Apple acknowledges training its own Knowledge using Google Tensor technology.

Article Hero Image
Apple’s new artificial intelligence research reveals that the company used Google hardware in order to build the foundations for Apple Intelligence. The research paper “Apple Intelligence Foundation Language Models”, is a technical document that details the sources of the language models at the heart of Apple’s new technology. A quote hidden in the paper suggests that Apple used Google hardware during early development. The paper states that Apple’s Foundation Model and the server technology which drives it were built initially on “v4 or v5p cloud TPU clusters”, using Apple software. The research contains a lot of information about how this was done and what data sources were used to train. CNBC reported on Monday that Apple rented space on Google-hosted clusters. However, the research does not directly support this claim, nor does it mention Google or Nvidia. It’s more likely that Apple purchased the hardware directly from the company and used it in its own data centres. In the end, it doesn’t matter that the model’s initial training was performed on Google’s hardware. Apple is said to be using its own hardware, derived from Apple Silicon, in its data centres to process Apple Intelligence requests. Apple’s alleged “Project ACDC” is intended to optimize AI applications in its data centers. Apple plans to invest over $5 billion in AI server upgrades over the next two-year period. The company wants to match the technological capabilities industry leaders such as Microsoft and Meta, by acquiring tens or thousands of AI servers, probably driven by Project ACDC. Apple has also acquired companies in Canada and France, which both work to compress data used in AI queries into data centers.