Meta Platforms, Inc. and NVIDIA announced a multi-year strategic cooperation agreement on Tuesday. Under the agreement, Meta will utilize millions of NVIDIA chips in its own data centers and through cloud service providers. The specific terms of the deal were not publicly disclosed.
Meta has long been a customer of NVIDIA, having already deployed NVIDIA GPUs in its own data centers and renting related products from cloud service providers like Oracle. However, this new collaboration encompasses several NVIDIA technologies, including the upcoming Vera Rubin chip and its proprietary networking technology, signifying a move towards deeper cooperation. The two companies also stated that their engineering teams will work together on the joint design of Meta's AI models.
Meta had previously considered using AI chips from competitors. In November, The Information reported that Meta was in discussions with Google, planning to invest in using Google TPUs in its own data centers by 2027 and to rent Google chips via Google Cloud this year. It is currently unclear whether the new agreement with NVIDIA will affect the TPU-related negotiations. A Meta spokesperson said "there is no further information to share" and did not respond to questions about whether the talks are still ongoing.
NVIDIA stated that Meta will continue to use NVIDIA's Grace central processing units and is considering a large-scale deployment of the upcoming Vera CPU in 2027. (CPUs are traditional data center chips used for running non-AI related tasks.)
Comments