At the 'Data Center and AI Technology Premiere' held on Tuesday, the leading semiconductor firm AMD disclosed its recent CPU and AI accelerator advancements in an effort to compete with Nvidia. Alisa Davidson This announcement comes at a crucial time as AI firms struggle with limited GPU availability amid the boom in generative AI. Throughout the event, AMD shared comprehensive insights into its strategic AI Platform.

The firm revealed extensive details about the new AMD Instinct™ MI300 Series of accelerators, including the introduction of the MI300X model, heralded as 'the world's most advanced accelerator for generative AI.' The MI300X is set to challenge Nvidia's leading H100 chipset.

At present, Nvidia retains a commanding lead in the GPU market, boasting over 80% market share. The H100 stands as the quickest GPU tailored for artificial intelligence, high-performance computing, and data analytics applications. April 29, 2025 by

Alisa Davidson

The H100 is notable for being the inaugural GPU powered by Nvidia's fourth-generation Tensor Cores, purpose-built to enhance AI training and inference efficiency. The performance of these Tensor Cores shows a remarkable increase, being up to seven times faster than those in the prior generation, specifically for models like GPT-3.

• Moreover, the H100 is equipped with Nvidia's innovative Hopper Memory, a high-bandwidth and low-latency memory architecture designed to optimize data-heavy tasks. This Hopper Memory is reported to be up to twice as fast compared to previous versions, making it the swiftest GPU memory solution available.
• Additionally, the H100 supports Nvidia's new Grace CPU architecture, aimed at facilitating exascale computing capabilities. Together, the H100 and Grace CPU can achieve performance levels that are tenfold greater than that of earlier generation configurations.
• The Nvidia H100 GPU features an impressive memory capacity of 188GB, which is currently unmatched in the GPU market.
• In terms of cost, Nvidia typically prices its chipset around $10,000, although it has surged to as much as $40,000 in some instances. In contrast, AMD has yet to disclose pricing for its new chipset.

With AMD's Instinct MI300X set to launch later this year, there is potential for it to disrupt Nvidia's H100 dominance in the marketplace. Reports suggest that April 29, 2025 AWS' cloud division is exploring the use of AMD's latest chip offerings.

The accelerator is loaded with 192GB of HBM3 memory, providing the necessary computational and memory efficiency for training and inference of extensive language models tailored for generative AI applications. Know More This substantial memory capacity can support large language models like Falcon-40, a 40 billion parameter model, all on a single MI300X accelerator.

Know More

• "AI is emerging as the pivotal technology that will define future computing, presenting AMD with its greatest opportunity for strategic growth. Our focus is unwavering as we work towards accelerating AMD AI platform deployments at scale in data centers, particularly highlighted by our Instinct MI300 accelerators coming this year and a growing ecosystem of enterprise-ready AI software fine-tuned for our hardware,\" stated AMD's Chair and CEO, Dr. Lisa Su.
• In conjunction with the chipset reveal, AMD introduced the ROCm software ecosystem, which consists of a suite of software tools and resources tailored for data center accelerators. During the event, they acknowledged partnerships with industry frontrunners that joined AMD for discussions.

The popular AI framework PyTorch partnered with AMD and the PyTorch Foundation to ensure the ROCm software stack is seamlessly integrated, allowing PyTorch 2.0 to support all AMD Instinct accelerators right from the start, giving developers access to a diverse array of AI models fueled by PyTorch on AMD hardware.

Read More

Additionally, Hugging Face, recognized for being an open platform for AI developers, has announced intentions to optimize thousands of their models for AMD systems specifically.

In May, AMD announced a revenue figure of $5.4 billion for the initial quarter of 2023, reflecting a 9% decline compared to the previous year. Following their AI strategy announcement, their stock saw an uptick of over 2% since market opening, currently sitting at $127 as of this report. Analysts from Barclays, Jefferies, and Wells Fargo have all adjusted AMD’s target stock price upward to a range of $140-$150.

Microsoft Edge Enhances Web Videos Using AMD and NVIDIA GPU

Please note that the information on this page should not be viewed as legal, financial, tax, or investment advice. It is crucial only to invest funds you can afford to lose, and if you have any uncertainties, consider seeking advice from an independent financial advisor. For additional information, we recommend reviewing the terms and conditions along with the support pages offered by the issuer or advertiser. MetaversePost is committed to delivering accurate and impartial reporting, but please understand that market dynamics can change unexpectedly.

Read more

• Cindy, a journalist at Metaverse Post, dedicates her writing to web3, NFTs, the metaverse, and AI topics, focusing heavily on interviews with key figures in the Web3 landscape. Having engaged with over 30 C-level executives, she brings their insights to a broader audience. Originally from Singapore, Cindy now resides in Tbilisi, Georgia, armed with a Bachelor's degree in Communications & Media Studies from the University of South Australia and a decade's worth of experience in the field of journalism.