AMD and Meta Announce Expanded Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs

[hist78]

The "pay your customer who pays you" loop continues..

It’s not exactly a simple “pay your customer who pays you” situation. In the Meta-AMD deal, AMD will generate real revenue from Meta’s real orders and increase the production scale of its products (also real), including EPYC server processors, AMD Instinct MI450 GPU/AI processors, and Helio rack systems.

In exchange, AMD is giving Meta performance‑based warrants to acquire AMD shares, which could dilute existing shareholders’ ownership. But the key is that the warrants are tied to performance. If AMD reaches certain stock price, revenue, and product delivery milestones, shareholders may end up better off than they would have been without the large orders Meta brings. Additionally, Meta must purchase a specified amount of AMD products in order to exercise the warrants.

Let’s face it: in this intense AI infrastructure race, Nvidia is capturing the majority of large customers and large volumes, while many “also‑ran” AI chip companies fight over the remaining market. Meanwhile, Microsoft, Google, Amazon, and Meta are all building their own server and AI processors for internal use. Without large customer orders, it’s very difficult for the #2 or #3 players to compete. The Meta-AMD deal isn’t perfect, but it’s a strong strategic move for AMD to secure a big customer's order.

This deal also highlights the very different mindsets of Intel and AMD. AMD is a fabless company, so it doesn’t need to spend massive capital on manufacturing facilities and equipment. It focuses on designing the best products and winning customer orders.

Intel, on the other hand, is an IDM. It must worry about everything AMD worries about plus one enormous burden: building extremely expensive fabs.

From 2020 to 2025, Intel spent a staggering $127 billion in CapEx to build those fabs.

To support this massive $127 billion CapEx, Intel raised cash from multiple sources in various forms: Brookfield ($11 billion), Apollo ($15 billion), the US government ($8.9 billion), Nvidia ($5 billion), and SoftBank ($2 billion). None of them committed or has the ability to bring orders to Intel.

But while raising capital to fund huge CapEx, Intel still hasn’t solved the most important part of the business: where are the orders, and where is the revenue and profit? Intel’s products still can’t compete effectively. As a result, Intel’s revenue has declined, and the company has incurred multiple years of net losses.

If a deal can’t meaningfully increase orders and can’t lead to profitability, it’s an ineffective deal or a bad deal.

 

[Xebec]

It’s not exactly a simple “pay your customer who pays you” situation. In the Meta-AMD deal, AMD will generate real revenue from Meta’s real orders and increase the production scale of its products (also real), including EPYC server processors, AMD Instinct MI450 GPU/AI processors, and Helio rack systems.

In exchange, AMD is giving Meta performance‑based warrants to acquire AMD shares, which could dilute existing shareholders’ ownership. But the key is that the warrants are tied to performance. If AMD reaches certain stock price, revenue, and product delivery milestones, shareholders may end up better off than they would have been without the large orders Meta brings. Additionally, Meta must purchase a specified amount of AMD products in order to exercise the warrants.

Let’s face it: in this intense AI infrastructure race, Nvidia is capturing the majority of large customers and large volumes, while many “also‑ran” AI chip companies fight over the remaining market. Meanwhile, Microsoft, Google, Amazon, and Meta are all building their own server and AI processors for internal use. Without large customer orders, it’s very difficult for the #2 or #3 players to compete. The Meta-AMD deal isn’t perfect, but it’s a strong strategic move for AMD to secure a big customer's order.

I agree it's a strong strategic move, and it's an unequal repsonse to Nvidia (i.e. innovative). But it's still basically just a discount on future sales -- if Meta buys enough of these GPUs, then they get 10% of a publicly traded company to go along with it. It's no different than a cash investment from Nvidia to OpenAI, or Intel's rebating techniques, except the stakes are higher. (And there's less tiering involved -- this is all or nothing).

It's both a strong move but also a desperate one.. imo

Also, if this deal can't meaningfully increase orders - AMD stock may get punished more now because there are higher shareholder expectations for sales volumes that didn't exist prior to the deal.