IBM and AMD Collaborate to Launch AI Accelerator Service by 2025

Generative AI inferencing involves computationally intensive tasks such as producing text, images, and other outputs from trained models

Tech giants IBM and Advanced Micro Devices (AMD) have announced a partnership to introduce AMD Instinct MI300X accelerators as a service on IBM Cloud, set to launch in the first half of 2025. The collaboration is poised to enhance performance and efficiency for generative artificial intelligence (AI) models and high-performance computing applications, catering to business customers. Key to this initiative is the integration of AMD Instinct MI300X accelerators into IBM's watsonx AI and data platform, alongside AI inferencing support in Red Hat Enterprise Linux. With 192GB of HBM3 memory, AMD Instinct MI300X accelerators are optimized for large-scale model inferencing and fine-tuning, enabling clients to run larger models with fewer GPUs, potentially reducing inferencing costs. This expanded capability will empower IBM watsonx clients with a robust AI infrastructure, scaling workloads seamlessly across hybrid cloud environments.

Generative AI inferencing involves computationally intensive tasks such as producing text, images, and other outputs from trained models. These operations demand high-performance solutions, particularly in real-time applications, making the IBM-AMD offering timely and impactful. The announcement comes amid reports that AMD plans to reduce its global workforce by 4%, impacting approximately 1,000 employees. The layoffs, revealed earlier this month, aim to strengthen AMD’s position in the competitive AI chip market, currently dominated by NVIDIA.

The IBM-AMD partnership marks a significant step forward in democratizing access to advanced AI capabilities, driving innovation and efficiency across industries.