Intel has officially put an end to the Falcon Shores project, which was intended to be its advanced GPU designed for high-performance computing and artificial intelligence applications.
This decision comes as Intel seeks to regain its footing after several disappointing product introductions and unprecedented financial losses, particularly as rivals like AMD and Nvidia continue to make significant strides.
During a conference call regarding the company’s fourth-quarter earnings, Intel co-CEO Michelle Johnston Holthaus announced that the company would use Falcon Shores solely as an internal test chip and would not proceed with a public release.
“The AI data center market is an appealing area for us,” Holthaus stated during the call. “[B]ut my current sentiment is one of dissatisfaction. We are not yet significantly involved in the cloud-based AI data center sector… One immediate step we are taking is to streamline our roadmap and focus our resources more effectively.”
The company’s attention will shift to Jaguar Shores, which Holthaus referred to as Intel’s chance to “create a comprehensive solution at rack scale… to better serve the AI data center market as a whole.”
Holthaus had previously curbed expectations for Falcon Shores, suggesting that it represents an “iterative” advancement over the firm’s earlier dedicated AI data center chip, Gaudi 3.
“We learned from the Gaudi experience that delivering just the silicon is insufficient,” Holthaus emphasized during the latest earnings call. “Falcon Shores will assist us in refining our approach to systems, networking, memory, and all related components. However, our customers are primarily interested in a comprehensive rack solution, which we aim to achieve with Jaguar Shores.”
The Gaudi 3 chip has generally been viewed as a failure for Intel. In November, the company stated it would not hit its target of $500 million in Gaudi 3 sales due to software-related complications. Currently, only a handful of major service providers, aside from IBM, have committed to implementing the chip.
Intel confronts substantial challenges in the AI data center chip segment. Competitor AMD anticipates generating approximately $7 billion in AI chip revenue by 2025, while Nvidia, the leading competitor, is projected by some analysts to reach $195 billion in revenue by fiscal 2026, according to various forecasts.
“As I reflect on our AI prospects, my emphasis is on understanding the challenges our customers face, particularly regarding the need to reduce costs and enhance computational efficiency,” Holthaus remarked. “Consequently, a universal solution will not be effective, and I see distinct opportunities to utilize our core resources in innovative ways to offer the most attractive total cost of ownership.”
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
