Amazon’s (AMZN) cloud services have received its biggest endorsement yet. Apple (AAPL) has confirmed that it is using Amazon Web Services custom artificial intelligence chips. While appearing at the annual AWS Reinvent conference, Apple’s senior director of machine learning, Benoit Dupin, confirmed that they are using the chips for services like search. In addition, they are evaluating the chips for use in Apple Intelligence.
Amazon AI Chips
The confirmation is a strong endorsement for Amazon, which is increasingly looking for ways to strengthen its edge on cloud computing. The e-commerce giant is entangled in fierce competition from Microsoft’s Azure and Google Cloud amid the growth of generative artificial intelligence spending.
Apple has emerged as one of Amazon Web Services’ biggest customers, having used the service for more than a decade for services like Siri, Apple Maps, and Apple Music. In addition, it has used the e-commerce giant’s Inferentia and Graviton chips to serve search services.
Apple AI Investment
The iPhone maker also plans to use Amazon’s Trainium 2 chip to pre-train its proprietary models. Apple’s decision indicates that Amazon’s chips can be used to create new AI models, in addition to being a more affordable option for inferring AI models compared to x86 central processors manufactured by Advanced Micro Devices (AMD) and Intel (INTC). Amazon has confirmed it is possible to rent its Trainium2 chip.
Apple confirmed in a research paper earlier this year that it had trained its iPhone AI service, which it calls Apple Intelligence, using TPU chips from Google Cloud. Most AI training is carried out on expensive Nvidia graphics processors. Startups and cloud providers are rushing to create cheaper alternatives and are investigating various strategies that might result in more efficient processing. Apple’s use of proprietary chips may warn other businesses about the viability of non-Nvidia training methods.
Amazon Nova Family
Meanwhile, Amazon has unveiled a suite of artificial intelligence models, marking an important milestone in the company’s push to rival other tech giants in the fast-growing generative AI sector. The company has unveiled a suite of foundational models to strengthen its position against Microsoft (MSFT), Google, Meta Platform (META) and Open AI.
Six AI models in the Amazon Nova family are capable of producing everything from text to videos. According to the company, the models are faster and at least 75% less expensive than similar offerings that are available on AWS servers.
The original lineup consists of Nova Micro for quick text processing, Nova Lite for simple multimedia tasks and Nova Premiere for sophisticated reasoning, which is scheduled for release in early 2025. With support for 200 languages and the ability to customize the models using customers’ proprietary data, Amazon hopes to draw in businesses creating specialized AI applications.
Until recently, the majority of Amazon’s AI offerings through its AWS cloud service had been restricted to giving users access to models from other businesses, such as Anthropic, an AI startup that it supports. AWS is still the market leader in cloud computing, which is required to power artificial intelligence tools and products, even though Google, Microsoft, and OpenAI have taken the lead in AI.
