Meta, a leading AI company in the U.S., has announced plans to purchase additional units from Nvidia’s competitor AMD just a week after revealing its acquisition of AI semiconductors from Nvidia. This move appears to be a strategic effort to secure supplies in the increasingly competitive AI chip market.
According to the Wall Street Journal (WSJ), Meta and AMD announced on February 24 that AMD will supply Meta over the next five years with MI450 series graphics processing units (GPUs), EPYC central processing units (CPUs), and Helios server racks, which were first showcased at last month’s Consumer Electronics Show (CES 2026).
The hardware supplied by AMD will consume 6 gigawatts (GW) of power, based on the power usage of AI-driven data centers. These AI semiconductors are specialized GPUs designed for AI development and operation, combining special GPUs with high-bandwidth memory (HBM) chips.
The initial supply of 1 GW will begin in the second half of this year, with plans for gradual expansion. The WSJ estimates the contract value to exceed $100 billion. AMD’s CEO Lisa Su stated that the value per GW is in the tens of billions of dollars.
To secure Meta’s long‑term commitment, AMD has offered equity benefits. The WSJ reported that AMD will grant Meta options to buy up to 160 million shares, about 10% of its total stock, at 0.01 USD per share, based on actual purchase volumes and other conditions.
The WSJ also noted that Meta’s aggressive procurement of AI chips has intensified the semiconductor war among AI companies. On February 17, Meta signed a multi-year contract with Nvidia for millions of GPUs and CPUs. Local media reported in last October that Meta is also considering acquiring billions of dollars worth of Google’s Tensor Processing Units (TPUs) for its data centers by 2027.
Meta CEO Mark Zuckerberg emphasized that the AMD deal is aimed at building efficient inference computing and personal superintelligence, calling it a significant step in diversifying Meta’s computing resources.
The AI industry is currently facing a semiconductor shortage, prompting even Google, an AI service provider, to enter the AI chip market with its TPUs.
Mark Lipacis, an analyst at Evercore ISI, argued that Nvidia, which dominates the AI chip market, develops cutting-edge products but rarely increases production. AMD, while still behind Nvidia in overall AI GPU performance and market share, is steadily expanding its footprint in data center AI. OpenAI, a Google competitor, agreed to purchase 6 GW of AI chips from AMD last October.
