Intel confirmed its next-generation GPU IP, code-named Xe Next, will follow the recently launched Xe3P-based Crescent Island data center accelerators. The Xe Next lineup will include two distinct product lines: inference-focused Crescent Island GPUs and the Jaguar Shores design for AI training and high-performance computing (HPC) workloads. According to Intel Vice President Anil Nanduri, Jaguar Shores is nearing completion and could enter production by late 2026 or early 2027, pending yield improvements from its 18A manufacturing process. The design may combine Intel’s internal production with TSMC’s 5nm fabrication, though specifics remain unclear.

HBM4 memory, supplied by SK hynix, will power both GPU lines, aligning with industry trends toward high-bandwidth memory for AI workloads. This dual-path strategy positions Intel to compete directly with NVIDIA and AMD in the data center accelerator market, where margins remain exceptionally high.

The Xe Next roadmap suggests Intel is doubling down on data center AI infrastructure despite earlier skepticism about post-Xe3P innovation. While Crescent Island targets inference efficiency, Jaguar Shores aims to address the growing demand for generative AI training and scientific computing. Analysts speculate this bifurcation could strain Intel’s resources but may also diversify its revenue streams.

By expanding beyond inference chips, Intel seeks to capture a larger share of the AI accelerator ecosystem. However, the success of Jaguar Shores hinges on resolving fabrication challenges and securing long-term partnerships with memory suppliers like SK hynix. Industry watchers will monitor production timelines and performance benchmarks as the Xe Next era unfolds.