Jotunn8 AI processor highlights HBM3E in inference ASIC design

GUC is showcasing VSORA's Jotunn8 inference processor as a data-center AI chip built with advanced packaging and HBM3E integration. The design combines chiplet architecture, die-to-die connectivity, and TSMC CoWoS-S packaging to address the memory-wall problem in large-scale AI inference. For RamTrend, this is another sign that HBM demand is not limited to the largest training accelerators. If inference ASICs increasingly pair custom compute with HBM3E, the buyer base for high-bandwidth memory can broaden across more AI hardware programs. The article does not disclose production volume, so the price impact remains directional.