Intel and IBM Cloud join forces with Gaudi 3 | Hardware store | 10 parts of computer | Types of hardware | Turtles AI

Intel has found its first cloud customer for the Gaudi 3 AI accelerator chip: IBM Cloud. The announcement marks a significant step for Intel in the AI chip market, where Nvidia dominates unchallenged. IBM Cloud will begin offering Gaudi 3 chips to customers early next year, with a focus on hybrid and on-premise environments. This collaboration between Intel and IBM highlights the importance of an open and collaborative ecosystem to develop innovative and affordable AI solutions.

Key Points:

• Gaudi 3 on IBM Cloud: IBM Cloud will be the first cloud provider to offer Intel’s Gaudi 3 chips, with availability planned for early 2024.
• Focus on Watsonx: IBM will integrate Gaudi 3 into its Watsonx data and AI platform, strengthening the platform’s AI capabilities.
• Competition and performance: Intel aims to compete with Nvidia and AMD in the AI chip market, despite financial and market difficulties.
• Technical innovations: Gaudi 3 includes new Ethernet connectivity technologies and Intel’s Xeon processors aimed at improving data center performance.

IBM Cloud will offer Intel’s Gaudi 3 AI accelerator chip starting in early 2024, marking an important step for Intel in the AI chip market. The official announcement came following a joint statement by the two companies outlining how Gaudi 3 accelerators will be available for hybrid and on-premise cloud environments. In addition, IBM disclosed that it plans to integrate support for Gaudi 3 within its Watsonx platform, focused on data and AI.

This move represents a major milestone for Intel, which is seeking to expand its presence in the AI chip market dominated by Nvidia and, to a lesser extent, AMD. Unveiled in December 2023, Gaudi 3 is the latest evolution of the series of accelerators developed by Habana Labs, which was acquired by Intel in 2019 for $2 billion. Gaudi 3 accelerators are designed to compete directly with Nvidia’s AI solutions, offering competitive performance and potentially lower costs. Early published benchmarks have highlighted the chip’s cost-effectiveness, but Intel will face the challenge of convincing customers already loyal to Nvidia to switch to its products.

In parallel with the announcement of the collaboration with IBM Cloud, Intel shared reference designs for Gaudi 3-equipped servers, which will be used by partners such as Lenovo, Dell, HPE, and Super Micro. These designs include significant technical innovations, such as new Ethernet connectivity designed to compete with Nvidia’s InfiniBand and integration with Intel’s Xeon 6 processors.

Despite its ambitions, Intel is facing a complex period. In April, the company estimated revenues of $500 million from Gaudi 3 in 2024, a modest figure compared to AMD’s projected $3.5 billion with its Instinct MI300 GPU and Nvidia’s expected $40 billion for its data center business. Nvidia’s dominance is set to strengthen further with the launch of its next AI chip, Blackwell, which is expected to offer up to four times the performance of the H100 model, against which Gaudi 3 stands as a direct competitor.

The optimism expressed by Intel, with CTO Greg Lavender stating in July that the company could aim for second place in the AI chip market, has been scaled back by financial reality. After posting a $1.6 billion loss in the second quarter of 2023, Intel announced a cost-cutting plan that includes cutting 15,000 jobs and projected savings of $10 billion by 2025. Nvidia’s delayed production of Blackwell gives Intel a limited but decisive window of opportunity to establish Gaudi 3 in the market.

 The alliance between Intel and IBM Cloud to implement Gaudi 3 represents a crucial attempt to strengthen Intel’s position in the AI field, albeit against a backdrop of severe competitive and financial challenges.