Meta deepens partnership with NVIDIA on Muse – Unite.AI AI models and infrastructure

NVIDIA and Meta will expand their collaboration on Meta’s Muse family of AI models and expand planned deployments of NVIDIA compute and networking infrastructure to Meta, the companies said in an NVIDIA Newsroom release dated October 26, 2026. NVIDIA will also invest $3 billion in Meta’s equity offering.

The companies said they will collaborate to expand work on Muse and the Muse family of models to improve inference and training performance, citing what the release describes as popular demand for useful AI agents. According to the release, Meta’s Muse models and agents run on NVIDIA infrastructure and have delivered what the companies called frontier-level benchmark performance and increased consumer downloads since launch. Meta and NVIDIA said they intend to further deepen their collaboration through joint workflow optimization work, with the stated goals of maximizing performance and improving total cost of ownership.

“AI agents represent a new computing paradigm: software that understands our goals, our reasons, uses tools and does the work on our behalf,” Jensen Huang, founder and CEO of NVIDIA, said in the release. He said the shift changes how people use computers, moving from operational applications to working with intelligent agents, and that the companies are expanding the partnership to optimize end-to-end computing infrastructure for capable, efficient and secure AI agents.

OpenShell and Sentry under review

According to the release, the companies will also consider further collaboration on the NVIDIA OpenShell reference architecture. This consideration applies to NVIDIA OpenShell, which the release describes as supporting secure deployment, and NVIDIA Sentry, described as a tool for monitoring agent behavior. The release summary also describes Sentry in relation to the monitoring and processing of structured, unstructured and graphical data.

NVIDIA infrastructure expanded to Meta

Meta intends to expand its implementation of NVIDIA infrastructure, including Vera Rubin and Vera CPUs, for next-generation AI agents, with the stated goal of providing maximum performance per watt. Meta will also deploy NVIDIA Spectrum-X Ethernet and BlueField-4 networking to support network efficiency and throughput, according to the release.

$3 billion equity investment proposal

NVIDIA will invest $3 billion in Meta’s equity offering, the release said. In the forward-looking statements section of the release, NVIDIA describes the transaction as the proposed investment in Meta, citing expectations regarding its expected completion and applicable closing conditions.

“Muse is how we bring personal superintelligence to everyone,” Meta founder and CEO Mark Zuckerberg said in the release. He said Meta is excited to continue working with NVIDIA to provide personal agents and frontier models for billions of people.

It builds on the February Infrastructure Partnership

The release states that the new announcement strengthens the companies’ long-term partnership in support of Meta’s AI infrastructure ambitions. On February 17, 2026, Meta announced a multi-year strategic partnership with NVIDIA to advance its long-term AI infrastructure roadmap, with large-scale deployment of NVIDIA technology to support Meta’s creation of data centers optimized for AI training and inference, as well as its core business. Meta said the advances will deliver substantial improvements in performance per watt, supporting more efficient AI operations at scale.

Meta’s February announcement also stated that the company had adopted NVIDIA Confidential Computing for WhatsApp Private Messaging, enabling AI-powered features across the messaging platform while ensuring the confidentiality and integrity of user data. Meta has also adopted the NVIDIA Spectrum-X Ethernet networking platform across its infrastructure, describing it as capable of delivering AI-scale networking with predictable, low-latency performance, while maximizing utilization and improving operational and energy efficiency. As part of what Meta calls a multi-generational collaboration, engineering teams from both companies would optimize and accelerate advanced AI models in Meta’s core workloads to improve the performance and efficiency of new AI capabilities used by billions of people.

In the February announcement, Huang said the partnership featured deep co-design of CPU, GPU, networking and software, bringing NVIDIA’s complete platform to Meta’s researchers and engineers. He said no other organization deploys AI at the scale of Meta, highlighting Meta’s pairing of frontier research and large-scale infrastructure behind personalization and recommendation systems that serve billions of users. Zuckerberg said in February that Meta was excited to expand its partnership with NVIDIA and build cutting-edge clusters on the Vera Rubin platform, with the stated goal of providing personal superintelligence.