The tweet was deleted by the author.
But we saved everything 🙂.
Meta has signed a major deal with Advanced Micro Devices to supply AI chips and servers. According to estimates, the value of the agreement could exceed $100 billion. For AMD, this is a chance to move closer to Nvidia, while for Meta it is a way to reduce dependence on a single supplier and expand its AI infrastructure.
Meta has agreed with AMD on AI chip and server supplies for five years. The deal covers up to 6 gigawatts of capacity for Meta’s data centers. Deliveries are scheduled to begin in the second half of 2026.
The key product in the agreement is AMD Instinct MI450 accelerators. Meta plans to configure part of these chips for inference — this is when a model has already been trained and needs to be quickly deployed in services such as search, recommendations, advertising, and chatbots.
In addition to GPUs, Meta will purchase AMD EPYC 6th-generation server processors (Venice). These processors are needed to run the servers: they manage the system, memory, and networking, while GPUs handle AI computations.
According to estimates, the agreement could exceed $100 billion. Other estimates mention up to $60 billion over five years. It all depends on how much equipment Meta ultimately selects and at what prices.
Another part of the deal involves AMD shares for Meta. Meta will receive the right to purchase up to 160 million shares — about 10% of the company. The price is symbolic — $0.01 per share. However, the shares will not be transferred immediately: the first portion will be granted after the delivery of 1 gigawatt, and the remaining tranches as purchases increase toward 6 gigawatts. The final portion will only be granted if AMD’s stock rises to $600.
The market reacted immediately: AMD shares rose from $196 to $213 in a single day. On the day of the news, the stock closed around $214, and premarket gains reached roughly 14%.
But why does Mark Zuckerberg’s company need this agreement? Meta is pursuing this deal because of the scale of its AI plans. The company has already launched the Meta Compute initiative and openly says it wants to build “tens of gigawatts” of capacity this decade and “hundreds of gigawatts or more” over time. Last year, Meta spent $72 billion on AI data centers, and this year it plans to spend up to $135 billion. At this scale, it is important for Meta not to rely on a single supplier.
Today, Nvidia remains the main supplier in the AI chip market, and Meta continues to buy millions of GPUs from it. At the same time, Meta is securing a second major supply channel through AMD. Two things are key in the AMD deal: guaranteed volumes for years ahead and customization for Meta’s needs. Meta will receive advanced accelerators and will also be able to influence what future versions of these chips look like.
For AMD, this is a way to accelerate the race with Nvidia. Contracts of this scale provide a major anchor customer and predictable demand for several years. Meta is already considered AMD’s second-largest customer and will now become even more important for its growth. At the same time, AMD is promoting its server architecture developed together with Meta within the Open Compute Project — this helps Meta build data centers faster from ready-made “building blocks,” while allowing AMD to sell not only chips but broader infrastructure solutions.
The Meta–AMD deal fits into a broader trend: major technology companies are steadily increasing their AI spending. According to Reuters estimates, Alphabet, Microsoft, Amazon, and Meta together could spend at least $630 billion this year. Most of that money goes toward data centers and chip purchases.
Demand for AI processors is so strong that companies are trying to lock in supply in advance. Meta is buying millions of chips from Nvidia, signing agreements with AMD, and at the same time developing its own processors. As a result, the market is becoming increasingly interconnected. Large customers take stakes in suppliers, and suppliers invest in clients.
Against this backdrop, the Meta–AMD deal looks not like a one-time purchase, but part of a broader reshaping of the market. Companies are locking in capacity years ahead and are ready to pay tens of billions of dollars to avoid being left without chips. For AMD, this is a chance to secure a stronger position among AI market leaders. For Meta, it is an opportunity to scale its services faster. And for the industry as a whole, it is a signal that the race for AI infrastructure is only accelerating.