Intel on Tuesday unveiled its new GPU, 'Crescent Island,' which emphasizes high energy efficiency and low cost for AI inference. Equipped with 160GB of LPDDR5X memory and based on the next-generation Xe3P microarchitecture, the card will begin customer testing in the second half of 2026 as part of Intel's strategy to build an open AI system architecture.
$Intel (INTC.US)$ On Tuesday, a data center GPU equipped with 160GB of memory and featuring high energy efficiency was announced, and it was added to the company’s AI accelerator portfolio, as part of Intel's new AI strategy centered around open systems and software architecture.
The GPU, codenamed 'Crescent Island,' is designed for air-cooled enterprise servers running inference workloads, emphasizing 'power and cost optimization.' Crescent Island is based on Intel’s Xe3P microarchitecture, which focuses on high performance per watt. It features 160GB of LPDDR5X memory and supports multiple data types, providing ample capacity for large language models (LLMs).
Intel’s announcement also noted that Crescent Island will support various data types and is positioned as 'highly suitable' for vendors offering tokens-as-a-service and AI inference use cases.
In addition to highlighting its energy efficiency, Crescent Island will feature an air-cooled design aimed at cost optimization. Intel is currently advancing its open-source software stack through its existing Arc Pro B-series GPUs in preparation for Crescent Island.
Intel stated that it plans to begin providing samples to customers in the second half of 2026. However, the company did not disclose an official release date — whether it will be launched before the end of 2026 remains unclear, though it is more likely to see a full-scale rollout in 2027. No product slides, prototypes, or further technical details have been released at this time.
Intel did not provide any updates on 'Jaguar Shores,' a next-generation GPU for rack-level platforms announced earlier this year.
At a press briefing last month, Intel’s Chief AI and Technology Officer, Sachin Katti, stated that Crescent Island offers 'enhanced memory bandwidth' and 'substantial memory capacity,' making it an 'ideal choice for token cloud services and enterprise-level inference scenarios.'
Crescent Island made its official debut at the 2025 OCP Global Summit, marking Intel's formal adoption of an annual GPU release cadence. This announcement came just a week after Intel heavily promoted its upcoming 'Panther Lake' and 'Clearwater Forest' CPUs, with the GPU launch being part of the same series of initiatives.
According to media reports, over the past two years, $NVIDIA (NVDA.US)$both Intel and AMD have shifted to an annual cadence of releasing new products. This move by Intel is also aimed at catching up. Over the past 15 years, Intel has experienced multiple failures in the field of accelerated chips, going through four CEOs, and has consistently failed to gain a foothold in the AI infrastructure market dominated by NVIDIA.
Sachin Katti, appointed by Intel CEO Lip-Bu Tan in April this year, is responsible for leading the company's new AI strategy. He stated that Intel is building a new vision for the AI hardware market centered around 'open systems and software architectures,' aiming to provide computational power that is 'appropriately scaled and cost-effective' to support future autonomous AI workloads.
He said:
"We will build scalable heterogeneous systems to provide a frictionless experience for agentic AI (autonomous AI) workloads, while leveraging open heterogeneous architectures to achieve optimal performance per dollar for these workloads."
Katti stated that this open strategy will offer customers and partners more choices at both the system and hardware levels, allowing multiple vendors to participate.
He added:
"As we continue to introduce more disruptive technologies, these new technologies can be seamlessly integrated into this open heterogeneous architecture."
AI Portfolio Strategist!One-click insight into holdings,Fully grasp opportunities and risks.
Editor/Joryn
