Meta sells $4B in stock to Nvidia, AMD as AI infrastructure deal expands

Meta Platforms will raise up to $95 billion in a stock sale that includes $4 billion in direct purchases by Nvidia and AMD, the companies disclosed today. Nvidia will take $3 billion of the equity; AMD the remaining $1 billion. Broadcom is expected to contribute another $1 billion, bringing the total from the three chipmakers to $5 billion.

The transaction follows a recent stock split by Meta, which was structured to raise capital while preserving voting control. The scale of the equity sale suggests significant near-term spending on AI hardware, though the company has not disclosed specific budget figures.

Nvidia’s $3 billion purchase is tied to expanded hardware deployments. Meta plans to increase its use of Nvidia’s Rubin GPUs and Vera CPUs, which are often paired in high-density AI training systems. The deal also includes networking upgrades, with Meta set to expand its use of Nvidia’s Spectrum-X Ethernet products—including SuperNICs and switches designed for AI workloads. These components are marketed as capable of detecting and responding to traffic anomalies at sub-microsecond intervals. Meta will also adopt Nvidia’s BlueField-4 data processing units, which accelerate networking, storage, and security tasks.

Beyond hardware, the companies have outlined plans to collaborate on software optimizations, including potential work involving Nvidia’s OpenShell Reference Architecture, a framework for running AI agents in isolated environments.

AMD’s $1 billion investment is connected to a separate infrastructure agreement. Meta has begun deploying AMD’s Helios racks in its data centers, with testing currently underway. The systems are built on AMD’s Instinct MI450 architecture, Epyc CPUs, and Pensando data processing units. Meta has indicated plans for further deployments, and future adoption of a custom AI accelerator based on AMD’s Instinct architecture may follow. AMD has stated that insights from Meta’s deployments could inform its future chip and rack designs.

The arrangement mirrors a growing trend in which chipmakers are deepening their relationships with major AI adopters. Recent months have seen similar partnerships between hyperscalers and semiconductor firms, though few have involved direct equity investments of this scale. While Nvidia has prioritized stock buybacks over startup acquisitions in recent quarters, AMD’s recent $8.2 billion purchase of World Labs and reports of Nvidia’s interest in Reflection AI suggest chipmakers remain selective about where they deploy capital beyond their core hardware businesses.

For Meta, the deals provide immediate funding while potentially strengthening its supply chain relationships. Whether other large AI adopters will pursue similar equity-for-hardware arrangements remains to be seen.

Sources: siliconangle.com