potentially double its revenue if the company maintains its market position.
Despite the positive outlook, it is essential to consider valuation. Currently, Broadcom’s stock is trading at 37 times its projected forward earnings. While the company is expected to experience robust earnings growth in the coming years, justifying such a premium price for the stock remains challenging.
If the stock price were to decrease to around 30 times the anticipated 2025 earnings, it may present a buying opportunity, especially given that analysts forecast over 30% earnings growth this year, resulting in a price/earnings-to-growth ratio of around 1. Additionally, if the adoption rate of Broadcom’s custom chips exceeds expectations as Meta and Google aim to broaden their usage in their data centers, the company could align with its current valuation.
For investors keen on the future trajectory of artificial intelligence, Broadcom is a company to monitor closely. However, it is essential to note that the stock appears to be currently priced at a premium, making it less advisable for immediate investment.
Randi Zuckerberg, a former director of market development and spokeswoman for Facebook and sister to Meta Platforms CEO Mark Zuckerberg, is a member of The Motley Fool’s board of directors. Suzanne Frey, an executive at Alphabet, is a member of The Motley Fool’s board of directors. Adam Levy has positions in Alphabet and Meta Platforms. The Motley Fool has positions in and recommends Alphabet, Meta Platforms, and Nvidia. The Motley Fool recommends Broadcom. The Motley Fool has a disclosure policy.
