NVIDIA Plans $3 Billion Equity Investment in Meta as Muse Partnership Expands

Expanded hardware and software plans

Meta intends to enlarge its deployment of NVIDIA infrastructure, including Vera Rubin and Vera CPUs, to run its next generation of AI agents, with the stated aim of achieving the greatest possible performance per watt. The companies say they will jointly extend work on Muse and the wider model family to sharpen both inference and training speed.

The release also points to joint optimization across workflows, which the companies say is meant to raise performance and lower total cost of ownership. The arrangement covers NVIDIA OpenShell, which the release says supports secure deployment, and NVIDIA Sentry, which it describes as a means of observing how agents behave.

Link to the February agreement

The new announcement rests on a multi-year strategic partnership Meta revealed with NVIDIA on February 17, 2026, which provides for large-scale use of NVIDIA technology in data centres handling AI training and inference as well as core business work. According to the release, the latest steps reinforce that long-term infrastructure relationship.

NVIDIA says Meta's Muse models and agents run on its infrastructure and have, in the companies' characterisation, reached frontier-level benchmark results and seen strong consumer downloads since launch. The firms frame the work as part of a move in which people stop operating individual applications and instead work alongside intelligent agents.

Open questions

• When is the proposed $3 billion equity investment expected to close?
• What performance-per-watt gains will the expanded deployment deliver in practice?