Meta and NVIDIA Sign Multi-Year AI Infrastructure Deal

Meta has entered into a multi-year strategic partnership with NVIDIA to supply and build out the social media company’s AI data center infrastructure. The deal represents one of the largest infrastructure commitments between a technology company and a chip manufacturer, centered on advancing AI training and inference capabilities at an unprecedented scale.

Under the agreement, NVIDIA will supply Meta with millions of Blackwell and Rubin graphics processing units. These chips will power Meta’s next generation of data centers, purpose-built for AI workloads. Meta will also integrate NVIDIA Spectrum-X Ethernet switches into its Facebook Open Switching System platform, providing AI-scale networking with low-latency performance and improved power efficiency across its infrastructure footprint.

The partnership includes what NVIDIA has described as the first large-scale deployment using Grace CPUs without their standard pairing with Blackwell GPUs. Meta has additionally adopted NVIDIA Confidential Computing for WhatsApp, enabling AI-powered features within the messaging platform while maintaining user data confidentiality and integrity.

Engineering teams from both companies will collaborate across multiple generations of hardware and software to optimize AI models for Meta’s core workloads. The collaboration spans CPU, GPU, networking, and software layers, with an emphasis on performance per watt and operational efficiency at scale.

Meta has announced capital expenditure plans of between 115 billion and 135 billion US dollars for 2026, the majority of which will be directed toward data centers and computing infrastructure. Unlike hyperscalers that rent computing capacity to other businesses, Meta is retaining all of this infrastructure for its own operations, including powering its personalization and recommendation systems used by billions of users globally.

The scale of this deal carries broader consequences for the wider technology industry. IDC has projected that AI-driven demand for chips will significantly impact the IT market over the next two years. As large technology companies secure GPU supply at this volume, enterprises seeking to build or expand their own AI infrastructure face increasingly constrained access to NVIDIA processors. Businesses dependent on NVIDIA chips for AI development may be forced to seek alternative suppliers or delay planned projects.