©Reuters. FILE PHOTO: Nvidia’s Grace Hopper superchips are displayed at Gigabyte during COMPUTEX Taipei, one of the world’s largest computer and technology trade shows, in Taipei, Taiwan, May 30, 2023. REUTERS/Ann Wang/File Photo
By Max A. Cherney and Stephen Nellis
SAN FRANCISCO (Reuters) – Nvidia (NASDAQ:) is building a new business unit focused on designing custom chips for cloud computing companies and others, including advanced artificial intelligence processors, according to nine sources familiar with the company’s plans. agency.
The dominant global designer and supplier of AI chips aims to capture a piece of a growing market for custom AI chips and protect itself from the growing number of companies interested in finding alternatives to its products.
The Santa Clara, California-based company currently controls about 80% of the market for high-end AI chips, a position that has caused its market value to rise 40% this year to reach 1, 73 trillion dollars, after more than tripling in 2023.
Its customers, including ChatGPT maker OpenAI, Microsoft (NASDAQ:), Alphabet (NASDAQ:) and Meta Platforms (NASDAQ:), have raced to grab Nvidia’s dwindling supply of chips to compete in the gaming industry. rapidly expanding generative artificial intelligence.
Nvidia’s H100 and A100 chips serve as a generalized, general-purpose AI processor for many of these major customers. But tech companies have started developing their own in-house chips for specific needs. This helps reduce energy consumption and can potentially reduce costs and design time.
Nvidia is now trying to play a role in helping these companies develop custom AI chips that have then flowed into rival companies such as Broadcom (NASDAQ:) and Marvell (NASDAQ:) Technology, according to the sources, who declined to be identified because they do not they were. authorized to speak publicly.
“If you’re really looking to optimize things like power or optimize costs for your application, you can’t afford to throw an H100 or an A100 in there,” said Greg Reichow, general partner at venture capital firm Eclipse Ventures. in an interview. “You want to have the right combination of processing and just the type of processing you need.”
Nvidia doesn’t disclose prices for the H100, which are higher than the previous-generation A100, but each chip can sell for $16,000 to $100,000 depending on volume purchased and other factors. Meta said it plans to bring its total inventory to 350,000 H100s this year.
Nvidia officials met with representatives from Amazon.com (NASDAQ:), Meta, Microsoft, Google and OpenAI to discuss making custom chips for them, according to two sources familiar with the meetings. In addition to data center chips, the company has pursued customers in the telecommunications, automotive and gaming industries.
In 2022, Nvidia said it will allow third-party customers to integrate some of its proprietary networking technology with its own chips. The company has not spoken about the program since, and Reuters is reporting its broader ambitions for the first time.
An Nvidia spokesperson declined to comment beyond the company’s 2022 announcement.
Dina McKinney, a former executive at Advanced Micro Devices (NASDAQ:) and Marvell, heads Nvidia’s custom unit, and her team’s goal is to make its technology available to customers in cloud, 5G wireless, gaming and in the automotive industry, according to a LinkedIn profile. Those mentions were deleted and her headline changed after Reuters asked Nvidia for comment.
Amazon, Google, Microsoft, Meta and OpenAI declined to comment.
$30 BILLION MARKET
The market for custom data center chips will grow to $10 billion this year and double by 2025, Alan Weckel of research firm 650 Group estimates.
The broader custom chip market was worth about $30 billion in 2023, or about 5% of annual global chip sales, according to Needham analyst Charles Shi.
Currently, custom silicon design for data centers is dominated by Broadcom and Marvell.
In a typical deal, a design partner like Nvidia would offer intellectual property and technology, but leave the fabrication, packaging and additional steps of the chips to Taiwan Semiconductor Manufacturing Co. or another contract chipmaker.
Nvidia’s entry into this territory has the potential to impact Broadcom and Marvell’s sales.
“With Broadcom’s custom silicon business hitting $10 billion and Marvell’s at around $2 billion, this is a real threat,” said Dylan Patel, founder of silicon research group SemiAnalysis. “It’s a really big negative: There’s more competition entering the fray.”
BEYOND AI
Nvidia is in talks with telecom infrastructure builder Ericsson (BS:) for a wireless chip that includes the chip designer’s graphics processing unit (GPU) technology, according to two sources familiar with the talks.
650 Group’s Weckle expects the market for custom telecom chips to remain stable at around $4 billion to $5 billion annually.
Ericsson declined to comment.
Nvidia also plans to target the automotive and gaming markets, according to sources and public social media posts.
Weckel expects the custom car market to grow steadily from the current $6 billion to $8 billion range, by 20% annually, and the $7 billion to $8 billion custom video game chip market could increase with next-generation consoles of Xbox and Microsoft. Sony (NYSE:).
The current Nintendo Switch (NYSE:) the handheld console already includes an Nvidia chip, the Tegra X1. According to a source, a new version of the Switch console expected this year will likely include a custom Nvidia design.
Nintendo declined to comment.