Nvidia and AMD Invest $4 Billion in Meta as It Raises $95 Billion for AI

ByAIERA Blog Editorial Team

Key Takeaways

  • Nvidia will invest $3 billion and AMD will invest $1 billion in Meta, a combined $4 billion commitment announced on 27 October 2026.
  • Meta announced a day earlier, on 26 October, that it planned to raise $95 billion through share sales to fund its artificial intelligence infrastructure.
  • The raise is split into $35 billion of underwritten public equity offerings and a further $60 billion through an at-the-market programme that sells shares gradually over time.
  • Demand was so strong that Meta upsized the offering, taking the total financing package to $97.5 billion.
  • The deals turn two of Meta's biggest chip suppliers into shareholders, deepening partnerships built around Meta's Muse AI agent and its data centre buildout.
  • Broadcom was also named in Meta's filing as interested in buying shares, but it had not announced a commitment when Nvidia and AMD confirmed theirs.

What happened

On 26 October 2026, Meta announced plans to raise $95 billion to pay for its artificial intelligence buildout. The structure had two parts. The first was $35 billion through underwritten public equity offerings, where banks buy the shares and resell them to investors. The second was a further $60 billion through an at-the-market programme, a mechanism that lets a company sell shares gradually at prevailing market prices rather than all at once.

A day later, on 27 October, Nvidia and AMD confirmed they would participate. Nvidia said it would invest $3 billion in Meta's equity offering. AMD said it would invest $1 billion. Together, the two commitments total $4 billion. Meta's original filing had named Nvidia, AMD and Broadcom as parties interested in purchasing up to $5 billion of the shares between them. Broadcom had not announced a stake of its own when the other two confirmed theirs.

The market response to the offering itself was emphatic. Investment banks reported that Meta's Class A offering was oversubscribed fivefold, and Meta upsized it. The common stock portion rose from $22.5 billion to $25 billion, and the total public offering reached $37.5 billion once $12.5 billion of mandatory convertible preferred shares were included. Combined with the $60 billion at-the-market programme, the total financing package reached $97.5 billion. The shares were priced at $625 each, and the sales were due to close on 29 October.

The scale of the raise reflects the scale of the spending behind it. Meta projected capital expenditure of between $135 billion and $145 billion in 2026, compared with $72.2 billion in 2025. The company generated $97 billion of operating cash in the first nine months of 2026 and still needed outside money to keep building. Third-quarter revenue reached $64.6 billion, up 26 percent year on year, yet investors remained focused on how quickly the company can convert its spending into returns.

What Nvidia is getting, and giving

Nvidia's $3 billion investment is paired with a deeper technical partnership. Meta will expand its use of Nvidia's Vera Rubin systems and Vera CPUs, along with Spectrum-X networking and BlueField-4 data processing units. The two companies will also work on tuning Meta's Muse models and agents for Nvidia hardware, and they are looking at Nvidia's OpenShell and Sentry tools for running and monitoring AI agents safely.

Muse is central to the story. Meta launched Muse, its AI agent, in September 2026, and both chipmakers describe their investments as support for it. Our earlier coverage of Meta Muse Wonder pricing and benchmarks gives background on the model family behind the agent. Nvidia chief executive Jensen Huang said AI agents represent a new computing paradigm, describing software that understands goals, reasons, uses tools and gets work done on behalf of users. Meta chief executive Mark Zuckerberg said the company looks forward to delivering the best personal agents and frontier models for billions of people.

For Nvidia, the investment also fits a wider pattern. The company has participated in 121 investment rounds in 2026, compared with 13 in 2022, as silicon vendors increasingly take equity stakes in their biggest AI customers. The Meta stake is therefore not an isolated bet. It is part of a strategy in which the supplier of the most important hardware in AI also becomes a shareholder in the companies building the largest AI systems.

What AMD is getting, and giving

AMD's $1 billion commitment comes with its own hardware story. The company's first Helios rack-scale systems are installed and being tested at Meta, and its EPYC processors power the computers behind Muse. Meta will also deploy a custom AMD Instinct GPU built for its workloads. AMD chief executive Lisa Su said the companies are building infrastructure at gigawatt scale for Meta's superintelligence ambitions.

The AMD relationship already had a large commercial base before the equity investment. In February 2026, Meta signed an agreement worth about $60 billion with AMD for six gigawatts of Instinct GPUs, with warrants convertible into roughly 10 percent of AMD's equity. A comparable February agreement with Nvidia was worth about $50 billion. Against purchase commitments of that size, the new equity investments are relatively small, which is one reason analysts have treated them as partnership signals rather than as financing that changes Meta's funding picture on its own.

For AMD, the strategic value is placement. Being a named supplier inside Meta's data centres, alongside Nvidia rather than instead of it, strengthens AMD's position in the AI accelerator market. The Helios racks and the custom Instinct GPU give AMD a visible role in the infrastructure behind one of the world's most widely used AI agents.

Why suppliers are becoming shareholders

The most interesting question in this deal is not the $4 billion figure. It is the direction of the money. In a normal supplier relationship, the customer pays the supplier for hardware. Here, the suppliers are also paying the customer, by buying its shares. The capital then comes back to them as orders for chips, networking equipment and processors.

This pattern has a name in market commentary: circular financing. The concern is straightforward. If a chipmaker's revenue depends on a customer whose spending is partly funded by the chipmaker's own investment, then growth can look stronger than the underlying demand. Investors have raised this question about several AI infrastructure deals in 2026, and the Meta raise is large enough that it will attract the same scrutiny.

There are reasonable answers to that concern, and they matter for understanding the deal. First, the equity stakes are small next to the purchase commitments and next to Meta's own cash generation. Meta made $97 billion of operating cash in nine months, and its February hardware agreements alone were worth about $110 billion combined. A $4 billion equity cheque does not fund the buildout by itself. Second, the partnerships have genuine technical content. Tuning Muse for Nvidia hardware, deploying Helios racks, and building a custom Instinct GPU are real engineering work, not just financial engineering. Third, Meta's offering was oversubscribed fivefold, which means plenty of investors with no hardware to sell were also willing to buy. The suppliers were not the only buyers, or even the largest ones.

None of that removes the risk entirely. If AI spending slows, or if Muse and Meta's other AI products do not generate returns fast enough, then both the equity stakes and the hardware orders are exposed to the same downturn. Suppliers holding customer shares and selling that customer hardware are, to some degree, doubling down on the same bet. That concentration is the honest reason to watch these deals closely, even when the individual amounts look modest.

What the $95 billion is actually for

Meta's spending is dominated by data centres, and data centres for AI are a different kind of building from the server farms of a decade ago. Training frontier models and running agents for billions of users requires enormous power, specialised cooling, and networking that can move data between thousands of accelerators without bottlenecks. The $135 billion to $145 billion of projected 2026 capital expenditure covers that construction, the hardware inside it, and the power to run it. The pressure is visible across the whole cloud industry: Amazon's cloud capacity is largely sold out through 2027, a sign of how tight AI infrastructure has become.

The at-the-market programme is worth understanding because it shapes how the money arrives. A traditional offering sells a fixed block of shares on one day. An at-the-market programme lets the company sell shares gradually, when it judges the price and market conditions to be favourable. For a programme of $60 billion, that flexibility matters. It means Meta is not forced to sell everything at once, but it also means existing shareholders face a long period of potential dilution rather than a single event.

The financing mix also matters. Meta is combining cash earned from its business, corporate bond issuance, strategic alliances and share sales. That breadth is a sign of how large AI infrastructure spending has become. When one of the richest companies in technology turns to shareholders as well as bond markets, it shows that the AI buildout has outgrown even Big Tech's cash flow. That is a useful benchmark for the whole industry, not just for Meta.

What to watch next

Three things will show whether this deal matters beyond the headlines. The first is Broadcom. The filing named it as interested, and if it confirms a stake, the pattern of supplier-shareholders becomes a trio rather than a pair. If it does not, that absence is also informative, suggesting that not every supplier sees equity as the right way to deepen a customer relationship.

The second is deployment. The announcements promise expanded Vera Rubin and Vera CPU deployments, Helios racks in testing, and a custom Instinct GPU. The pace at which those systems move from testing into production at Meta, and the performance they deliver for Muse, will matter more than the share purchases themselves. Hardware announcements are easy. Production deployments at gigawatt scale are hard.

The third is Meta's spending discipline. The company raised the money, and the market funded it enthusiastically. The harder test is whether the data centres and models built with it generate revenue and profit on a timetable that satisfies investors. Third-quarter revenue growth of 26 percent shows the core business is strong. The question is whether AI spending grows into that strength or outruns it.

For readers following AI infrastructure, the Meta raise is also a useful reminder of how the industry's money now flows. Chipmakers are no longer just vendors. They are investors, partners and, in some cases, co-designers of the systems their customers run. That blurring of roles is one of the defining features of the AI buildout, and the Nvidia and AMD investments in Meta are among its clearest examples.

Sources

Similar Posts