Apple reports it utilizes Amazon's personalized AI technology.

• Apple is currently utilizing Amazon Web Services' custom AI chips for services such as searching and will assess whether its latest AI chip can be employed to pre-train its AI models encompassed in Apple Intelligence.
• At its annual AWS Reinvent conference, Apple disclosed its utilization of Amazon's exclusive chips.
• The endorsement of Amazon's cloud service by Apple, which showcased its use of Amazon's chips at its conference, is a strong indication of its competition with Microsoft Azure and Google Cloud for AI spending.

The company is currently utilizing Web Services' custom artificial intelligence chips for services such as search and will assess whether its latest AI chip can be used to pre-train its models like Apple Intelligence.

At the annual AWS Reinvent conference on Tuesday, Apple disclosed its utilization of Amazon's exclusive chips. Benoit Dupin, Apple's senior director of machine learning and AI, spoke about how Apple employs the cloud service. This is a unique instance of the corporation acknowledging a supplier as a client.

Apple's Dupin stated that our relationship is strong, and our infrastructure is both dependable and capable of meeting the needs of our global customer base.

The endorsement of Amazon's cloud service by Apple, which appeared at its conference and adopted its chips, is a strong indication that it is competing with Microsoft Azure and Google Cloud for AI spending. Despite this, Apple also utilizes those cloud services.

Benoit stated that Apple had utilized AWS for more than a decade to provide services such as Siri, Apple Maps, and Apple Music. Additionally, Apple has employed Amazon's Trainium and Graviton chips to improve search services, resulting in a 40% efficiency increase, according to Benoit.

On Tuesday, Amazon made its Trainium2 chip generally available for rent, indicating that its chips are not only cost-effective for inference of AI models compared to x86 central processors from Intel and AMD, but can also be used for AI model development. Benoit also suggested that Apple would use Amazon's Trainium2 chip for pre-training its proprietary models.

Dupin stated that in the initial stages of evaluating Trainium2, we anticipate early results showing up to a 50% increase in efficiency with pre-training.

Apple Intelligence, the AI service used on iPhones, was trained using Google Cloud's TPU chips, as stated in a research paper released earlier this year by Apple.

Nvidia graphics processors are commonly used for AI training, but cloud providers and startups are looking for cheaper alternatives. They are exploring different methods to achieve more efficient processing. Apple's use of custom chips may indicate that non-Nvidia approaches can be effective.

On Tuesday, AWS is predicted to reveal information about renting Nvidia Blackwell-based AI servers.

This fall, Apple released its first major generative AI product, Apple Intelligence, which includes services that can summarize notifications, rewrite emails, and generate new emojis. The product will integrate with OpenAI's ChatGPT later this month, and Siri will receive new abilities to control apps and speak naturally next year.

Apple's AI approach is different from that of other companies. Instead of relying on large clusters of Nvidia-based servers in the cloud, Apple uses its own M-series chips to do as much processing as possible. Then, it sends complicated queries to Apple-operated servers.