Qualcomm made an AI inferencing card with 128GB RAM without using HBM: https://www.qualcomm.com/news/onq/2023/11/introducing-qualcomm-cloud-ai-100-ultra https:

Qualcomm made an AI inferencing card with 128GB RAM without using HBM: https://www.qualcomm.com/news/onq/2023/11/introducing-qualcomm-cloud-ai-100-ultra https://www.qualcomm.com/news/onq/2023/11/introducing-qualco... Would someone with “basic ‘knowledge’ of hardware” explain why a GPU cannot be made with the same memory configuration?