Meta has entered a multi-year partnership with NVIDIA to expand its AI infrastructure, relying on a large-scale rollout of NVIDIA’s CPUs and GPUs.

The agreement covers on-premises and cloud-based data centres, as Meta aims to support its long-term AI infrastructure plans.

The collaboration will see Meta construct hyperscale data centres focused on both AI training and inference, deploying millions of NVIDIA Blackwell and Rubin GPUs.

NVIDIA Spectrum-X Ethernet switches will become integrated within Meta’s Facebook Open Switching System platform to support these operations.

Meta is also increasing the use of Arm-based NVIDIA Grace CPUs in its data centre applications, targeting improvements in performance per watt.

This marks the first significant deployment of NVIDIA Grace CPUs alone, backed by joint work on software optimisation and enhancements to CPU ecosystem libraries intended to boost efficiency with each generation.

Both companies are working together to prepare for the potential large-scale adoption of NVIDIA Vera CPUs from 2027, which is expected to further extend the energy-efficiency profile of Meta’s AI compute resources and contribute to developments across the Arm software sector.

NVIDIA founder and CEO Jensen Huang said: “Through deep codesign across CPUs, GPUs, networking and software, we are bringing the full NVIDIA platform to Meta’s researchers and engineers as they build the foundation for the next AI frontier.”

In addition, Meta will implement NVIDIA GB300-based systems across its on-premises facilities and through partnerships with NVIDIA Cloud Partners.

This unified architecture is intended to maximise scalability and performance while streamlining operational processes.

Meta has already applied the NVIDIA Spectrum-X Ethernet networking platform throughout its infrastructure for AI-focused networking requirements, aiming for low-latency connectivity and optimised power use.

Meta founder and CEO Mark Zuckerberg said: “We’re excited to expand our partnership with NVIDIA to build leading-edge clusters using their Vera Rubin platform to deliver personal superintelligence to everyone in the world.”

Meta has also introduced NVIDIA Confidential Computing into WhatsApp’s processing framework, enhancing privacy protections for users while enabling advanced AI capabilities within the messaging service.

The companies plan to extend these confidential computing features across more products in Meta’s portfolio.

Engineering teams from both firms are collaborating closely to co-design next-generation AI models that improve performance and efficiency for key workloads used globally.

These joint efforts combine Meta’s production-scale demands with advancements within NVIDIA’s technology stack.