AMD and Meta push AI infrastructure to a new scale
AMD and Meta have officially announced a major expansion of their strategic partnership. With this agreement, both companies aim to accelerate the development and deployment of advanced AI models at an unprecedented scale. Most importantly, the deal includes the installation of AMD Instinct GPUs supporting power capacity of up to six gigawatts.
As a result, this collaboration significantly extends their existing relationship. At the same time, it aligns chip, system, and software roadmaps to deliver AI platforms tailored specifically to Meta’s workloads. Consequently, the agreement represents one of the largest AI infrastructure commitments seen so far.
A long-term, multi-generation deployment plan
Under the expanded partnership, Meta and AMD have agreed on a long-term, multi-generation deployment strategy. In practice, Meta will deploy customized AMD Instinct GPUs across its global AI infrastructure. Altogether, these deployments will scale progressively until reaching the six-gigawatt target.
Furthermore, the platform is built on AMD’s rack-scale Helios architecture, which was introduced at the Open Compute Project Global Summit 2025. According to the announcement, shipments for the first gigawatt are expected to begin in the second half of 2026. This phase will rely on custom AMD Instinct GPUs based on the MI450 architecture, optimized specifically for Meta’s AI workloads.
MI450 GPUs, EPYC processors, and Helios architecture
In the initial rollout, the AI platform combines several core AMD technologies. These include custom MI450-based Instinct GPUs, 6th Gen AMD EPYC processors codenamed Venice, and the ROCm software stack. Together, these components form a rack-scale system designed for high performance and energy efficiency.
In addition, the Helios architecture was co-developed by AMD and Meta through the Open Compute Project. Because of this collaboration, the system supports scalable, rack-level AI infrastructure capable of handling rapidly growing workloads. Therefore, Meta gains a flexible foundation that can expand alongside its AI ambitions.
Executive perspectives highlight strategic importance
Dr. Lisa Su, Chair and CEO of AMD, emphasized the scale and significance of the partnership. She stated that the long-term collaboration spans Instinct GPUs, EPYC processors, and rack-scale AI systems. According to her, this alignment enables AMD to deliver highly efficient infrastructure optimized for Meta’s demanding AI workloads while accelerating one of the largest AI deployments in the industry.
Meanwhile, Mark Zuckerberg, Founder and CEO of Meta, highlighted the strategic value of working closely with AMD. He explained that the partnership supports high-performance inference computing and advances Meta’s vision of personal superintelligence. Moreover, he noted that AMD will remain a key partner as Meta continues to diversify and scale its compute infrastructure.
Expanding collaboration beyond GPUs
Beyond graphics accelerators, the partnership also strengthens collaboration around AMD EPYC processors. Over multiple generations, Meta has already deployed millions of EPYC CPUs across its global infrastructure. At the same time, Meta continues to operate AMD Instinct MI300 and MI350 GPUs within its AI systems.
Looking ahead, Meta will serve as a lead customer for upcoming 6th Gen AMD EPYC processors codenamed Venice and Verano. These next-generation CPUs feature workload-specific optimizations designed to deliver strong performance per dollar and per watt. As AI infrastructure grows more complex, processors remain a strategic pillar alongside GPUs.
Performance-based alignment and financial impact
As part of the agreement, AMD has issued performance-based warrants to Meta covering up to 160 million shares of AMD common stock. These warrants are tied to specific milestones related to Instinct GPU shipments. Initially, the first tranche activates with the delivery of the first gigawatt. Subsequently, additional tranches scale with Meta’s orders up to the six-gigawatt level.
According to Jean Hu, Executive Vice President and Chief Financial Officer of AMD, the partnership is expected to drive substantial revenue growth over the coming years. Furthermore, the performance-based structure closely aligns both companies in execution, operational success, and long-term value creation.
A global push for next-generation AI
Overall, the expanded AMD and Meta collaboration spans silicon, systems, and software. Together, both companies aim to power AI infrastructure at a truly global scale. Ultimately, this effort is expected to accelerate AI innovation and deliver AI-driven services and experiences to billions of users worldwide.
AMD will hold a conference call today at 7:30 a.m. CT to discuss www.amd.com the announcement.
