Challenge Nvidia and AMD! Intel Unveils New Affordable AI Chip "Crescent Island," Focused on Air Cooling and Low Cost

Intel Challenges Nvidia and AMD on Price and Cooling! According to foreign media reports, Intel plans to ship a new AI chip codenamed "Crescent Island" by the end of 2026. Unlike mainstream high-end AI chips that pursue maximum computing power, this chip focuses on the "AI inference" market, and by adopting more affordable LPDDR5 memory and traditional air cooling design, it significantly reduces data center construction costs. This is an important step for Intel to revitalize its AI product line after new CEO Lip-Bu Tan took office.
(Background recap: Breaking news » Apple and Intel form a shocking alliance! Rumors say Intel will OEM chips for Apple, causing stock prices to soar 14% to a record high)
(Additional background: Intel (INTC) stock skyrocketed 11% to a 52-week high! Market value surpasses $530 billion)

Intel is adjusting its artificial intelligence hardware strategy, trying to break through in the AI chip market with more cost-effective solutions.

According to a report by the Financial Times on June 1, 2026, Intel plans to begin limited shipments of a new AI chip called "Crescent Island" by the end of this year. Unlike Nvidia’s dominant model training market, this chip will be entirely focused on AI inference tasks, which involve processing computations when end users make actual requests.

Dropping HBM and liquid cooling, Crescent Island emphasizes low deployment costs

Kevork Kechichian, head of Intel’s Data Center Group, said the company is "starting from the basics," focusing on the inference market to avoid past failures in the training market. To offer differentiated value in a highly competitive market, Crescent Island has chosen a hardware architecture that is completely different from Nvidia and AMD flagship products:

| Hardware Specification Differences | | --- | | Intel Crescent Island | | Nvidia / AMD High-End Competitors | | --- | --- | --- | | Memory Architecture | LPDDR5 (lower cost, ample supply) | HBM high-bandwidth memory (costly, limited capacity) | | Cooling Requirements | Traditional air cooling | Liquid cooling infrastructure | | Main Application Scenarios | AI inference | Model training and large-scale inference |

By adopting air cooling and LPDDR5 memory, data center customers will not need to invest heavily in liquid cooling infrastructure, greatly reducing the total cost of ownership (TCO) of AI servers.

New CEO Takes Charge, Intel Evaluates Launching China-Specific Version

This product is the first AI infrastructure product launched after former CEO Pat Gelsinger stepped down last year and Lip-Bu Tan took over as the new CEO. The development cycle of Crescent Island was only 18 months. Prior to this, Intel had canceled the development of its successor to the Gaudi GPU last year due to disappointing sales. Under Lip-Bu Tan’s significant cost-cutting and strategic reshaping, Intel’s stock has rebounded strongly by over 200% this year.

In terms of production and sales strategy, Intel plans to keep Crescent Island manufactured in its own foundries (Intel Foundry), aiming to reduce dependence on TSMC and further cut costs.

Additionally, focusing on the huge AI demand in the Chinese market, Kechichian confirmed that Intel is actively evaluating whether to launch a version of Crescent Island that complies with U.S. export control regulations. Since Nvidia and AMD’s high-end AI chips are still restricted by U.S.-China trade bans, Intel believes its affordable inference-focused products will have a strong opportunity to fill the gap in the Chinese market.