Advanced Micro Devices said on Thursday it plans to start mass production of a new version of its artificial intelligence (AI) chip called the MI325X in the fourth quarter of the year, as it seeks to strengthen its presence in a market dominated by Nvidia.
At an event in San Francisco, AMD CEO Lisa Su said the company plans to release its next-generation MI350 series chips in the second half of 2025. Those chips include increased amounts of memory and will have a new core architecture that AMD said will significantly improve performance over the previous MI300X and MI250X chips.
The announcements were largely expected based on AMD’s revelations earlier this year. They failed to cheer up investors, who caused AMD’s shares to fall by almost five percent in afternoon trading. Some analysts attributed the decline to a lack of major new cloud computing customers for the chips.
Shares of rival Nvidia rose 1.5 percent, while Intel fell 1.6 percent.
Demand for AI processors from big tech companies like Microsoft and Meta Platforms has far outstripped supply from Nvidia and AMD, allowing semiconductor companies to sell as much as they can produce.
That’s led to a huge rally in chip stocks over the past two years, with AMD shares up about 30 percent since a recent low in early August.
“So far, there are no new buyers announced,” said Summit Insights research analyst Kinngai Chan, adding that shares rose ahead of the event in anticipation of “something new.”
Santa Clara, Calif.-based AMD said vendors such as Super Micro Computer will begin shipping its MI325X AI chip to customers in the first quarter of 2025. AMD’s design aims to compete with Nvidia’s Blackwell architecture.
The MI325X chip uses the same architecture as the already available MI300X, which AMD launched last year. The new chip includes a new type of memory that AMD said will speed up AI calculations.
AMD’s next-generation AI chips are likely to put further pressure on Intel, which has struggled to implement a coherent AI chip strategy. Intel has predicted AI chip sales of more than $500 million (roughly Rs. 4,203 crore) in 2024.
New server, PC chips
AMD’s Su also said at the event that the company currently has no plans to use contract chipmakers other than Taiwan’s TSMC for advanced manufacturing processes, which are used to produce high-speed AI chips.
“We would like to use more capacity outside of Taiwan. We are very aggressive in using TSMC’s facility in Arizona,” Su said.
AMD also introduced several networking chips that help speed up data transfers between chips and systems inside data centers.
The company has announced the availability of a new version of the server’s central processing unit (CPU) design. The family of chips previously codenamed Turin includes a version of one designed to keep graphics processing units (GPUs) fed with data – which will speed up AI processing.
The flagship chip has nearly 200 processing cores and is priced at $14,813 (roughly Rs. 12.45 lakh). The entire line of processors uses the Zen 5 architecture, which offers a speed increase of up to 37 percent for advanced AI data crunching.
In addition to chips for data centers, AMD announced three new PC chips aimed at notebook computers, based on the Zen 5 architecture. The new chips are tuned to run AI applications and will be able to run Microsoft’s Copilot+ software.
In July, AMD raised its forecast for AI chips to $4.5 billion (roughly Rs. 37,834 crore) for the year from a previous target of $4 billion (roughly Rs. 33,630 crore). Demand for MI300X chips has skyrocketed due to the frenzy around building and deploying generative AI products.
This year, analysts expect AMD to report data center revenue of $12.83 billion (roughly Rs. 10,78,711 crore), according to LSEG estimates. Wall Street expects Nvidia to report data center revenue of $110.36 billion (roughly Rs. 9,27,877 crore). Data center revenue is a substitute for the AI chips needed to build and run AI applications.
Analysts’ rising earnings expectations kept AMD and Nvidia’s values in check despite the share jump. Both companies trade at more than 33 times their 12-month forward earnings estimates, compared with the benchmark S&P 500 at 22.3.
© Thomson Reuters 2024
(This story was not edited by NDTV staff and was automatically generated from a syndicated feed.)