Meta has expanded its partnership with Nvidia to include millions of AI chips, including the company’s new standalone Grace central processing units and next-generation Vera Rubin systems, in a major move to accelerate its data center build-out. The multiyear deal, announced Tuesday, marks a significant escalation in the collaboration between the two tech giants, with Meta becoming the first company to deploy Nvidia’s Grace CPUs as standalone chips in its data centers—rather than pairing them with GPUs in traditional server designs. Meta CEO Mark Zuckerberg said the expanded partnership supports his vision of delivering personal superintelligence to people worldwide, a goal he first outlined in July. While financial details were not disclosed, analysts estimate the deal is worth tens of billions of dollars. Chip analyst Ben Bajarin of Creative Strategies noted that a substantial portion of Meta’s planned $135 billion AI spending in 2026 will go toward this infrastructure build-out. The partnership builds on a decade-long relationship, but this new agreement broadens the scope to include Nvidia’s full suite of technologies. Standalone Grace CPUs are designed to handle complex inference and agentic workloads, serving as a high-performance complement to Nvidia’s Grace Blackwell and upcoming Vera Rubin server racks. Meta’s deployment of these chips at scale is seen as a validation of Nvidia’s integrated CPU-GPU strategy. The Vera Rubin systems are expected to be deployed in 2027. In addition to processors, Meta will use Nvidia’s Spectrum-X Ethernet switches to connect GPUs in large-scale AI clusters and integrate Nvidia’s security features into AI-powered tools on WhatsApp. The deal is part of Meta’s broader $600 billion investment in U.S.-based infrastructure by 2028, including plans for 30 data centers—26 of which will be in the United States. Two of its largest AI facilities are currently under construction: the 1-gigawatt Prometheus site in New Albany, Ohio, and the 5-gigawatt Hyperion site in Richland Parish, Louisiana. Despite its heavy reliance on Nvidia, Meta is not exclusively dependent on the company. In November, reports surfaced that Meta was exploring the use of Google’s tensor processing units in 2027. The company also designs its own custom silicon and uses AMD chips, which recently won a major deal with OpenAI as AI firms seek alternatives amid Nvidia’s supply constraints. Nvidia’s current Blackwell GPUs remain in high demand and have been on backorder for months. With this new deal, Meta has secured a reliable supply of both existing and next-generation chips. Engineering teams from both companies will collaborate closely to co-develop and optimize AI models. Meta is currently working on Avocado, a new frontier AI model intended to succeed its Llama series. The latest Llama release earlier this year failed to impress developers, highlighting the pressure to deliver cutting-edge results. Meta’s stock has been volatile recently, dropping sharply in October after announcing aggressive AI spending, then surging 10% in January on stronger-than-expected sales guidance. The new Nvidia deal has boosted investor confidence, sending Meta and Nvidia shares higher in after-hours trading, while AMD shares fell about 4%.