In an attempt to take on market leader Nvidia, Advanced Micro Devices introduced its newest AI processors on Monday along with a roadmap for developing AI chips over the next two years.
The MI325X accelerator was unveiled by AMD CEO Lisa Su at the Computex technology trade expo in Taipei. It is scheduled for release in the fourth quarter of 2024.
Towards the bell, Nvidia’s stock increased by roughly 3%, while AMD’s gained 1.4%.
The drive to create generative AI applications has resulted in an enormous demand for the cutting-edge CPUs needed in AI data centers that can handle these intricate applications.
AMD has been fighting to challenge Nvidia, which now holds a commanding 80% market share in the lucrative AI chip sector.
AMD has now followed suit, with Nvidia having made it apparent to investors since last year that it intends to reduce its release cycle to once a year.
“AI is clearly our number one priority as a company and we have really harnessed all of the development capability within the company to do that,” Su told the media.
The reason behind this yearly cycle is that the market demands newer products with newer features. We always have the most competitive portfolio because we have the next big item every year.”
AMD also unveiled the MI350 chip series, which will be built on a revolutionary chip architecture and should go on sale in 2025.
AMD stated that it anticipates the MI350 to perform 35 times better in inference—the process of calculating generative AI responses—than the MI300 family of AI chips already on the market. AMD also unveiled the MI400 series, which will debut in 2026 and is built on the “Next” architecture.
CEO of Nvidia Jensen Huang announced on Sunday that GPUs, CPUs, and networking chips will be a part of the company’s next-generation AI chip platform, dubbed Rubin, which is scheduled for release in 2026.
Investors looking to assess the endurance of the huge genAI rally—which hasn’t showed any signs of slowing down—have been requesting longer-term updates from semiconductor companies. These investors have thrown billions of dollars into Wall Street’s picks-and-shovels trade.
Since the beginning of 2023, AMD’s stock has more than doubled. This increase is still insignificant when compared to the more than seven-fold increase in Nvidia’s stock price during the same timeframe.
In April, Su from AMD stated that the company anticipates selling $4 billion worth of AI chips by 2024, a $500 million increase over its previous projection.
AMD also said during the Computex event that the second half of 2024 will probably see the release of its most recent generation of central processor units.
Although AI chips are usually the top priority for enterprises when it comes to data centers, AMD’s CPUs are sometimes employed in conjunction with graphics processing units (GPUs), albeit the ratio is skewed in favor of GPUs.
The new neural processing units (NPUs) from AMD, which are intended to handle on-device AI tasks in AI PCs, have been described in full.
Chipmakers have been counting on increased AI capabilities to propel PC market expansion as it bounces back from a protracted downturn.
Devices with AMD’s AI PC chips will be released by PC manufacturers including HP and Lenovo. AMD claimed that their CPUs meet or surpass Microsoft’s Copilot+ PC specifications.
Shayne Heffernan
