Apple's technical paper on Monday revealed that two artificial intelligence models of Apple Intelligence are pre-trained on cloud chips designed by Google. A Google chip called TPU (Tensor Processing Units).

Google's TPU tensor processing unit was launched in 2015 and publicly released in 2017. Originally created for internal workloads, tensor processing units now have wider applications. Apple’s statement is tantamount to verifying that top technology companies in the field of artificial intelligence training models are looking for chips to replace Huida GPUs.

Huida's expensive graphics processing units (GPUs) dominate the high-end artificial intelligence training chip market. In the past few years, the demand has been so great that the procurement quantity has been insufficient. Open AI, Microsoft, and Anthropic all use Huida in their models. GPUs, while other technology companies including Google, Meta, Oracle and Tesla are also snapping up Huida chips to build their artificial intelligence systems and products.

Reuters noted that while Apple did not explicitly say it did not use any Nvidia chips, the company was not mentioned in its description. Apple doesn't mention Google or Nvidia in its 47-page technical article, but it does note that its Apple base model AFM and AFM server are trained on Google Cloud TPU clusters. This means Apple rents servers from cloud providers to perform calculations. Apple stated in the article: This system allows us to more efficiently scale up training AFM models, including on-device AFM, AFM servers, and larger models.

Apple uses Google chips to train two artificial intelligence models

Reuters pointed out that Apple's report shows that it will rely on Google chips, rather than graphics chips designed by industry leader Huida, to build the artificial intelligence infrastructure for its upcoming Apple Intelligence.

Apple's decision to rely on Google's cloud infrastructure is noteworthy because Huida holds about 80% of the market and makes the most popular AI graphics processors. Including chips produced by Google, Amazon and other cloud computing companies, all need to cooperate with Huida.

In the research report, Apple did not explicitly say that it does not use Nvidia chips, but its description of artificial intelligence hardware and software did not mention Nvidia hardware equipment.

To train artificial intelligence models, the iPhone maker uses two forms of Google's Tensor Processing Units (TPUs), which are organized into large groups of chips. To build AI models that can run on iPhones and other devices, Apple uses 2,048 TPUv5p chips. Apple deploys 8,192 TPUv4 processors for its server AI models.

Nvidia focuses on GPUs, graphics processing units widely used in artificial intelligence work. Google and Nvidia differ in the way they sell chips and systems as independent products. Google sells access rights to TPUs through Google Cloud Platform. Customers interested in purchasing rights must create software through Google's cloud platform system to use these chips.

Apple is rolling out some Apple Intelligence features to its beta users. Reuters reported on the use of TPU chips in June, but Apple has just published research that reveals its full reliance on Google hardware.

Apple engineers said in the report that using Google's chips could create larger and more complex models than the two models studied in the report. Apple launched a series of new artificial intelligence features at its developer conference in June, including integrating OpenAI’s ChatGPT technology into the software.

Investors are concerned about the future development of Apple Intelligence on iPhone

Apple Intelligence was launched at the 2024 Apple Developer Conference, integrating the personal intelligence systems in iOS 18, iPad OS 18 and macOS Sequoia.

Apple Intelligence composes multiple generative models for users' daily tasks and dynamically learns to publish their current tasks. Built-in base models fine-tune user actions such as writing and polishing text, prioritizing and summarizing notifications, creating interesting images for conversations with family and friends, and simplifying in-app functionality to reduce cross-usage The number of applications.

The new iPhones and Siri functions launched by Apple in recent years have not been satisfactory. The newly introduced Apple Intelligence is only applicable to the iOS 18 version. You must have at least an iPhone 15 Pro or M1 and higher-end versions of MacOS 15 and iPadOS 17 to use it.

Experts point out that Bloomberg reports that Apple has delayed the launch of Apple Intelligence until October at the end of the year, which means it will miss the upcoming iPhone firmware update in September. Developers can only start testing the new artificial intelligence this week at the earliest. Function, investors and Apple fans expect that the new Apple iPhone 16 mobile phone at the end of the year will have Apple Intelligence artificial intelligence enhanced functions, but they may be disappointed.

Apple is slow in developing hardware and software applications, and its relationship with Google is even more difficult to separate. Will Apple Intelligence be a dazzling product that eliminates Siri, or will it be Siri 2.0 that replaces Siri? The iPhone was once an epoch-making cutting-edge product launched by Steve Jobs. With the rapid development of artificial intelligence, the iPhone is gradually no longer the only choice for smartphone users. Apple's new artificial intelligence functions will definitely affect Apple's future status and stock price in the technology industry. .

This article: Apple wants to break up Huida? The new AI system is rumored to be cooperating with Google first appeared on Chain News ABMedia.