Intel's next-gen 'Jaguar Shores' Gaudi AI accelerator rumored to use new HBM4E memory

Intel's next-generation Jaguar Shores data center AI accelerator platform is rumored to be using newer HBM4E memory, which could launch sometime in the second half of 2027.

Back at the Intel AI Summit Seoul in South Korea back in July 2025, the company seemed to be ready for HBM4 from SK hynix for Jaguar Shores, with a release in 2026. However, Intel hasn't had a stable or successful time with its Gaudi AI accelerators and its release schedule, and it knows it has almost insurmountable competition from AMD and more so NVIDIA, so timelines can change, and specifications -- like using faster HBM4E -- can change, too.

The new information regarding Intel's use of HBM4E on its next-gen Jaguar Shores AI platform is from leaker @Bionic_Squash on X, who replied to @harukaze5719 regarding Jaguar Shores using HBM4, with a simply reply of "Jaguar is HBM4E".

• Read more: Intel kills off Falcon Shores AI chip for 'rack-scale solution' with Jaguar Shores

HBM4 will be rolling out in 2026 debuting with NVIDIA's next-gen Rubin AI GPUs and Vera Rubin AI platforms, with NVIDIA recently requesting that HBM suppliers SK hynix, Samsung, and Micron accelerate the development of 16-Hi HBM which should be HBM4E, for the second half of 2026.

• Read more: SK hynix confirms HBM4 + HBM4E memory are coming for next-gen AI GPUs
• Read more: Micron ships 11Gbps HBM4, also confirms that it is working on faster HBM4E
• Read more: Samsung accelerates HBM4E process: aims for 3.25TB/sec bandwidth ready for Rubin AI GPUs

NVIDIA will eat up most of that supply, and in the middle of a DRAM crisis that won't be ending for the foreseeable future, Intel securing bleeding-edge HBM4E memory chips seems a little odd, but I guess we'll see in the months ahead.