Meta Orders Millions of Nvidia Chips to Power Massive AI Infrastructure Expansion

Nvidia has concluded a long-term supply contract with Meta Platforms that provides for the sale of millions of AI chips over several years. The agreement encompasses both current and future processor generations from the semiconductor manufacturer and puts Nvidia in direct competition with established CPU providers such as Intel and AMD. The chip company announced the partnership in mid-February 2026, but did not disclose a specific contract value.

The agreement includes the current Blackwell architecture as well as the upcoming Rubin generation of AI accelerators. Additionally, Nvidia is supplying standalone installations of its Grace and Vera central processors, which are based on ARM technology. The company initially introduced these CPUs starting in 2023 as companions to its AI chips, but is now pursuing expansion into additional market segments, including the operation of AI agents and everyday technical tasks such as database management.

Meta is building massive data centers

Meta plans to build an AI data center called “Hyperion” in Holly Ridge in the U.S. state of Louisiana for approximately 10 billion dollars, which is intended to be the size of a quarter of Manhattan. This aggressive strategy has also led to a rift between Mark Zuckerberg and his longtime AI Chief Yann LeCun. LeCun believes that current Transformer models will not lead to superintelligence and that World Models based on a different architecture must be developed instead; he has already left Meta and founded a corresponding startup with AMI Labs.

Meanwhile, Meta has come to the conclusion that it must build significantly more computing power to improve AI models. Another factor is inference: if billions of users worldwide use Meta AI multiple times daily, corresponding computing power is needed to process that. In general, Meta has fallen far behind the competition with its own AI models and is likely to want to abandon the path of open LLMs that it has taken with Llama.

Estimated volume of 50 billion dollars

Industry analysts estimate the total value of the agreement at approximately 50 billion U.S. dollars. Meta ranks among the four largest customers, which accounted for approximately 61 percent of Nvidia’s revenue in the most recent business quarter. With the deal, the graphics card manufacturer secures significant business in an environment of growing competition from alternative processor architectures. According to Nvidia, the Grace CPUs promise approximately half the power consumption for certain common tasks, while the successor generation Vera is expected to exceed these efficiency gains even further.

The announcement comes against the backdrop of earlier reports of possible negotiations between Meta and Google. In November 2025, reports of the Facebook company’s billion-dollar investments in Google’s Tensor Processing Units briefly caused Nvidia stock to decline. To date, however, these plans have not materialized, and Google continues to use its TPU chips primarily in its own data centers. Meta is developing its own AI semiconductors in parallel, but apparently continues to rely heavily on external suppliers.

Energy efficiency becomes a critical factor

The power consumption of massive data centers is increasingly becoming a limiting factor for AI infrastructure expansion in the United States. Some states are already considering restrictions on new data centers. Meta emphasizes that it is building the world’s largest AI infrastructure and is relying on Nvidia’s technology to significantly increase performance per watt. In addition to the AI accelerators, the social media company has also ordered extensive quantities of Ethernet switches from Nvidia to optimize the networking of its data centers.

The announcement was made strategically one week before the release of Nvidia’s quarterly results and signals continued demand for the chip company’s products. Meta CEO Mark Zuckerberg formulated the ambitious goal of providing personal superintelligence to every person in the world using the ordered technology. For Nvidia, the contract demonstrates that the company can defend its dominant position in the AI chip market despite growing competition and in-house developments by hyperscalers.