Google to Supply Meta with AI Chips

This deal is part of Meta's broader strategy to diversify its AI chip suppliers beyond Nvidia, which currently dominates the market with an estimated 80-90% share. Meta has also recently announced a multi-billion dollar agreement with AMD to supply AI chips. This diversification helps Meta mitigate risks associated with relying on a single supplier, such as supply chain disruptions and pricing power. The agreement centers on Google's custom-designed Tensor Processing Units (TPUs), which are specialized chips optimized for machine learning workloads. Unlike general-purpose GPUs, TPUs are designed specifically for the matrix operations that are fundamental to neural networks, which can make them more efficient for AI tasks. Google has been developing TPUs for over a decade and uses them to power its own AI services like Search, Photos, and Gemini. For Google, this partnership represents a significant step in commercializing its custom silicon and challenging Nvidia's dominance. While Google has historically used TPUs for its internal needs and offered them through its cloud platform, this deal with Meta signifies a push to supply chips directly to other major tech companies. Meta's investment in its AI infrastructure is substantial, with the company committing between $115 billion and $135 billion in capital expenditures for 2026, a significant increase from $72 billion in 2025. This spending is driven by the massive computational power required for training large-scale AI models and running AI-powered features across its platforms like Facebook, Instagram, and WhatsApp. The collaboration between the two tech giants may extend beyond simply renting the chips. There are reports that Meta is in discussions to purchase TPUs from Google outright for its own data centers as early as next year. This would mark a strategic shift for Google, which has not traditionally sold its custom hardware directly to other large technology firms. Google's latest TPUs, known as Trillium, are the sixth generation of the chip and offer significant performance and efficiency gains over previous versions. The company has even used AI to help design these next-generation chips, accelerating the development process. This continuous innovation is crucial as the demands of AI models continue to grow exponentially.