‹ Back

Meta Expands AI Compute with 6GW AMD GPU | News

 

Meta Platforms and Advanced Micro Devices have agreed to expand their strategic partnership with a definitive multi-year deal to deploy up to 6 gigawatts of AMD Instinct GPUs in Meta’s AI infrastructure.

 

 

Meta, based in Menlo Park, and AMD, headquartered in Santa Clara, announced on February 24 that the expanded agreement will power Meta’s next generation of artificial intelligence systems across multiple generations of AMD’s accelerated computing products.

 

The deal builds on an existing collaboration between the companies and aligns their product roadmaps across GPU, CPU, systems and software development to support large-scale AI workloads tailored to Meta’s operational requirements.

 

Under terms outlined in the press release from AMD, the first deployment is expected to begin in the second half of 2026, using a custom version of AMD’s Instinct GPU based on the MI450 architecture. These GPUs will be integrated into rack-scale systems developed through the AMD Helios architecture, a platform jointly engineered with Meta under the Open Compute Project.

 

AMD further stated that the systems supporting this deployment will incorporate 6th Generation AMD EPYC CPUs, codenamed “Venice,” running ROCm software, to optimize performance and efficiency.

 

In formal statements accompanying the announcement, AMD Chair and Chief Executive Officer Lisa Su said the expanded collaboration “aligns our roadmaps to deliver high-performance, energy-efficient infrastructure optimized for Meta’s workloads,” and emphasized AMD’s role in supporting what she described as one of the largest AI deployments in the industry.

 

Meta founder and Chief Executive Officer Mark Zuckerberg was quoted saying the long-term partnership with AMD represents a step toward diversifying the company’s compute infrastructure and helping to deploy efficient inference compute for its AI services.

 

The expanded partnership also enhances Meta’s use of AMD EPYC processors, with the company designated as a lead customer for both “Venice” and a next-generation EPYC CPU, “Verano,” designed with workload-specific optimizations for AI deployment.

 

To align strategic interests between the companies, AMD granted Meta performance-based warrants for up to 160 million shares of AMD common stock. Those warrants vest in tranches tied to shipment milestones as Meta’s purchases scale toward the full 6 gigawatts, and vesting also depends on specific stock price thresholds and technical or commercial criteria being met.

 

AMD Chief Financial Officer Jean Hu stated that the structure of the partnership is expected to support substantial multi-year revenue growth and be accretive to non-GAAP earnings per share, reflecting AMD’s long-term financial model.

 

The expanded agreement leverages existing deployments of AMD technology by Meta, which has previously integrated millions of AMD EPYC CPUs and significant installations of AMD Instinct MI300 and MI350 series GPUs throughout its global infrastructure.

 

AMD’s press release noted that the collaboration across silicon, systems and software is intended to enable AI infrastructure deployment at global scale and to accelerate delivery of AI-powered services to billions of end users.

 

The first shipments supporting the initial gigawatt of the 6 gigawatt agreement are scheduled for the latter half of 2026, marking the beginning of the expanded strategic deployment designed to meet Meta’s projected AI compute needs.