Nvidia to invest $3 billion in Meta, expanding Muse AI collaboration

According to information released by both parties, Nvidia will invest $3 billion in Meta's stock issuance plan. Meta had previously announced plans to raise funds through a public offering and a subsequent share sale program to accelerate the construction of AI infrastructure. Meta then raised its overall fundraising target from $95 billion to $97.5 billion, making it one of the largest fundraising plans in the company's history.

Nvidia said the partnership focuses on improving the training and inference performance of Muse and its related models while reducing operating costs. The two parties plan to pursue deeper optimization for AI workflows to improve computing efficiency and improve total cost of ownership.

As part of the partnership, Meta will further expand its deployment of Nvidia's next-generation AI infrastructure, including the Vera Rubin system and Vera CPU, while adopting the Spectrum-X Ethernet networking solution and BlueField-4 data processing technology to improve data transmission efficiency, network throughput, and computing performance per unit of energy consumption.

In addition to hardware cooperation, the two parties are also evaluating the possibility of further adopting Nvidia's OpenShell reference architecture. The related cooperation will include using OpenShell to enable the secure deployment of AI agents, as well as using the Sentry tool to monitor and manage the behavior of AI agents, thereby improving the security and controllability of the system.

Nvidia CEO Jensen Huang said that AI agents are representing a new computing paradigm. This type of software can not only understand users' goals but also reason, invoke tools, and complete tasks on users' behalf. He believes that this shift is changing the way people use computers, moving users from operating applications to working collaboratively with intelligent agents.

Meta CEO Mark Zuckerberg said that Muse is a core component of Meta's strategy to promote "personal superintelligence." The company hopes to use this platform to bring advanced AI capabilities to billions of users and, together with Nvidia, build future-oriented personal intelligent agents and frontier models.

According to the introduction, Meta regards Muse as its next-generation personal AI assistant, which, with the user's authorization, can access various services such as email, calendars, payment services, health apps, shopping platforms, and smart home systems, and can perform practical tasks such as sending emails and arranging travel. Users can revoke the relevant permissions at any time.

Meta said that Muse and its model family have achieved leading performance in multiple benchmarks since their release, while user downloads continue to grow. However, the company did not disclose specific figures.

This investment has also once again sparked market discussion about the "circular financing" model. Because Meta will continue to purchase Nvidia AI hardware on a large scale in the future, outsiders believe that chip suppliers providing financial support to customers, who then purchase more chip products, is becoming an increasingly common business model in the AI industry.

Meta expects its capital expenditure for this year to be between $135 billion and $145 billion, the vast majority of which will be used to build AI infrastructure. At the same time, the company's latest financial report shows that its quarterly revenue grew 26% year over year to $64.6 billion. Analysts believe that as AI competition continues to heat up, the partnership between Meta and Nvidia will deepen further, and both parties are betting that AI agents will become the next-generation computing platform after the mobile internet.