Intel Signs Awakening At AI Arena, New Chips Ready To Launch Next Year

JAKARTA - Intel Corp on Tuesday, October 14 announced plans to launch a new artificial intelligence (AI) chip for data centers next year. This step is the latest attempt by the United States chip giant to re-compete in the AI chip market now controlled by AMD and Nvidia.

The new chip, called Crescent Island, will be optimized for energy efficiency and supports various applications, including running an AI system or an inference process, Intel Chief Technology Officer Badminton Katti said at the Open Compute Summit in San Jose.

"This chip emphasizes our focus on optimized inference for AI, to provide the best performance per dollar and economic efficiency of the best token," Katti said, quoted by VOI from Reuters.

The launch of Crescent Island marks Intel's latest attempt to capitalize on a surge in global investment in AI that has generated billions of dollars for competitors such as AMD and Nvidia. However, Intel's move is still lagging behind competitors, reflecting the huge challenges the company faces to seize the chip market share and AI systems.

Intel CEO, Lip-Bu Tan, previously promised to revive the company's AI initiative after several projects, including the Gaudi chip line and the Falcon Shores processor, were suspended.

Crescent Island will be equipped with 160 memory gigabytes with speeds lower than the high bandwidth memory (HBM) used by AMD and Nvidia on their data center chips. The chip design will be based on Intel's consumer GPU architecture, although the company has not disclosed the manufacturing process used.

Since the emergence of a genrative AI phenomenon after the launch of ChatGPT by OpenAI in November 2022, demand for GPUs for data centers has surged sharply. Many cloud companies and startups are racing to secure chip supplies, triggering scarcity and high prices in the market.

At the exhibition, Katti added that Intel plans to launch an AI chip for data centers every year, following the annual pattern that AMD, Nvidia, and a number of major cloud service providers have also implemented.

Nvidia has so far dominated the chip market used to train large AI models such as ChatGPT. Tan stated that Intel will now focus more on chip designs designed to run these AI models behind-the-scenes systems so that AI software can operate.

"Instead of trying to build chips for all types of workloads, we will focus on inference," Katti said.

Intel also adopts an open and modular approach, allowing customers to combine chips from various vendors.

As part of a collaborative strategy, Nvidia recently invested $5 billion to buy about 4% of Intel's shares. This investment makes Nvidia one of Intel's biggest shareholders and opens future chip development partnerships for PCs and data centers.

According to Katti, the deal is an important step for Intel to ensure that its central processor (CPU) will be a major part of any AI system sold in the global market.