Nvidia will invest $3 billion in Meta's equity offering, the two companies announced, as part of an expanded partnership on Meta's Muse family of models and the Muse personal AI agent. Alongside the stake, Meta plans to expand its use of Nvidia's next-generation Vera Rubin hardware.
Meta has since priced the offering it launched on Monday. The company raised the common-stock portion from $22.5 billion to $25 billion and set the price at $625 a share. A separate mandatory convertible preferred stock sale carries a gross value of $12.5 billion, bringing the expected total to as much as $37.5 billion before options and costs.
Nvidia's stake is expected to close once customary conditions are met. When Meta first launched the sale, Nvidia, AMD and Broadcom had together indicated interest in buying up to $5 billion of the shares. The immediate offerings are part of a broader plan to raise up to $95 billion, which also includes a $60 billion at-the-market share program set to begin by early 2027.
“AI agents represent a new computing paradigm.”— Jensen Huang, Founder and CEO of Nvidia
Key facts
- 01Nvidia will invest $3B in Meta's equity offering, with closing subject to customary conditions.
- 02Meta raised the common-stock portion of its offering from $22.5B to $25B and priced shares at $625 each.
- 03A separate mandatory convertible preferred sale of $12.5B brings the expected total to up to $37.5B.
- 04Meta will expand deployment of Nvidia Vera Rubin GPUs, Vera CPUs, Spectrum-X Ethernet and BlueField-4.
- 05The companies will consider using Nvidia OpenShell and Sentry to deploy and monitor Muse AI agents.
On the infrastructure side, Meta intends to expand its deployment of Nvidia Vera Rubin GPUs and Vera CPUs for its next generation of AI agents. It will also roll out Nvidia Spectrum-X Ethernet networking and BlueField-4 to improve network efficiency and throughput. The work builds on an existing multiyear, multigenerational agreement that covers millions of Blackwell and Rubin GPUs.
The two companies will jointly optimize Muse training and inference workflows, with the stated goals of higher performance and lower total cost of ownership. Meta said its Muse models and agents, which already run on Nvidia infrastructure, have delivered frontier-level benchmark results and strong consumer download growth since launch.
“Muse is how we bring personal superintelligence to everyone.”— Mark Zuckerberg, Founder and CEO of Meta
Muse Spark, a multimodal reasoning model from Meta Superintelligence Labs that handles text, images, external tools and multiple agents, launched in April. The broader Muse agent followed in September and can book travel, fill in forms and make purchases on a user's behalf. It quickly became one of the most downloaded apps in the US.
Nvidia and Meta will also consider collaborating on the Nvidia OpenShell Reference Architecture. That would mean using OpenShell as a runtime for deploying agents safely and Nvidia Sentry for monitoring agent behavior.
The financing supports a capital spending plan that Meta expects to land between $135 billion and $145 billion in 2026. Meta also works with AMD, its own custom chips and outside cloud providers, and has been exploring selling excess compute capacity.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
