Meta set the table on Monday with plans to raise as much as $95 billion for AI infrastructure: an initial $35 billion underwritten public equity offering followed by a $60 billion at-the-market program. The company said Nvidia, AMD and Broadcom had expressed interest in buying up to $5 billion of shares.¹

By Tuesday, two of those prospective buyers had committed. Nvidia said it would purchase $3 billion of Meta stock, while AMD said it would invest $1 billion, turning Meta’s capital raise into a $4 billion endorsement from two of the companies supplying the AI boom’s computing muscle.²

For Meta, the immediate prize is financing for an infrastructure push built around its Muse agent platform and increasingly vast model-training requirements. Nvidia said the companies would expand work on Muse and its family of models to improve training and inference performance. Mark Zuckerberg cast the partnership as a route to “the best personal agents and frontier models for billions of people.”¹

Nvidia’s argument is broader: agents are not merely another application, but a shift in how people use computers. “AI agents represent a new computing paradigm — software that understands our goals, reasons, uses tools and gets work done on our behalf,” CEO Jensen Huang said.²

AMD, meanwhile, emphasized the industrial scale of the arrangement. Its Helios racks are already running at Meta and its EPYC processors power Muse, according to the company. CEO Lisa Su said the partners were building infrastructure “at gigawatt scale” while shaping future silicon, software and systems for Meta’s superintelligence ambitions.²

The investments therefore do double duty: Meta gains cash and supply-chain alignment for a costly AI expansion; Nvidia and AMD gain a larger stake in a customer whose appetite for chips is poised to grow with every new agent and model.