Complex data center workloads, such as training machine learning models and running artificial intelligence (AI) applications, take significantly longer to run on central processing units (CPUs) alone. To do so, we use specialized semiconductors to speed up compute-intensive AI tasks.
In the semiconductor industry, Nvidia (NVDA 0.39%) Graphics processing units (GPUs) have emerged as an industry standard. In fact, the company has captured 80% to 95% market share for AI accelerators, according to analysts. However, NVIDIA shareholders recently received some worrying news from rival Broadcom. (AVGO 3.15%).
Nvidia shareholders received bad news from Broadcom
Broadcom sells a variety of semiconductor products, including combined Wi-Fi and Bluetooth chips for Apple and Samsung smartphones and network chips for Arista switches. But Wall Street is particularly fascinated by its leadership in application-specific integrated circuits (ASICs). ASICs are chips built specifically for specific use cases, such as accelerating artificial intelligence (AI) workloads.
Analysts estimate that Broadcom has about 60% market share for custom AI chips thanks to its relationships with three hyperscalers, a term used to describe companies with vast data center footprints. are. Broadcom hasn’t identified Hyperscale’s customers, but analysts generally believe it to be Google parent Alphabet, Metaplatforms, and TikTok parent ByteDance.
Broadcom estimates that revenue from its three existing hyperscale customers will grow from $12.2 billion in 2024 to $60 billion to $90 billion in 2027. In other words, the company expects sales of custom AI chips to grow at least 70% annually over the next three years. It takes a few years, but it happens probably as quickly as 95% per year.
This is unfortunate for Nvidia shareholders, as it means Broadcom will likely gain market share with AI accelerators. In fact, Morgan Stanley analysts predict that ASICs will account for 13% of AI accelerator sales by 2027, up from 11% in 2024. We also believe that this number could reach 15% by 2030. But there’s more bad news for Nvidia shareholders. .
Broadcom CEO Hock Tan told analysts on a fourth-quarter earnings call that Broadcom has two new hyperscalers that are likely to become revenue-generating customers by 2027. He said he had selected. That means revenue from custom AI chips could actually grow at more than 95% annually over the next year. A few years. Importantly, Broadcom does not identify its customers, but analysts believe they are Apple and ChatGPT creator OpenAI.
Image source: Getty Images.
Nvidia shareholders still have good reason to be optimistic
Broadcom said it has selected two additional hyperscalers as potential customers. CEO Hock Tan himself used this term because Broadcom doesn’t develop ASICs for small businesses, and small businesses aren’t interested in using custom AI chips.
There are two reasons for this. First, ASIC designs are expensive, so customers need a large enough data center footprint to justify the cost. Piper Sandler analyst Harsh Kumar recently told CNBC that each chip costs about $500 million to design, so it’s not economically wise to work with customers who can only order a few thousand units. Ta. Instead, orders must be between 250,000 and 500,000 units.
Second, custom chips are designed for specific workloads and lack supporting software development tools, making them less flexible. Nvidia offers a robust ecosystem of code libraries and pre-trained models that streamline GPU application development. No such tools exist for ASICs. This means that implementing custom chips requires a high level of technical expertise and Broadcom’s customer base is limited.
Additionally, companies experimenting with ASICs may ultimately decide that the costs outweigh the benefits. New Street Research’s Antoine Chkaiban says Google and Amazon are the only two companies deploying custom AI silicon at scale. Therefore, Nvidia is well-positioned to maintain its leadership in AI accelerators. Analysts at Bank of America expect it to capture 75% market share in 2030, down only slightly from 80% in 2024.
Looking ahead, Wall Street believes Nvidia’s adjusted earnings will grow 34% annually through fiscal 2027 (ending in January 2027). With this consensus, the current valuation of 53 times adjusted earnings looks very reasonable. Future investors can buy a few shares with confidence, and current shareholders have good reason to be optimistic about the future.
John Mackey, former CEO of Amazon subsidiary Whole Foods Market, is a member of the Motley Fool’s board of directors. Alphabet executive Suzanne Frye is a member of The Motley Fool’s board of directors. Bank of America is an advertising partner of Motley Fool Money. Randi Zuckerberg is a former head of market development and spokesperson at Facebook, sister of Meta Platforms CEO Mark Zuckerberg, and a member of the Motley Fool’s board of directors. Trevor Jennewine has held positions at Amazon, Arista Networks, and Nvidia. The Motley Fool has positions in and recommends Alphabet, Amazon, Apple, Arista Networks, Bank of America, Meta Platforms, and Nvidia. The Motley Fool recommends Broadcom. The Motley Fool has a disclosure policy.
