The reason behind Meta's partnership with AMD to build a $60 billion AI infrastructure has been revealed.

Meta and Advanced Micro Devices (AMD) have just announced a multi-year partnership agreement to build AI infrastructure for the next generation of artificial intelligence.

 

Under the agreement, AMD will supply AI hardware systems to support Meta's growing workloads, while Meta will expand its large-scale data centers to deploy these systems. The project aims to deploy up to 6GW of computing power from AMD's Instinct GPUs, meeting Meta's growing global AI processing needs.

Mark Zuckerberg stated that this collaboration is an important step in Meta's strategy to diversify its computing infrastructure. He also expects AMD to become a long-term partner for many years to come.

Concerns about an 'AI bubble'

The deal between Meta and AMD is estimated to be worth around $60 billion. According to analyst Alvin Nguyen from Forrester, US technology companies could spend up to $660 billion on AI assets in the near future.

On an industry-wide scale, spending on AI data centers is projected to reach $3 trillion, with participation from tech giants like OpenAI, Google, and leading AI chip manufacturer NVIDIA.

However, this massive investment scale has also raised concerns about the potential for an 'AI bubble'. Some investors worry that if the development of general artificial intelligence (AGI) slows down, companies may struggle to generate returns commensurate with the trillions of dollars invested.

However, there are currently no clear signs that the AI ​​market is slowing down. The market value of many technology companies continues to grow strongly, with NVIDIA once approaching a $5 trillion market capitalization, indicating that the demand for AI infrastructure remains very high.

 

AI at hyperscale

Meta stated that collaborating with AMD on both software and hardware will help the company innovate faster and deploy AI on a large scale.

The first batches of GPUs are expected to be shipped in the second half of 2026. This system will utilize the Helios rack-scale architecture, an open-source hardware platform designed to make data center infrastructure more efficient, flexible, and sustainable.

Lisa Su, President and CEO of AMD, said this partnership will help the two companies align their development paths between Instinct GPUs, EPYC CPUs, and rack-scale AI systems.

According to her, combining these technologies will create a high-performance yet energy-efficient computing infrastructure, while also helping AMD become one of the centers of the global AI infrastructure development wave.

Meta is preparing to invest more than $600 billion by 2028 to build gigawatt-scale data centers in several US states, including Louisiana, Texas, Wisconsin, and Wyoming.

One of the largest projects is Prometheus Hyperscale, a multi-gigawatt AI data center expected to become operational in Ohio in 2026. Additionally, another project called Hyperion could expand capacity to 5 gigawatts in the following years.

In a post on the social media platform Threads, Mark Zuckerberg stated that one of Meta's new data centers will be roughly the size of Manhattan Island, equivalent to about 22.8 square miles.

Meta Compute – Long-term AI Strategy

Meta stated that the agreement with AMD is part of the Meta Compute initiative, a program aimed at expanding technology infrastructure for what the company calls the era of 'personal superintelligence'.

According to Meta, this strategy allows the company to develop new hardware designed in conjunction with internal software systems to handle ever-increasing volumes of data and computation.

Meta also emphasized that the collaboration with AMD will help them continue to innovate in the field of AI and ensure the ability to deliver powerful AI experiences to billions of users worldwide.