Nvidia bets on personal agents in new Meta partnership

Nvidia’s decision to invest $3 billion in Meta's Muse agent and Muse family of models to improve inference and training performance is a strategic bet on the economics of AI agents. However, challenges exist that could collapse its wager.Nvidia and Meta revealed on Monday that the Muse models – Wonder and Wonder Full - and agents will run on Nvidia infrastructure. Meta plans to expand its deployment of Nvidia hardware, including using Vera Rubin and Vera CPUs for its next-generation of AI agents. The vendors are also looking at other collaborations, including using Nvidia’s OpenShell Reference Architecture for agentic security.The partnership comes after Meta's consumer Muse agent has seen massive popularity among consumers, with 10.9 million downloads since its introduction on Sept. 8 and 2.5 million daily active users. The partnership enables Nvidia to bet on the success of personal AI agents while continuing to be a major financial supporter of and GPU provider for a host of vendors in the AI market. In addition to Meta, Nvidia has invested in other AI vendors, including OpenAI, Anthropic and SpaceXSI.Related:Anthropic rolls out Claude Duet, but enterprises aren’t readyThe inference storyNvidia’s support for Muse will significantly boost inference for the Muse agent, said Sid Nag, analyst and founder at Tekonyx.“The real story here is the foundation for everyday agentic activity,” Nag said. “Agents are going to generate multiple inference requests. If Muse becomes the window to consumers … that could create a significant need for massive infrastructure. As a consequence of that, Nvidia gains the opportunity to optimize these workloads."However, the tie-up goes beyond hardware. If Meta and Nvidia deepen their software partnership - agreeing to use OpenShell to help keep Muse safely in the right environment and limit its access to real credentials - it could lead to greater integration, Nag added.“The demand for this could be significant, and Nvidia is positioning itself to be the supplier of that infrastructure, but also creating the software integration with Meta, alongside OpenShell and a host of Nvidia software,” he said. Other Nvidia software includes Nvidia Sentry, which monitors agent behavior.While Muse is now largely for consumers and small businesses, its continued popularity could extend into the enterprise, since there is a clear pattern that many employees “carry their consumer behavior into the workplace,” Nag continued. If that happens, Nvidia stands to gain.“They have a significant amount of engineering resources to help drive the future compute demand as a consequence of Muse’s adoption in the consumer space, and then it bleeds into the enterprise space,” Nag said.Related:Prompt: Enterprise controls can't keep up with AI agentsBut it is still uncertain, especially amid mounting security concerns about agentic AI systems, whether personal AI agents such as Muse can make money. Despite its wide popularity, Meta has yet to make the agent a profit engine.It’s also unclear what Meta’s obligations are regarding this investment.“Is Meta required to demonstrate a certain amount of growth as a condition?” Nag said. “Compute alone cannot resolve bottlenecks.”About the AuthorNews Writer, AI BusinessEsther Shittu has covered AI technologies and industry trends since 2021. As co-host of the Targeting AI podcast, she talks with experts, thought leaders and practitioners exploring critical AI developments. Before AI Business, she wrote for SearchEnterpriseAI, the New York Daily News, Bklyner and the Brooklyn Daily Eagle. When she's not diving deep into the world of AI, she spends her time on passion projects and raising her three daughters.