Nvidia Invests $3 Billion in the Meta Muse Partnership

1. What did Nvidia and Meta actually announce?

On Tuesday, October 27, 2026, Nvidia and Meta announced they will deepen their partnership on Muse and the Muse family of models, and Nvidia will invest $3 billion in Meta's equity offering. The announcement came in a joint release carried by GlobeNewswire, and Reuters reported the same day that Meta had raised its equity offering to $97.5 billion, the largest fundraising in its history.

Muse models and agents already run on Nvidia infrastructure, and the two companies said the expanded work will improve inference and training performance, the two jobs that decide how good an AI agent feels to use. Inference is the work of answering your request. Training is the work of teaching the model in the first place. The release says the goal is to maximize performance and improve total cost of ownership, which is the full cost of running these systems, not just the price of the chips.

The deal builds on a multiyear partnership the two companies announced in February 2026, covering AI graphics processors, central processing units, and networking technology. This October announcement extends that work specifically to Muse, which Meta describes as its personal AI agent, and to the Muse family of models behind it.

Jensen Huang, founder and chief executive of Nvidia, framed the deal as a shift in how people use computers, from operating applications to working with intelligent agents. He said the companies are expanding the partnership to optimize the computing infrastructure end to end, so Muse can serve billions of people. Meta's chief executive Mark Zuckerberg has made Muse central to his plan for personal superintelligence, the idea that everyone gets an AI helper that knows them and acts for them. Muse is known inside Meta by the name Hatch, according to Reuters.

2. What does $97.5 billion buy, and where does Nvidia's $3 billion sit?

The $97.5 billion is Meta selling new shares to raise cash, and Nvidia is buying $3 billion of them. Put simply, Nvidia is becoming a Meta shareholder while remaining one of Meta's biggest suppliers. Three billion is about 3 percent of the total offering, so Nvidia is a meaningful buyer but far from the whole raise.

Meta announced the offering on Monday, October 26, 2026, at up to $95 billion, then raised the total to $97.5 billion the next day. The sale has three parts: $25 billion of common stock priced at $625 a share, $12.5 billion of preferred shares priced at $50 apiece, and about $60 billion in an open market sale program that runs after Meta's new Class C shares begin trading in early 2027. Meta said the money will fund its investment in AI infrastructure and its consumer AI tools.

Nvidia is not the only chipmaker buying in. Meta said Nvidia, AMD, and Broadcom together agreed to buy as much as $5 billion of the common shares. That detail matters, because it shows Meta is spreading its supplier relationships rather than betting everything on one vendor, even while it deepens the work with Nvidia.

For scale, $97.5 billion is the largest fundraising in Meta's history, according to Reuters. It is also a measure of how expensive the AI buildout has become. Training and running models for billions of people takes data centers, power, and chips at a scale no company has attempted before, and Meta is raising the cash to pay for it in public, in one of the biggest share sales ever attempted.

3. What are Vera Rubin, Vera, Spectrum X, and BlueField 4, in plain language?

Four hardware names appear in the announcement, and each one does a different job. Here they are in plain language.

• Vera Rubin is Nvidia's next generation AI platform, the chips that train models and run them. When you ask Muse a question, the answer is computed on chips like these, in a Meta data center. Vera is the central processing unit that works beside those chips, handling the general computing around the AI work so the AI chips can stay focused on the model.
• Spectrum X is the networking that connects thousands of chips so they behave like one machine. Training a large model is not one chip doing a huge job, it is thousands of chips doing pieces of the job at the same time, and the network decides how fast they can share results. Faster networking means less time wasted waiting on other chips.
• BlueField 4 is a data processing unit, a smaller chip that takes over networking, storage, and security work. By moving that work off the main chips, more of their capacity goes to the model itself. Meta said it will deploy Spectrum X networking and BlueField 4 to support network efficiency and throughput, which is a plain way of saying data moves faster and less of it gets stuck.
• OpenShell and Sentry are frameworks, not chips. OpenShell is a reference design for deploying software safely, built for a world where agents act on your behalf. Sentry is a tool for monitoring how AI agents behave once they are running. Both are described as areas to explore, not as products shipping inside Muse today, and that distinction matters. Nothing in the announcement says either one is already part of the Muse app.

4. What does this mean for you if you use Muse?

Nothing in the Muse app changes today. There is no new button, no new setting, and no new signup step from this deal. The change is underneath, in the machines that answer you, and it arrives over time rather than all at once.

The first thing you should feel, eventually, is speed. Inference performance is the time between asking Muse something and getting a useful answer back. If the joint work delivers what the release promises, answers should come back faster and stay fast when millions of people are using Muse at the same time. The release also talks about total cost of ownership, which matters to you in a quieter way. If each request costs Meta less to answer, the free tier is easier to sustain and the paid tiers are easier to keep priced where they are. That is the logic of the deal, not a promise Meta made, and no price change was announced.

The second thing is reliability at scale. More than 10 million people had already set up Muse in the United States and Canada, and the app keeps adding countries. Singapore launched on October 28, 2026, as the fourth country, after the United States on September 8, Canada on September 18, and Japan on October 23. Every new country adds millions of potential users, and the infrastructure has to grow ahead of them, not behind them. This deal is part of that growth.

The third thing is trust in what an agent is allowed to do. Muse can reach into your email, calendar, payments, health, shopping, and smart home apps, with your permission, and you can take that permission away at any time. An agent with that much reach needs guardrails, and the OpenShell and Sentry work is aimed at exactly that: safer deployment and closer monitoring of what agents do. If those frameworks mature into real products, they are the kind of change you notice as fewer mistakes and clearer limits, not as a headline feature.

5. What does this deal not change?

This is the part most coverage skips, so here it is plainly. The announcement changes nothing about what Muse does, what it costs, or where it is available.

• No new Muse features were announced. The release talks about models, infrastructure, and frameworks, not about a new capability you will see in the app. If you were hoping for a new trick, this is not that announcement.
• No price change was announced. Nothing in the release touches what Muse costs, and no change to the free tier or the paid plans was mentioned. The money in this deal is for building infrastructure, not for repricing the app.
• No new country was launched by this deal. Muse is live in the United States, Canada, Japan, and Singapore, and new launches are announced separately. The Singapore launch on October 28, 2026, happened the day after this announcement and had nothing to do with it.
• Nvidia did not buy Muse. It bought $3 billion of Meta shares. Muse stays a Meta product, built by Meta, running on hardware Meta buys from Nvidia and others. The distinction matters because an equity investment is not a takeover, and nothing about who controls Muse has changed.
• No timeline was given for faster responses. The performance gains are stated aims, and the release gives no date for when you should expect to feel them. If a post tells you Muse will be twice as fast by a certain date, that number came from somewhere else, not from this announcement.

6. Why is Nvidia buying Meta stock at all?

The obvious question is why a chipmaker would buy stock in its own customer. The answer is alignment. Nvidia already sells Meta the chips that run Muse, so Nvidia's business grows when Muse grows. Owning Meta shares adds a second reason to want the same outcome, and it puts Nvidia's money behind a partnership it already depends on.

There is a second reason, and it is about the market for AI chips. Meta is one of the largest buyers of AI hardware in the world, and it is now buying from Nvidia, AMD, and Broadcom at the same time. For Nvidia, a $3 billion investment is a way to stay close to its biggest customer while that customer deliberately keeps other suppliers in the picture. For Meta, having its suppliers as shareholders is a way to keep them committed to its buildout.

It also fits the pattern of this year. Meta and Nvidia announced a multiyear partnership in February 2026, and the October announcement extends it rather than starting something new. The relationship has been moving in this direction for months, and the investment makes the direction public in a way a press release alone could not.

The scale is the story. A $97.5 billion raise, the largest in Meta's history, says the AI buildout is the most expensive project Meta has ever funded. Nvidia putting $3 billion of its own money into that raise says the company that sells the picks and shovels believes the mine is worth digging.

7. What happens next, and what should you watch?

Three things are worth watching from here.

First, the hardware. Meta said it will expand its deployment of Vera Rubin systems and Vera central processing units, and deploy Spectrum X networking and BlueField 4. Watch for Meta talking about these systems going live in its data centers, because that is when the performance work stops being a press release and starts being something you can feel.

Second, the frameworks. OpenShell and Sentry are only described as areas the companies may explore. Watch for either one turning into a named product or a real feature, because safer deployment and better monitoring are the changes that would matter most to anyone who lets an agent act on their behalf.

Third, the countries. Muse is live in four countries as of October 28, 2026, and the waitlist keeps growing everywhere else. Infrastructure deals like this one make expansion easier, but they do not announce it. The next launch will come as its own announcement, and the full list of Muse available countries records every launch as it happens.

One more thing to watch is the money itself. The offering includes an open market sale program of about $60 billion tied to Meta's new Class C shares, which begin trading in early 2027. How much of that program Meta actually uses will say a lot about how fast it intends to keep building, and how much of the AI bet it funds with new shares rather than with cash it already has.

For now, the deal is best read as a foundation, not a finish line. The fastest AI agent in the world still needs machines to run on, and this announcement is about who supplies those machines, who pays for them, and how closely the two companies will work together to make them faster.

Get started

New to Muse? See where it is live today on the full list of Muse available countries.

Frequently asked questions

Did Nvidia buy Muse?

No. Nvidia invested $3 billion in Meta's equity offering, which means it bought Meta shares. Muse stays a Meta product, and the partnership is about the models and the infrastructure underneath them.

Does this change what Muse costs?

No. The announcement says nothing about Muse pricing, and no change to the free tier or the paid plans was announced. The money is for AI infrastructure, not for repricing the app.

Will Muse get faster because of this deal?

That is the stated aim, but no timeline was given. Meta and Nvidia said they will work together to improve inference and training performance, which should mean quicker answers and steadier service over time. Treat it as a direction, not a dated promise.

Does this deal launch Muse in new countries?

No. No new country was announced with this deal. Muse is live in the United States, Canada, Japan, and Singapore as of October 28, 2026, and new launches are announced separately.

What are OpenShell and Sentry?

OpenShell is Nvidia's framework for deploying software safely, and Sentry is a tool for monitoring how AI agents behave. The companies said they may explore working together on both, so neither is a shipped Muse feature yet.

Who else is investing in Meta's offering?

Meta said Nvidia, AMD, and Broadcom together agreed to buy as much as $5 billion of the common shares. The full offering is $97.5 billion, the largest fundraising in Meta's history, and it also includes an open market sale program of about $60 billion.

Built with Muse is an independent community project. Not affiliated with or endorsed by Meta.