Intel is showcasing three separate chip architectures aimed at supporting agentic AI and enterprise-scale computing: the Diamond Rapids processor for high-performance orchestration, the Crescent Island GPU for efficient AI inference, and the Wildcat Lake system-on-chip for intelligent client and edge computing.
Pushkar Ranade, Intel's chief technology officer, said agentic AI is fundamentally changing how computing systems are designed and delivered, from the transistor and chip packaging level all the way up through full system architecture. He said the future depends on tightly integrating general-purpose compute with purpose-built acceleration, advanced packaging, and open chiplet technologies to build systems that can adapt to different workloads while staying within real-world constraints around power, cost, and deployment.
All three processors are built on Intel Foundry technologies, including the Intel 18A process family, Foveros Direct 3D packaging (a technique for stacking chip components vertically), and early adoption of UCIe, an industry standard for connecting chiplets from different sources together. Together, Intel says these capabilities provide a differentiated, scalable foundation for next-generation AI infrastructure.
At Hot Chips 2026, Intel discussed each of these three architectures in detail. Diamond Rapids, the codename for Intel's next-generation Xeon processor, is designed as a flexible chip for enterprise-scale agentic AI. Built entirely on Intel's own technology, including the power- and performance-optimized Intel 18A-P process, Diamond Rapids introduces a new architecture combining adjustable compute building blocks, a unified memory system, and flexible input/output connections, aimed at delivering consistent high performance, efficiency, and scalability. Intel also highlighted the chip's advanced construction, using Foveros Direct 3D packaging and UCIe-S interconnects, along with a high-bandwidth memory subsystem, new Advanced Performance Extensions (APX), and enhanced Advanced Matrix Extensions (AMX) meant to accelerate next-generation AI workloads. Diamond Rapids will offer up to 256 cores with 1.28 GB of last-level cache, 16 memory channels running at 12,800 MT/s, and 128 lanes of PCIe Gen6 and CXL 3.0 connectivity.
Crescent Island is designed to change the economics of real-time AI inference by delivering more output tokens and greater business value per deployment. Described as a low-power, easy-to-deploy GPU with strong compute capability and high memory capacity, Crescent Island is intended to support larger AI models, longer context windows, and more simultaneous AI agents, all while fitting within existing air-cooled data center infrastructure rather than requiring specialized cooling. By improving the economics of inference at scale, Intel says the chip helps enterprises accelerate AI adoption, use their infrastructure more efficiently, and get better returns on AI investment. Built on Intel's established Xe graphics architecture, Crescent Island is optimized specifically for agentic AI workloads, aiming to sustain inference performance, maximize the rate of tokens processed, and reduce cooling requirements. Specifications include 32 Xe cores and 256 XMX engines based on the Xe3P architecture, up to 480GB of LPDDR5X memory, and a 350-watt, air-cooled PCIe card design.
The third architecture, Wildcat Lake, launched commercially as the Intel Core Series 3 processor line, is aimed at bringing right-sized AI capability to more affordable laptops and edge computing platforms. Built on the Intel 18A manufacturing process, Wildcat Lake combines new CPU cores optimized for strong single-thread x86 performance with integrated Xe3 graphics that include XMX acceleration, along with a neural processing unit (NPU) delivering up to 17 TOPS to support hybrid AI, meaning AI processing split between the CPU, GPU, and NPU. Wildcat Lake also marks the first time Intel has used UCIe interconnect technology in a shipping processor, enabling more cost-effective multi-chip package designs for mainstream AI-capable devices. It includes 2 performance cores and 4 efficiency cores, support for up to LPDDR5X-7467 memory, and Wi-Fi 7 and Bluetooth 6.0 connectivity.
Intel frames the broader significance of this lineup around the idea that agentic AI's future is inherently heterogeneous, spanning cloud, enterprise, and edge environments where balancing performance against cost is essential. Diamond Rapids provides a scalable compute foundation, Crescent Island addresses the cost of inference specifically, and Wildcat Lake brings AI capability into mainstream client and edge devices. Anchored by Intel Foundry's manufacturing technology and open industry standards, Intel is positioning this combined portfolio to support deploying agentic AI at scale.
