Test report DSG-2598 · Rev E · tested October 30, 2026
AI Datacenter InfrastructureDevice under test
Nvidia commits $3bn to Meta's $97.5bn equity offering
Nvidia will buy $3bn of Meta's $97.5bn equity offering and supply Vera Rubin, Spectrum-X and BlueField-4 hardware for the Muse AI agent platform.
- Read
- 3 min
- Words
- 604
- Node
- 28nm
- Operator
- Marcus Bennett
Spec summary
- Nvidia will invest $3bn in Meta's $97.5bn equity offering.
- Meta will deploy Vera Rubin, Vera CPUs, Spectrum-X Ethernet and BlueField-4 DPUs.
- Meta raised the offering size to $97.5bn on October 26: $25bn common stock plus $12.5bn in convertible preferred depositary shares.
- 250 million convertible preferred depositary shares priced at $50 each.
- Meta plans up to $60bn in additional share sales once Class C shares trade in Q1 2027.

Nvidia will invest $3 billion in Meta's $97.5 billion equity offering, under an expanded partnership that ties the chipmaker's capital directly to Meta's buildout of AI agent infrastructure.
The two companies will focus the collaboration on Meta's Muse family of AI agents. Nvidia said the partners will work on "optimizations across workflows" to raise inference and training performance and improve total cost of ownership.
The deal goes beyond capital. Meta has committed to expand its deployment of Nvidia hardware across several product lines:
- Vera Rubin platforms and Vera CPUs
- Spectrum-X Ethernet networking technology
- BlueField-4 DPUs
The companies will also explore "additional collaboration opportunities" around Nvidia's OpenShell and Sentry offerings, aimed at supporting the safe deployment of AI agents.
What does the investment change for Meta's infrastructure plans?
It locks in supply and engineering alignment at the hardware layer. Meta gets a committed silicon roadmap for Muse, plus a co-optimization program covering the full stack from networking to processing. Nvidia, in turn, secures both a major customer and a $3 billion equity position in that customer's growth.
Jensen Huang, Nvidia's founder and CEO, framed the deal around a shift in how software operates.
"AI agents represent a new computing paradigm — software that understands our goals, reasons, uses tools and gets work done on our behalf," Huang said.
He added: "This changes how we use computers, from operating applications to working with intelligent agents. Nvidia and Meta are expanding our partnership to optimize the computing infrastructure end-to-end for capable, efficient and secure AI agents, so Muse can empower billions of people to achieve more and pursue greater ambitions."
How is Meta funding the buildout?
Meta announced on October 26 that it had raised the size of its planned equity offering to $97.5 billion. The structure breaks into three parts.
First, $25 billion comes through common stock. Second, an additional $12.5 billion will be raised through the sale of 250 million convertible preferred depositary shares at $50 each. Third, Meta plans to sell up to $60 billion in shares after its planned non-voting Class C shares begin trading in the first quarter of 2027.
Nvidia's $3 billion commitment sits inside that overall offering.
What is Muse?
Meta introduced the Muse AI agent in September 2026. The company describes it as a personal AI agent "that proactively helps with people's goals and suggests ideas." It runs on a dedicated virtual machine that hosts both the agent and a user's data, keeping the workload isolated.
The product first reached consumers through the Charm device, a Tamagotchi-like gadget. Meta has since launched a version aimed at small and medium enterprises. That edition integrates tools such as Dropbox, Slack and Zoom into Muse's agentic workflow.
Why does the agent design matter for hardware?
Agents that reason, call external tools and hold user data on a dedicated VM generate sustained compute load. That load spans training, inference and networking, which explains the breadth of Nvidia products in the agreement: CPUs for the host layer, Spectrum-X for east-west traffic between nodes, and BlueField-4 DPUs for offloaded data-path processing and security.
For hardware vendors and data center operators, the deal signals where agent platforms are heading: end-to-end co-optimized infrastructure rather than component-by-component procurement. The safety angle — OpenShell and Sentry — points to a growing requirement that agent deployments carry hardened execution environments.
The offering proceeds give Meta the balance sheet for that buildout, and Nvidia now holds a direct financial stake in the outcome.
via wired.com (Original)
More from Marcus Bennett
Show full bio
News editor covering regulation, policy and their impact on hardware layer of artificial intelligence at Die Signal.
516 articles
Same lot · LOT-C204
- DSG-381914nmAmazon Seeks to Offload $8bn of Nvidia Chips to Investors
- DSG-891845nmNvidia to Invest in Chip Rival d-Matrix as Its Challengers Choose Partnership
- DSG-820310nmNvidia reserves AI accelerator capacity through 2028
- DSG-270128nmAmazon Weighs $8 Billion Nvidia Chip Sale to Investors
- DSG-530928nmAmazon Seeks to Offload $8 Billion of Nvidia Chips to Investors