Alphabet is intensifying its challenge to Nvidia's dominance in the AI chip market by offering its Tensor Processing Units (TPUs) to third parties. Meta is reportedly in discussions with Google to integrate TPUs into its data centres starting in 2027 and may also rent TPU capacity from Google Cloud next year. This shift marks a departure from Google's previous strategy of solely using TPUs for its own cloud infrastructure and renting access to others.
Google is pitching on-premise TPU deployment as a way for companies to maintain greater control over sensitive data and meet strict regulatory requirements. Google Cloud executives believe that increased TPU adoption could capture a significant portion of Nvidia's annual revenue, potentially reaching 10%.
News of Google's intensified AI chip rivalry has already impacted the stock market. Alphabet shares have risen, while Nvidia and AMD have experienced declines. Google's TPUs are reportedly cheaper than Nvidia's chips and were used to train its advanced Gemini 3 AI model. Meta currently relies heavily on Nvidia GPUs for its AI systems.
