Apple reportedly entered a multiyear agreement with Google in January 2026 to access Gemini technology, with the deal valued at $1 billion annually. Since July 2024, Apple has used Google TPUs to train its AFM models, including a cluster of 8,192 TPU v4 chips for its first server model and a 2,048-chip group of TPU v5p chips for its on-device model. The choice more than doubled floating point operations per second, tripled high-bandwidth memory and reduced model training time by up to 2.8 times. In June 2026, Apple unveiled a third generation of AFM models developed with Google, while its AFM 3 Cloud Pro model continued to use NVIDIA Blackwell GPUs through Google Cloud.
Source: Read the original article
Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
