Nvidia and AMD Are Buying $4 Billion of Their Own Customer's Stock

Meta has upsized its equity raise to $97.5 billion, the largest in its history. Nvidia is putting in $3 billion and AMD $1 billion — money that will largely flow back to them as orders for chips. The AI boom has started financing itself.

The companies selling Meta the chips are also buying the stock Meta is selling to pay for them.

On Tuesday, Nvidia said it will put $3 billion into Meta's equity offering. AMD said it will put in $1 billion. Meta had already told investors that Nvidia, AMD and Broadcom were interested in buying up to $5 billion of Class A shares in total. The raise itself was upsized to $97.5 billion — the largest in Meta's history — from the $95 billion announced a day earlier, according to Reuters.

The structure matters. Meta now plans to raise $25 billion in common stock at $625 a share, up from $22.5 billion disclosed on Monday. It priced 250 million convertible preferred depositary shares at $50 each, another $12.5 billion. And it has a separate at-the-market program to sell up to $60 billion more over time, expected to begin after its planned non-voting Class C shares start trading in the first quarter of 2027. Meta closed Monday at $630.56, down nearly 2 percent, and was down about 1 percent on Tuesday.

Why does the company that owns Facebook, Instagram and WhatsApp need to sell a slice of itself? Because AI infrastructure has become a cash furnace. Meta burned $2.12 billion of free cash flow in the third quarter — its first negative quarter in more than a decade, against positive free cash flow of $10.63 billion a year earlier. It has committed to roughly $825 billion in future spending, mostly data centers, AI infrastructure and cloud capacity, and it raised the low end of this year's capital expenditure forecast to $135 billion to $145 billion. At the end of September it held $82.45 billion in cash and marketable securities against $80.92 billion in long-term debt. Revenue is still growing fast — $64.6 billion in the quarter, up 26 percent — but the balance sheet has stopped being a self-funding machine, as CNBC reported.

So the chipmakers arrived with checkbooks.

For Nvidia, the $3 billion is not charity, and it is not really a bet on Meta's advertising business. It is a bet on Muse, Meta's AI agent, and on the hardware that runs it. The two companies said they will expand their partnership on Muse and the Muse family of models, with Meta deploying more Nvidia infrastructure — Vera Rubin systems and Vera CPUs, Spectrum-X Ethernet networking and BlueField-4 processors — while the pair optimize training and inference for cost and performance. They will also explore Nvidia's OpenShell reference architecture and a monitoring tool called Sentry for keeping agents in line, according to Nvidia.

"AI agents represent a new computing paradigm — software that understands our goals, reasons, uses tools and gets work done on our behalf," Jensen Huang said in the announcement. "This changes how we use computers, from operating applications to working with intelligent agents."

Muse is the demand story that justifies the spending. Launched in September, it is known internally as Hatch and was modeled on the open-source agent OpenClaw, according to Reuters. It can reach into a user's email, calendar, payments, health, shopping and smart-home apps, with access revocable at any time. It began as a consumer product tied to Meta's Charm device and has a small-business version that plugs into Dropbox, Slack and Zoom. Every one of those sessions is inference, and inference is chips.

That is the logic. Here is the circle. Nvidia has spent the past two years becoming the AI industry's largest shareholder as well as its largest supplier. It has invested in OpenAI, Anthropic, Intel and xAI, and participated in 121 investment rounds so far this year, according to Dealroom data cited by CNBC — up from 13 in 2022. AMD, which committed up to $5 billion to Anthropic in July, has been in 25 funding rounds this year, up from two in 2022. Reuters put the obvious concern plainly: the Meta deal "adds to concerns about an increasingly circular flow of capital in the AI industry, as chip suppliers help finance the expansion of one of their biggest customers."

Vendor financing is not new, and it is not illegal. Equipment makers lent to carriers during the telecom build-out of the late 1990s, and when demand stalled the equipment orders and the loans soured together. What makes this version notable is the scale and the concentration: a dollar of Meta's capital spending is a dollar of Nvidia's revenue, and now part of the capital behind that dollar comes from Nvidia itself. If Meta's AI returns disappoint, Nvidia's revenue outlook and the value of its Meta stake would fall at the same time. The $4 billion is only about 4 percent of the $97.5 billion raise, small enough to dismiss as a rounding error, large enough to signal that the buyer needed help filling the book.

AMD's $1 billion is the more interesting tell. Its Helios racks are now running at Meta and its EPYC processors power Muse, the company said. Paying for a seat inside a customer that also named Broadcom as a possible buyer is a way of staying inside the tent while Meta spreads its orders around. Nvidia's $3 billion does something similar, tying the relationship to the Muse roadmap just as Meta's own models become the showcase for Nvidia's next-generation racks.

Meta is not alone in making the shift. Alphabet increased its equity offerings to $84.75 billion in June. SpaceX said Monday that an affiliate of Apollo Global Management will buy AI chips and server hardware and lease them back to the company, in a deal with total principal payments of about $40 billion. Goldman Sachs estimates AI-related capital spending could total about $7.6 trillion from 2026 through 2031. The funding mix is moving from cash flow toward debt, equity and supplier balance sheets — which is what happens when the build-out outruns the earnings it is supposed to produce.

"We expect investors to continue to question the financing and potential for future capital raises," William Blair analysts wrote on Monday, while adding that the raise also "reinforces both Meta's confidence in the long-term returns from AI investment and the increasingly capital-intensive nature of pursuing that opportunity."

Both things can be true. The chips are real. The orders are real. The equity is real. What is not yet settled is whether the demand would exist at this size without the suppliers helping to pay for it. Meta's next cash-flow report, and Nvidia's next disclosure of what its investment portfolio is worth, will say more than either press release.

Sources: Reuters, via The Star, Reuters, via The Star, CNBC, CNBC, NVIDIA Newsroom