Meta plans to deploy Google TPUs in its AI data centres from 2027 and rent TPU capacity from Google Cloud in 2026, marking a strategic shift for Google
Google’s move challenges Nvidia, which holds 80% of the AI GPU market. A share of Meta’s $72B capex could shift significant revenue to Google
Gemini3, built on TPUs, now has 650M users and rivals ChatGPT. The model showcases the power of Google’s hardware and growing AI influence
The Meta-Google deal sent Nvidia shares down 2% and Alphabet up 8%, signaling confidence in Google’s hardware ambitions and market impact
TPUs excel at matrix math for deep learning, while GPUs handle versatile computations. TPUs may be more cost-effective for training large-language models
With Amazon, Microsoft, and others developing in-house chips, Google’s TPU expansion could reshape the AI hardware market and intensify competition
Custom AI hardware like TPUs may drive a sea change, as giants compete and collaborate, shaping the innovation and economics of the AI ecosystem