Meta’s two largest chip suppliers are becoming shareholders as well as vendors. Nvidia said it will invest $3 billion in Meta’s equity offering, and AMD said it will invest $1 billion, as Meta raises up to $97.5 billion to pay for the data centres and computing power behind its Muse AI assistant and its wider superintelligence plans. The priced part of the offering was expected to close on 29 October, subject to conditions.
What Meta is raising
Meta announced its fundraising plan on 26 October, initially targeting about $95 billion, before increasing the total to $97.5 billion a day later. The plan has two parts: about $37.5 billion in priced offerings, and a further $60 billion through an at-the-market share sale programme that has not yet started and may not happen in full.
The priced portion consists of $25 billion of Class A common stock, priced at $625 per share, and $12.5 billion of mandatory convertible preferred stock, priced at $50 per depositary share. Reports said orders for the common stock reached about five times the shares on offer, with more than half allocated to long-term and existing investors.
There is an important timing detail. The common stock was sold through forward sale agreements, so Meta said it does not expect proceeds from that part immediately, with settlement possible up to November 2027. The preferred stock carries a 6.50% annual dividend, costing about $812.5 million a year on the base issue if paid in cash.
What Nvidia is putting in
Nvidia’s $3 billion commitment came alongside a deeper partnership on Muse, Meta’s family of AI models and agents. The companies said they will work together to improve training and inference performance, and to lower the total cost of running agents at scale. Meta said its Muse models already run on Nvidia infrastructure.
As part of that work, Meta plans to expand its deployment of Nvidia’s Vera Rubin systems and Vera central processing units, plus Spectrum-X Ethernet networking and BlueField-4 technology. The companies are also considering work on Nvidia’s OpenShell framework for safe agent deployment and its Sentry tool for monitoring agent behaviour, while Muse is already expanding into new markets.
What AMD is putting in
AMD’s $1 billion commitment came with its own expanded partnership, spanning processors, graphics chips, networking, systems and software. AMD said its initial Helios rack-scale systems are installed at Meta and are being tested, ahead of broader deployment. It said Meta has already deployed millions of its EPYC processors across several generations.
AMD also said its Instinct graphics processors are used in Meta production workloads for training, ranking and recommendation, and that its EPYC processors support the computing environments behind Muse. The companies are working on a custom Instinct graphics processor for Meta’s workloads, based on the MI450 architecture, as well as future Verano processors and MI500-series graphics chips.
Why suppliers are buying in
Meta had earlier said Nvidia, AMD and Broadcom had shown interest in buying up to $5 billion of the offering. The confirmed $4 billion from Nvidia and AMD accounts for most of that indicated interest, but it remains a small share of the $37.5 billion priced offering and a smaller share of the full $97.5 billion plan.
For both chipmakers, the logic is commercial as much as financial. Meta is one of the largest buyers of AI hardware in the world, and its spending plans for 2026 run to between $135 billion and $145 billion in capital expenditure. A stake in the customer, alongside a hardware roadmap, makes it harder for a rival supplier to displace them.
Why It Matters
The deal shows how the AI infrastructure boom is now being financed, not just built. Suppliers are no longer only selling chips to customers; they are also putting their own capital into those customers’ fundraising, which blurs the line between a sale and an investment.
Investors have begun asking whether that circular flow makes headline spending a weaker signal of genuine end-user demand, a question that also hangs over other large chip financing deals in the sector, and whether suppliers can keep funding customers indefinitely without returns of their own.
For Meta, the immediate benefit is clear: a heavily oversubscribed offering and two suppliers backing its plans. The longer test is whether the capacity it buys turns into revenue from Muse and its other AI products, rather than into larger data centres. If not, the preferred dividend cost and dilution from new shares will weigh on shareholders for years.
Background
Meta launched Muse in September 2026 as its personal AI assistant, and has since positioned it at the centre of its plan to deliver personal superintelligence to billions of users, a strategy that depends on far more computing power than its social networks ever required.
Source: Nvidia announcement and AMD announcement
0 Comments