First AI PC platform built on 18A

Intel CEO Lip-Bu Tan holds a wafer of CPU tiles for the Intel Core Ultra series 3, code-named Panther Lake, at the Intel Ocotillo campus in Chandler, Arizona. Panther Lake is the first client system-on-chips (SoC) built on the Intel 18A process node. (Credit: Intel Corporation)

Intel’s newly announced Core Ultra Series 3 “Panther Lake” architecture represents a fundamental redesign of the client computing platform—one that aligns the evolution of PC performance with the acceleration demands of AI workloads. Fabricated on the Intel 18A process node, Panther Lake integrates transistor-level innovations, scalable heterogeneous compute resources, and next-generation power delivery technologies that together define the next era of energy-efficient, AI-capable systems.

18A: The Foundation for Next-Gen Performance

At the heart of Panther Lake lies Intel’s most advanced process node—18A—introducing two key technologies:

• RibbonFET, Intel’s gate-all-around transistor, which replaces FinFET to improve electrostatic control, switching speed, and scaling efficiency.

• PowerVia, a backside power delivery network that separates signal and power routing to reduce IR-drop and noise while improving current efficiency.

Together, these enable up to 15 % higher performance per watt and 30 % greater transistor density compared to Intel 3. For system designers, this translates to higher sustained frequencies under the same thermal envelope and more flexible die stacking options for AI-integrated workloads.

Scalable Multi-Chiplet Architecture

Panther Lake extends Intel’s disaggregated architecture strategy with a modular, multi-chiplet design. Each package can combine compute, graphics, and AI accelerators based on application needs—whether in ultrabooks, workstations, or edge compute systems.

Key compute features include:

• Up to 16 performance cores (P-cores) and efficiency cores (E-cores) based on next-generation microarchitecture.

• A redesigned Intel NPU capable of exceeding 100 TOPS (trillions of operations per second) of AI performance.

• Integrated Intel Arc™ graphics with up to 12 Xe cores, supporting real-time rendering, media processing, and AI-assisted image upscaling.

This heterogeneous compute fabric allows workloads to dynamically shift between CPU, GPU, and NPU depending on data characteristics and latency requirements—crucial for AI inference tasks that mix matrix, vector, and scalar operations.

Packaging and Power Delivery Innovations

Manufactured at Fab 52 in Chandler, Arizona, Panther Lake leverages Intel’s advanced EMIB and Foveros packaging technologies to interconnect chiplets with high-bandwidth, low-latency interfaces. Combined with PowerVia’s backside power grid, designers achieve cleaner signal paths, reduced parasitic losses, and better thermal management—key factors for sustaining high AI throughput under mobile form-factor constraints.

Enabling the AI-PC Ecosystem

The rise of AI-centric computing is driving a paradigm shift in client platform design. Panther Lake is Intel’s most comprehensive response to this transformation. Its balanced XPU (CPU + GPU + NPU) composition enables local execution of generative AI, vision processing, and contextual computing workloads without full dependency on cloud services—improving privacy, responsiveness, and energy efficiency.

Beyond PCs, the same 18A node will power Xeon 6+ “Clearwater Forest” processors for data centers and AI inference accelerators. This node-level convergence simplifies cross-platform software optimization, allowing developers to scale AI applications from edge to cloud using a consistent Intel toolchain and instruction set.

Toward Smarter, More Efficient Systems

Panther Lake symbolizes a clear engineering pivot—from raw performance scaling toward intelligent efficiency. It exemplifies how architectural synergy, packaging advancements, and AI-native compute units can coexist in a power-constrained environment. As device intelligence increasingly migrates from the cloud to the edge, Intel’s latest platform stands as both a technical milestone and a roadmap for the next decade of client-side computing.