Qualcomm has validated first-generation High Bandwidth Compute (HBC) on working silicon at its San Diego lab. Testing exercised the complete memory read-write path and near-memory computations.
Engineers used industry-standard automated test equipment with randomized weights and activations. The data paths and dimensions were representative of production. HBC places compute closer to memory to reduce data movement when running AI models.
Qualcomm expects Dragonfly AI250 to begin sampling in the first half of 2027. The platform will integrate HBC into an AI inference accelerator and bring the technology into Dragonfly Rack infrastructure.
