Intel announces Jaguar Shores as successor to Falcon Shores, unifying Gaudi ASICs and Xe-HPC GPUs in a single product line.

Intel has revealed Jaguar Shores as the next successor in its AI accelerator lineup, following Falcon Shores, according to HPCwire. Intel's Habana Labs division disclosed this codename during a technical workshop at the SC2024 conference. When inquired for additional details, Intel declined to comment, as most specifications are likely not yet finalized.

The announcement comes as the industry awaits the launch of Intel's next-generation dedicated Gaudi 3 ASICs, which have been pushed into 2025 to compete against Nvidia's last-generation Hopper-based H100. Intel's current lineup positions the Ponte Vecchio GPU in the world's third-fastest supercomputer, Aurora, sitting alongside HPC and AI GPUs. Intel originally merged its CPU and GPU families into a disaggregated XPU (CPU+GPU) codenamed "Falcon Shores" to streamline its AI offerings. However, those plans have been abandoned, and Falcon Shores will now arrive in a GPU-only configuration by late 2025.

Jaguar Shores is likely a GPGPU (General-Purpose GPU)—akin to Nvidia's B100, B200, and B300 chips—fabricated using an Angstrom-grade node (Intel 18A/14A). HPCwire suggests that Intel teased a Falcon Shores successor last year, slated for a 2026 launch. However, it remains uncertain whether Jaguar Shores will be able to compete against Nvidia's Rubin chips by that timeline.

While Intel's AI accelerators currently pale in contrast to AMD and Nvidia's offerings, the company shows potential signs of recovery in other segments. The latest Granite Rapids CPUs featuring upwards of 128-core configurations signal a potential resurgence for Intel, at least in servers and workstations. Nevertheless, Intel's future in AI still appears uncertain, despite the company's efforts to reserve an X3D-esque server cache to strengthen its data center position.