NVIDIA Invests $3 Billion in Meta’s Equity Offering as Muse Work Expands

NVIDIA said on October 26 that it will invest $3 billion in Meta’s equity offering, alongside a wider partnership on Meta’s Muse family of models. Meta intends to expand its deployment of NVIDIA infrastructure, while the agent-security pieces of the deal are listed only as options.

What was announced

According to NVIDIA’s announcement, Meta intends to expand its use of Vera Rubin and Vera CPUs and will deploy Spectrum-X Ethernet networking and BlueField-4. The two companies plan joint optimization work on inference and training performance, with total cost of ownership named as a goal. The release gives no deployment size, no price for the equity and no closing date, and its forward-looking statements describe the investment as proposed and subject to closing conditions.

Why it matters

The hardware and the money are stated as plans. The agent controls are hedged further. The release says Meta and NVIDIA “will also consider additional collaboration opportunities” on NVIDIA’s OpenShell reference architecture for safe agent deployment and on NVIDIA Sentry for monitoring agent behavior. Neither product is described as part of the committed deployment.

Meta’s October 12 post introducing Muse Wonder says the models are built to ask for clarification and to check in before “following through on anything important”, and that a full report with additional safety evaluations “will be available soon.” Muse Wonder is available now through the Meta Model API and Muse Code.

What it means for the buying committee

A team pricing Muse Wonder through the Meta Model API can verify the model behavior Meta describes and the hardware NVIDIA lists. The monitoring layer around agents has no committed owner in either company’s text. Ask Meta for the safety report before it is final, and ask which agent-monitoring controls ship with the API. Other buyers of the same stack should read the announcement against NVIDIA’s investment in Higgsfield’s Series B extension and the custom-silicon route detailed in Microsoft’s Maia 300 inference chip when they compare supply risk.

Source: NVIDIA Newsroom