NVIDIA will invest $3 billion in Meta’s equity offering as the two companies expand their partnership around Muse, Meta’s personal AI agent, NVIDIA announced on October 27. The expanded deal covers joint work to improve Muse’s inference and training performance, a wider Meta rollout of NVIDIA’s Vera Rubin systems, and possible collaboration on agent safety and monitoring tools. Meta said Muse models and agents have delivered frontier-level benchmark results and surging consumer downloads since launch, and NVIDIA chief executive Jensen Huang said the partnership will optimise the computing infrastructure behind capable AI agents from end to end.
Under the expanded partnership, Meta and NVIDIA will work on optimisations across Muse workflows to improve performance and total cost of ownership. The companies said they will also consider collaborating on NVIDIA’s OpenShell reference architecture for safe agent deployment and on NVIDIA Sentry, a tool for monitoring agent behaviour across structured, unstructured and graph data. Meta intends to expand its deployment of NVIDIA Vera Rubin and Vera CPUs for next-generation AI agents, alongside Spectrum-X Ethernet networking and BlueField-4 networking hardware to support network efficiency and throughput.
What The Expanded Partnership Covers
According to NVIDIA, the partnership centres on Muse and the Muse family of models, which run on NVIDIA infrastructure. The two companies will partner on optimisations across workflows, with the stated aim of maximising performance and improving total cost of ownership for agent workloads. NVIDIA framed agents as a new computing paradigm, describing software that understands goals, reasons, uses tools and gets work done on a user’s behalf.
The infrastructure side of the announcement is specific. Meta will expand its deployment of Vera Rubin and Vera CPUs, which NVIDIA positions for maximum performance per watt on next-generation AI agents. Spectrum-X Ethernet networking and BlueField-4 data processing units will support network efficiency and throughput for those workloads. The companies also left the door open to further work on OpenShell for safe deployment and Sentry for monitoring agent behaviour, though both are described as collaborations under consideration rather than shipped products.
The $3 Billion Investment
NVIDIA said it will invest $3 billion in Meta’s equity offering, a commitment disclosed in the same announcement as the expanded partnership. NVIDIA did not disclose further terms of the investment, such as the share class or timing, beyond describing it as an investment in Meta’s equity offering. The announcement’s forward-looking statements note that the proposed investment remains subject to applicable closing conditions.
The investment lands as Meta raises large sums to fund its AI build-out. Meta’s $35 billion stock sale sits alongside a planned at-the-market programme of about $60 billion, and chipmakers NVIDIA, AMD and Broadcom had already been named as cornerstone investors indicating interest in up to $5 billion of stock in aggregate. NVIDIA’s $3 billion commitment puts a firm figure on its own share of that cornerstone interest. Meta chief executive Mark Zuckerberg said Muse is how Meta brings personal superintelligence to everyone, and that the company is excited to keep working with NVIDIA to deliver personal agents and frontier models for billions of people.
Why It Matters
Muse sits at the centre of Meta’s AI strategy, and this deal ties its future directly to NVIDIA’s hardware roadmap. Faster, cheaper inference on Vera Rubin systems would let Meta serve more agent workloads without a matching rise in cost, while Spectrum-X and BlueField-4 address the networking bottleneck that grows as agents run longer, multi-step tasks. For NVIDIA, the partnership locks in one of the world’s largest AI deployers as a showcase customer for its Rubin generation, and the $3 billion investment gives it a financial stake in the customer it supplies.
The safety strand of the announcement matters too. OpenShell and Sentry speak to the trust problem facing agents that act on a user’s behalf: companies deploying agents need ways to keep them inside safe bounds and to watch what they do. If the two companies turn those considerations into shared tools, other agent builders may adopt similar guardrails, and enterprise buyers weighing agent deployments will have a clearer reference point for what safe deployment looks like.
Background
Meta and NVIDIA have worked together for years on AI infrastructure, with Meta building giant training clusters on NVIDIA hardware and NVIDIA supplying the chips behind Meta’s model development. Muse launched as Meta’s personal AI agent and has since expanded to new markets, including Muse’s launch in Japan, while Meta has poured spending into data centres and its Prometheus supercluster in Ohio. The expanded partnership extends that relationship from supplier and customer into joint optimisation work and a direct equity investment.
Source: NVIDIA: NVIDIA Muse partnership; Reuters: Reuters Muse report
0 Comments