Nvidia puts $3 billion into Meta's raise as the two expand work on Muse

A $3 billion anchor for a record raise

Nvidia and Meta said that they would deepen their partnership on the Muse AI assistant and related AI models, with Nvidia investing $3 billion in the Facebook parent's equity offering, Reuters reported. Earlier the same day, Meta raised the target for its latest equity offering to $97.5 billion, the largest fundraising campaign in its history.

The structure matters more than the headline number. Nvidia is not buying capacity or paying a prepayment; it is taking equity in the customer whose infrastructure it supplies. The investment anchors a raise that Meta says will fund its AI infrastructure build-out while it spends on AI infrastructure.

What Muse is, and what it can reach

Muse, known internally as Hatch, is central to CEO Mark Zuckerberg's stated plan to deliver what he calls personal superintelligence to the billions of people who use Meta's services daily. It is modelled on the open-source AI agent OpenClaw.

Its access scope is the part worth reading closely. Muse can reach a user's email, calendar, payments, health, shopping and smart-home applications, according to the Reuters account, and users are able to revoke access at any time. That list turns an assistant from something that answers questions into something that acts across accounts, which is why the permission layer rather than the model itself decides how much damage a mistake can do.

Performance work on Nvidia hardware

The collaboration is aimed at making Meta's Muse models and agents, which run on Nvidia infrastructure, work more closely with the chipmaker to improve performance and reduce costs. Those are the two variables that decide whether an assistant can be offered free to billions of people: how much compute a request consumes, and what that compute costs per unit.

The companies may also explore Nvidia's OpenShell framework, including using it to roll out software safely and using a tool called Sentry to monitor how AI agents behave, Reuters reported. That is a monitoring arrangement rather than a model feature, and it addresses the question an operator faces once agents run unattended.

Why a chip supplier buys the customer's shares

Nvidia is paid by the companies that build AI capacity, and it has also put money into some of them. Taking stock in Meta ties the supplier's return to the software layer that will consume its hardware, and it puts a large, public company inside the group of customers whose spending plans Nvidia can partly influence.

For Meta, the appeal is narrower. A $3 billion equity commitment from a supplier is a signal to other investors in a very large offering, and it comes with a joint engineering relationship attached. Neither side has described the arrangement as exclusive, and the Reuters account does not say that Nvidia's hardware becomes the only option.

What the deal does not change

Nothing in the reporting changes how Muse handles permissions for an ordinary user. The ability to revoke access remains the control the company describes, and the investment does not add a new consent step or a new audit right for users.

The raise is also larger than the investment by a wide margin, which means the $3 billion is an anchor rather than the substance of the financing. The remaining $94.5 billion is arithmetic, the target minus Nvidia's commitment, and the willingness of other buyers to pay is the part that will show whether the market accepts Meta's spending plan.

What to watch after the announcement

Three things would make this concrete: the final size of the offering when it closes, the first published benchmarks for Muse agents on Nvidia hardware, and whether OpenShell or Sentry appears in a shipping product rather than a discussion of what the companies may explore.

The broader context is who pays for the AI build-out. Related coverage on this site includes Meta's $97.5 billion AI equity raise, OpenAI's plan to buy Google AI chips and the DOJ's questions about Nvidia's Hugging Face deal. The first two describe compute being bought with cash and shares; the third shows that a supplier's investments draw scrutiny of their own.

What the raise is actually paying for

Meta says the capital is for AI infrastructure, as it competes with rival AI developers. That is a capital-intensive plan with a long lag between spending and revenue, which is why the size of the offering matters more than the size of any single investor's commitment.

It also explains why a supplier would want equity rather than a supply contract. A contract is paid per unit shipped; equity participates in the outcome of the build-out. Nvidia supplies the infrastructure these models run on, so the investment adds a second way to benefit from the same demand while putting money into the customer's construction budget.

There is a second reading of the $3 billion that the reporting supports. Meta's offering is large enough that a single investor's participation is mostly a signal about confidence in the spending plan rather than a solution to financing it. The remaining buyers are being asked to fund AI infrastructure on the strength of an assistant that is still being built, and the supplier's cheque is the part of that story an investor can see today.