Meta has signed a $60 billion deal with Advanced Micro Devices to purchase artificial intelligence chips over the next five years, marking one of the largest AI infrastructure investments despite growing concerns about massive tech spending. The agreement includes Meta acquiring a 10% stake in the California-based semiconductor company, according to industry reports. This partnership represents a strategic shift for the Facebook owner as US tech companies prepare to spend an estimated $660 billion on AI assets this year.
The five-year arrangement will see AMD supply 6 gigawatts worth of chips to Meta, beginning with 1 gigawatt of the company’s upcoming MI450 hardware in the second half of this year, AMD chief executive Lisa Su confirmed. Meta will purchase both flagship graphics processing units and central processing units, including customised variants designed specifically for the social media platform’s requirements.
Diversifying AI Chip Supply Beyond Nvidia
According to Alvin Nguyen, an analyst at Forrester, the deal reflects a broader effort among leading AI companies to diversify their chip suppliers beyond Nvidia, the world’s largest chipmaker. Supply chain bottlenecks at Nvidia have prompted major players to seek alternative sources for AI infrastructure. OpenAI struck a similar partnership with AMD last year, highlighting this industry trend.
Meta has simultaneously negotiated separate agreements with Nvidia to purchase millions of AI chips and has been in discussions with Google about using the company’s tensor processors for AI work, Reuters reported. This multi-vendor approach addresses the limitations of relying on a single supplier as companies scale their AI operations.
Meta’s Strategic Pivot in AI Infrastructure
The AMD deal may signal a shift in Meta’s AI strategy, moving away from direct competition with companies like OpenAI and Anthropic toward becoming a major AI infrastructure provider, Nguyen suggested. Meta invested substantial resources into AI research last year, including a talent spending spree that reportedly offered signing bonuses up to $100 million to attract top employees from competitors.
However, this aggressive recruitment appeared to slow amid widespread fears of an AI bubble. The company now seems to be differentiating its approach by focusing on data centers and hosting AI infrastructure for other organizations. A massive data center under construction in Louisiana, estimated to cost $27 billion, supports this strategic direction.
Customized Solutions and Energy Efficiency
The custom CPU that AMD will develop for Meta will be optimized to deliver high performance while minimizing energy consumption, Su said. The deal encompasses two generations of AMD’s central processors, representing what Su described as “a big bet on AMD” by the social media giant.
Additionally, AMD offers technologies that allow workloads designed for Nvidia systems to be converted for use on its platforms, making the transition more seamless for companies diversifying their chip suppliers. Nguyen noted that AMD’s chips perform as well as, and in some cases better than, competing products, despite the company’s smaller market share.
Multi-Vendor Approach to Meet Growing Demands
Meta plans to continue purchasing chips from multiple vendors while simultaneously developing its in-house processors, according to Santosh Janardhan, the company’s infrastructure head. The scale at which Meta is building data centers and AI infrastructure necessitates working with multiple chip manufacturers and employing various technological approaches.
Meanwhile, the artificial intelligence landscape continues evolving rapidly, with new agentic AI tools causing market volatility and concerns about job displacement. Anthropic announced this week that it will launch a plugin allowing its Claude Cowork tool to integrate more seamlessly into workflows, connecting with Google Drive, Gmail, DocuSign, and other platforms.
Industry observers expect AI companies to continue diversifying their chip suppliers, potentially extending partnerships to other manufacturers such as Intel. The timing and specific details of Meta’s full chip deployment strategy remain to be confirmed as the company implements its multi-billion dollar infrastructure buildout.
