Meta has announced a multiyear partnership with NVIDIA to scale its AI infrastructure across on premises and cloud environments.
Under the agreement, Meta will deploy NVIDIA CPUs and millions of Blackwell and Rubin GPUs to power hyperscale data centres focused on AI training and inference. The collaboration also includes integration of NVIDIA Spectrum-X Ethernet networking into Meta’s Facebook Open Switching System to improve network efficiency and throughput.
Meta is expanding deployment of Arm based NVIDIA Grace CPUs to improve performance per watt in its data centres. The companies are also planning future deployment of NVIDIA Vera CPUs to further enhance energy efficient AI compute.
Meta will use NVIDIA GB300 based systems to create a unified AI architecture across its infrastructure footprint. It has also adopted NVIDIA Confidential Computing for WhatsApp to enable AI powered features while protecting user data. The capability may expand across other Meta platforms.
Engineering teams from both companies are working on deep codesign to optimise large AI models and improve performance across Meta’s global services.
Leave a comment