NVIDIA and Meta said they will expand their work on Meta's Muse and the Muse family of models, and NVIDIA will put $3 billion into Meta's stock sale. The joint announcement on NVIDIA's newsroom also says Meta intends to expand its deployment of NVIDIA Vera Rubin and Vera CPUs. On October 26, Meta also priced its stock offerings and made the common stock part larger.
What the announcement says
The NVIDIA release lists four pieces:
- Model work. The two companies will "partner to expand their work on Muse and the Muse family of models to improve inference and training performance."
- Hardware. "Meta intends to expand the deployment of NVIDIA infrastructure, including Vera Rubin and Vera CPUs for next-generation AI agents." Meta will also deploy NVIDIA Spectrum-X Ethernet networking and BlueField-4.
- Agent security tools, maybe. Meta and NVIDIA "will also consider" more work on NVIDIA's OpenShell Reference Architecture, including OpenShell for safe deployment and NVIDIA Sentry for monitoring agent behavior.
- Money. "In addition, NVIDIA will invest $3 billion in Meta's equity offering."
The release gives no unit counts, no gigawatt figures, no dates and no contract value for the hardware. "Intends to expand" is the release's wording. The OpenShell and Sentry items are things the companies will "consider," not signed work.
Where the $3 billion goes
On October 26, Meta launched a financing plan to pay for its AI build-out (our explainer). In that launch release, Meta named NVIDIA, AMD and Broadcom as "Cornerstone Investors" that had, "severally and not jointly, indicated an interest in purchasing up to an aggregate of $5.0 billion" of Class A common stock. Meta said those indications were "not binding agreements or commitments to purchase."
NVIDIA's new release gives a figure for NVIDIA alone: "NVIDIA will invest $3 billion in Meta's equity offering." The release does not say which of Meta's offerings the money goes into, and it does not say how the $3 billion relates to the up to $5.0 billion indication. NVIDIA's forward-looking statements refer to "NVIDIA's proposed investment in Meta, including its anticipated completion and any applicable closing conditions," so the investment is not complete yet.
The pricing
Meta's pricing release sets the terms:
- Common stock: $25.0 billion of Class A common stock at $625.00 per share, "upsized to $25.0 billion from the previously announced offering size of $22.5 billion." The shares are sold through forward sale agreements with Morgan Stanley and Goldman Sachs, for 40,000,000 shares.
- Preferred: 250,000,000 depositary shares at $50 each, each a 1/20th interest in a share of 6.50% Series A Mandatory Convertible Preferred Stock. Meta expects net proceeds of about $12.5 billion from this part.
- Timing: both offerings are expected to close on October 29, 2026. Meta expects about $24.9 billion of net proceeds, before expenses, from the common stock part when the forward agreements settle, which it expects "on or prior to November 19, 2027."
Why it matters
A supplier is funding its customer. Per the NVIDIA release, Meta's Muse models and agents "leverage NVIDIA's infrastructure." Now NVIDIA plans to put $3 billion into a stock sale that Meta says "supports Meta's continued investment in AI infrastructure." In the launch release, Meta said it expects "$135 to $145 billion of capital expenditures in 2026" and plans to fund it through "a balanced mix of operating cash flow, investment-grade debt, strategic partnerships, and equity capital." It joins other recent deals that pair chips with new financing, such as the $63B Broadcom SPV deal for Anthropic chips.
Vera CPUs are part of the pitch. The release names Vera CPUs next to Vera Rubin, and ties both to "next-generation AI agents." Modal, for one, already plans to use Vera CPUs for agent sandboxes.
Muse demand is the reason both companies give. The NVIDIA release says Muse models and agents "have delivered frontier-level benchmark performance and surging consumer downloads since launch." Meta's launch release says it "onboarded more than 10 million Muse users in the U.S. and Canada within the early weeks of launch," and that shopping is the leading use case. Meta also said it expects "to improve cost efficiency through better utilization of off-peak capacity, more efficient models and inference optimization." The joint work on inference performance, which the NVIDIA release says aims to "improve total cost of ownership," points at the same goal. For background on Meta's current compute, see our explainer on Meta's Prometheus supercluster in Ohio.
What to watch
- Whether NVIDIA's $3 billion closes with the offerings on October 29, and whether AMD and Broadcom say what they bought.
- Any numbers for the Vera Rubin and Vera CPU expansion: units, sites, power or timing. The release has none.
- Whether the OpenShell and Sentry items move from "consider" to a signed plan.
Sources
- NVIDIA Newsroom: NVIDIA and Meta to Expand Collaboration on Muse and Muse Family of Models Amid Robust Growth
- Meta (PR Newswire): Meta Announces Pricing of Offering of Depositary Shares and Offering of Class A Common Stock
- Meta Investor Relations: Meta Announces Launch of Offering of Depositary Shares and Offering of Class A Common Stock
