Intel’s Upcoming Falcon Shores AI GPUs Scheduled for Release in Late 2025

Release Date

Intel has announced that its next-generation Falcon Shores AI GPUs are scheduled to be released in late 2025. While the precise time frame for broad market availability may extend into 2026, the anticipation surrounding these GPUs is already considerable in the tech community.

Power Consumption

A notable characteristic of the Falcon Shores GPUs is their high power consumption, reaching up to 1500W. This marks a significant increase compared to previous models and competitors like Nvidia’s Blackwell B200 GPU. Consequently, Falcon Shores will be exclusively water-cooled, with no air-cooled variants in sight, underscoring the need for advanced cooling solutions to manage such power-intensive hardware.

Manufacturing Process

The chips for Falcon Shores will be manufactured using TSMC’s cutting-edge 3nm process node. They will also utilize CoWoS-R advanced packaging technology, highlighting a significant collaborative effort between Intel and TSMC. This manufacturing approach is aimed at achieving superior performance and efficiency.

Architecture and Integration

Falcon Shores will feature a hybrid architecture by merging Intel’s existing Gaudi 3 technology with the Xe architecture from Ponte Vecchio. This integration is designed to provide a more robust platform for deep learning, capitalizing on the Gaudi AI IP’s capabilities within Falcon Shores to enhance AI processing power.

Market Strategy

Intel is aiming to position Falcon Shores as a cost-effective alternative to Nvidia’s offerings in the AI hardware space. The strategy includes integrating Ethernet to maintain competitive costs and presenting a comprehensive product portfolio that spans high-end, mid-range, and low-end options, thus catering to a diverse range of client needs.

Impact of Restructuring

Despite undergoing significant restructuring, Intel remains committed to advancing its Falcon Shores and Gaudi AI GPU lineups. These AI-focused investments are deemed crucial for securing Intel’s competitiveness and innovative edge within the data center and AI markets.

Performance and Competition

Falcon Shores is designed to directly compete with Nvidia’s Blackwell products, promising high-performance computing capabilities tailor-made for AI applications. A standout feature is the integration of x86 CPU cores with Xe-HPC GPU cores into a single socket, a strategic move intended to maximize processing power and efficiency.

Networking and Open Technologies

With Falcon Shores, Intel places significant emphasis on utilizing Ethernet and embracing open technologies, as advocated by the Ultra Ethernet Consortium. This approach aims to ensure reliable and secure high-speed communications for AI operations, presenting a distinct alternative to Nvidia’s more proprietary and integrated solutions.