AMD Posts $1 Billion Accounting Charge for Giving OpenAI and Meta Equity Stakes in Exchange for Chips

What Happened

On Tuesday (November 4), AMD reported a $1.07 billion non-cash accounting charge for its third fiscal quarter, which ended on September 26. The amount equals 66 cents per share and is tied to warrants (stock purchase options) granted to OpenAI and Meta Platforms as part of two megadeal chip supply contracts for AI chips. Despite the charge, the company projected faster-than-expected growth for 2027, driven by data center sales, and shares rose 8% in after-hours trading, also pulling up Intel, Nvidia, and Broadcom stock.

How the Warrant Agreement Works

In February, AMD signed a contract to sell up to $60 billion in AI chips to Meta over five years. As part of the deal, Meta received warrants that could give it the right to buy up to 10% of AMD, provided certain purchase targets are met. The structure mirrors an earlier agreement, signed last year with OpenAI, which also received the option to acquire around 10% of the chipmaker. According to the original report, OpenAI's warrants have triggers tied to AMD's stock price; Meta's do not.

The Price of Securing Billion-Dollar Demand

AMD's chief financial officer, Jean Hu, said on the earnings call with analysts that shareholder dilution only kicks in once OpenAI and Meta meet the vesting and exercise conditions, which include product shipment targets and other commercial requirements. The company promised to share more details on that timeline in 2027. In short: if the warrants are fully exercised, OpenAI and Meta could each buy a significant stake in AMD for pennies on the dollar, which would substantially increase the company's share count.

Analysts read the move as a vote of confidence in AMD's AI chips and software, but it is not expected to shake Nvidia's dominance, which continues to sell every unit it can manufacture.

While I do believe incentives played a role in some of their sales, I believe the bigger issue is that there is pent-up demand for a serious alternative to Nvidia and AMD's products are now starting to compete on price/performance metrics.

Bob O'Donnell, chief analyst at TECHnalysis Research

A Pattern That Became the Industry Norm in AI

AMD didn't invent this model, but it's using it at large scale. The company also struck a chip purchase agreement with Anthropic, a competitor of OpenAI, which included a commitment for AMD itself to invest in the startup. Other AI infrastructure companies have made similar deals with major clients:

• Chip startup Cerebras granted OpenAI warrants tied to a $20 billion commitment.
• Marvell issued warrants to Alphabet (Google) giving the company the right to buy a stake worth up to $12.2 billion.
• Qualcomm gave Amazon warrants in a $60 billion contract for AI data center chips.

The result is a web of cross-ownership in which the same handful of companies simultaneously act as each other's customer, supplier, and investor, what the original report calls the circular nature of the AI boom.

What This Means for Infrastructure Decision-Makers in Brazil

For Brazilian founders and technical leaders who depend on GPU capacity or cloud contracts to train and run models, these deals show something concrete: securing supply of cutting-edge chips is no longer just a matter of money, it's becoming a form of equity currency. This favors buyers operating at the scale of tens of billions of dollars, a level no Brazilian company reaches on its own, but which the major cloud providers used here (AWS, Azure, Google Cloud) negotiate directly with these same chipmakers.

In practice, the effect reaches Brazil indirectly. If Meta and OpenAI secure priority supply of AMD chips in exchange for equity, buyers of GPU capacity without this kind of agreement, including smaller providers and cloud resellers in the country, may see more price and delivery-time variation in the coming cycles. Anyone negotiating a GPU-as-a-service contract or cloud credits to train models in Brazil has good reason to ask their supplier whether the quoted price reflects guaranteed capacity or the regular queue.

There's also a business-model lesson here. Equity for purchase volume is a tool that large technology buyers, not just chip buyers, can negotiate when the supplier needs revenue predictability to justify heavy capex, and the same logic applies to those negotiating large infrastructure, data, or model-licensing contracts in Brazil. What remains open is the timeline: AMD itself said it will only detail when the vesting conditions are expected to be met in 2027, which leaves the real size of future dilution uncertain until then.