Latest news with #TensorProcessingUnit
Globe and Mail
17-02-2025
- Business
- Globe and Mail
Meta Platforms and Google Just Shared Fantastic News for Broadcom Investors
Big tech is planning a massive spending spree in 2025. Four of the largest tech companies in the world are targeting a combined $320 billion in capital expenditures (capex) this year, and most of that spending will go toward building and refreshing artificial intelligence (AI) data centers for training and using generative AI. Nvidia (NASDAQ: NVDA) is sure to be a big winner from all of that spending. Practically all of the big tech companies reiterated the importance of their relationships with the leading designer of graphics processing units (GPUs). But comments from the management of Meta Platforms and Google parent Alphabet suggested a bright future for another chipmaker. Broadcom (NASDAQ: AVGO) could end up being a big winner from the increase in spending by big tech in 2025 and beyond. Here's what Meta and Google said, and what it means for Broadcom investors. The rise of custom chips Broadcom makes a lot of different types of semiconductors. When it comes to AI data centers, it plays an essential role. It makes chips for network switches, the devices that make sure data moves quickly and efficiently among servers. In effect, Broadcom's chips make sure Nvidia's chips are working as effectively as possible in Meta's and Google's data centers. But Broadcom also makes some other chips that Meta and Google deploy in their data centers. Its technology is behind Meta's custom AI chip, the Meta Training and Inference Accelerator (MTIA). Google, meanwhile, has been working with Broadcom since 2015 to develop its Tensor Processing Unit (TPU). Both Meta and Google had good things to say about their custom chips during their fourth-quarter earnings calls. During Meta's call, chief financial officer Susan Li told analysts: "In 2024, we started deploying MTIA to our ranking and recommendation influence workloads for ads and organic content. We expect to further ramp adoption of MTIA for these use cases throughout 2025." Over the long run, Meta sees its chips taking over more and more of its AI data centers from Nvidia and general GPUs. "Next year, we're hoping to expand MTIA to support some of our core AI training workloads, and over time, some of our GenAI use cases," Li said later in the call. Alphabet CEO Sundar Pichai said that in the fourth quarter, "We saw strong uptake of Trillium, our sixth-generation TPU." The company only launched its newest TPUs to the general public in December, giving developers less than three weeks to ramp up use before the end of the quarter. So, the comments suggest 2025 could be a big year for TPUs. Custom AI accelerator chips are more efficient than general-purpose GPUs for many types of AI workloads. As big tech scales up its data centers, eking out small efficiencies can have a big impact on the bottom line. As a result, expanding the uses for custom chips and having full control over the technology stack become increasingly important as capex budgets grow. Broadcom expects even stronger results long term Alphabet and Meta aren't the only companies partnering with Broadcom to build custom AI accelerators. TikTok parent company ByteDance is one of its three largest customers. And during the fourth-quarter earnings call, Broadcom management announced two new customers building their own next-generation chips with its help. Many believe those are Apple and OpenAI. Apple is notably a fan of Broadcom's chips. It used Google's previous-generation TPUs to train its Apple Intelligence system last year. A push by Apple to take more control of its AI training and development could quickly turn it into one of Broadcom's biggest customers. The company expects the market for custom AI accelerators to grow to between $60 billion and $90 billion within three years. Last year, Broadcom accounted for roughly 70% of custom AI accelerators and generated about $12 billion in sales. For reference, its total sales across its diversified operations for 2024 were just $51.6 billion, so the growth in AI chips alone could nearly double its revenue if it maintains its share of the market. But even with a bright future, valuation still matters. Broadcom's stock trades for 37 times forward earnings as of this writing. And while the chipmaker is poised to see very strong earnings growth over the next few years, it's hard to justify paying such a premium for the stock. If the stock price came down to a value closer to 30 times 2025 earnings, it may be worth buying considering analysts expect more than 30% earnings growth this year, giving it a price/earnings-to-growth ratio of around 1. Or if adoption of Broadcom's custom chips proves even stronger than expected at this point as Meta and Google look to expand their use in their data centers, it could end up growing into that valuation. Broadcom is certainly a company and a stock to keep an eye on for investors interested in the future of artificial intelligence. But right now, it's too expensive to recommend buying. Should you invest $1,000 in Broadcom right now? Before you buy stock in Broadcom, consider this: The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Broadcom wasn't one of them. The 10 stocks that made the cut could produce monster returns in the coming years. Consider when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you'd have $850,946!* Stock Advisor provides investors with an easy-to-follow blueprint for success, including guidance on building a portfolio, regular updates from analysts, and two new stock picks each month. The Stock Advisor service has more than quadrupled the return of S&P 500 since 2002*. *Stock Advisor returns as of February 7, 2025 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.
Yahoo
17-02-2025
- Business
- Yahoo
Meta Platforms and Google Just Shared Fantastic News for Broadcom Investors
Big tech is planning a massive spending spree in 2025. Four of the largest tech companies in the world are targeting a combined $320 billion in capital expenditures (capex) this year, and most of that spending will go toward building and refreshing artificial intelligence (AI) data centers for training and using generative AI. Nvidia (NASDAQ: NVDA) is sure to be a big winner from all of that spending. Practically all of the big tech companies reiterated the importance of their relationships with the leading designer of graphics processing units (GPUs). But comments from the management of Meta Platforms and Google parent Alphabet suggested a bright future for another chipmaker. Broadcom (NASDAQ: AVGO) could end up being a big winner from the increase in spending by big tech in 2025 and beyond. Here's what Meta and Google said, and what it means for Broadcom investors. Broadcom makes a lot of different types of semiconductors. When it comes to AI data centers, it plays an essential role. It makes chips for network switches, the devices that make sure data moves quickly and efficiently among servers. In effect, Broadcom's chips make sure Nvidia's chips are working as effectively as possible in Meta's and Google's data centers. But Broadcom also makes some other chips that Meta and Google deploy in their data centers. Its technology is behind Meta's custom AI chip, the Meta Training and Inference Accelerator (MTIA). Google, meanwhile, has been working with Broadcom since 2015 to develop its Tensor Processing Unit (TPU). Both Meta and Google had good things to say about their custom chips during their fourth-quarter earnings calls. During Meta's call, chief financial officer Susan Li told analysts: "In 2024, we started deploying MTIA to our ranking and recommendation influence workloads for ads and organic content. We expect to further ramp adoption of MTIA for these use cases throughout 2025." Over the long run, Meta sees its chips taking over more and more of its AI data centers from Nvidia and general GPUs. "Next year, we're hoping to expand MTIA to support some of our core AI training workloads, and over time, some of our GenAI use cases," Li said later in the call. Alphabet CEO Sundar Pichai said that in the fourth quarter, "We saw strong uptake of Trillium, our sixth-generation TPU." The company only launched its newest TPUs to the general public in December, giving developers less than three weeks to ramp up use before the end of the quarter. So, the comments suggest 2025 could be a big year for TPUs. Custom AI accelerator chips are more efficient than general-purpose GPUs for many types of AI workloads. As big tech scales up its data centers, eking out small efficiencies can have a big impact on the bottom line. As a result, expanding the uses for custom chips and having full control over the technology stack become increasingly important as capex budgets grow. Alphabet and Meta aren't the only companies partnering with Broadcom to build custom AI accelerators. TikTok parent company ByteDance is one of its three largest customers. And during the fourth-quarter earnings call, Broadcom management announced two new customers building their own next-generation chips with its help. Many believe those are Apple and OpenAI. Apple is notably a fan of Broadcom's chips. It used Google's previous-generation TPUs to train its Apple Intelligence system last year. A push by Apple to take more control of its AI training and development could quickly turn it into one of Broadcom's biggest customers. The company expects the market for custom AI accelerators to grow to between $60 billion and $90 billion within three years. Last year, Broadcom accounted for roughly 70% of custom AI accelerators and generated about $12 billion in sales. For reference, its total sales across its diversified operations for 2024 were just $51.6 billion, so the growth in AI chips alone could nearly double its revenue if it maintains its share of the market. But even with a bright future, valuation still matters. Broadcom's stock trades for 37 times forward earnings as of this writing. And while the chipmaker is poised to see very strong earnings growth over the next few years, it's hard to justify paying such a premium for the stock. If the stock price came down to a value closer to 30 times 2025 earnings, it may be worth buying considering analysts expect more than 30% earnings growth this year, giving it a price/earnings-to-growth ratio of around 1. Or if adoption of Broadcom's custom chips proves even stronger than expected at this point as Meta and Google look to expand their use in their data centers, it could end up growing into that valuation. Broadcom is certainly a company and a stock to keep an eye on for investors interested in the future of artificial intelligence. But right now, it's too expensive to recommend buying. Before you buy stock in Broadcom, consider this: The Motley Fool Stock Advisor analyst team just identified what they believe are the for investors to buy now… and Broadcom wasn't one of them. The 10 stocks that made the cut could produce monster returns in the coming years. Consider when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you'd have $850,946!* Now, it's worth noting Stock Advisor's total average return is 959% — a market-crushing outperformance compared to 178% for the S&P 500. Don't miss out on the latest top 10 list. Learn more » *Stock Advisor returns as of February 7, 2025 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. Meta Platforms and Google Just Shared Fantastic News for Broadcom Investors was originally published by The Motley Fool
Yahoo
05-02-2025
- Business
- Yahoo
Google software engineer accused of stealing AI technology for China
SAN FRANCISCO (KRON) — A former Google software engineer stole artificial intelligence (AI) technology from the Silicon Valley tech giant, prosecutors said Tuesday. A federal grand jury returned an indictment Tuesday charging 38-year-old Linwei Ding, also known as Leon Ding, with seven counts of economic espionage and seven counts of theft of trade secrets. Google, which is headquartered in Mountain View, hired Ding as a software engineer in 2019. Between 2022 and 2023, Ding uploaded more than 1,000 files containing confidential information from Google's network to his personal Google Cloud account, according to the indictment. 'While Ding was employed by Google, he secretly affiliated himself with two People's Republic of China (PRC)-based technology companies,' the U.S. Attorney's Office Northern District of California wrote. By mid-2023, Ding had founded his own technology company focused on AI and machine learning in China and was acting as the company's CEO, according to investigators. 'Ding intended to benefit the PRC government by stealing trade secrets from Google,' prosecutors wrote. Salesforce cutting 1,000 jobs, hiring salespeople for AI China sponsors talent programs incentivized people to engaged in research and development outside the PRC to transmit that knowledge and research back in exchange for salaries, research funds, lab space, or other incentives, according to federal investigators. Ding is accused of stealing technology from Google relating to: The hardware infrastructure and software platform that allows Google's supercomputing data center to train and serve large AI models. Trade secrets about the architecture and functionality of Google's Tensor Processing Unit (TPU) chips and systems and Google's Graphics Processing Unit (GPU) systems, the software that allows the chips to communicate and execute tasks, and the software that orchestrates thousands of chips into a supercomputer capable of training and executing cutting-edge AI workloads. Trade secrets on Google's custom-designed SmartNIC, a type of network interface card used to enhance Google's GPU, high performance, and cloud networking products. The case is being prosecuted by Assistant U.S. Attorneys Casey Boome and Molly Priedeman, as well as trial attorneys Stephen Marzen and Yifei Zheng of the National Security Division's Counterintelligence and Export Control Section. Copyright 2025 Nexstar Media, Inc. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.
